Abstract
People increasingly live their lives online, which means that their identities are increasingly constituted by and displayed through their activities on digital platforms. Existing theorizing about the psychology of digital identity has emphasized the social roles that people perceive and aim to verify online. These accounts can explain how digital identities are shaped by the social environment but not how they come together to create social life online. Moreover, there are unique features of digital identities, imposed by the digital platforms on which they are enacted, that cannot be accounted for by existing theories of offline social identity. To address these limitations, in the current article we propose a social digital identity theory that outlines how a person’s digital identity is shaped by their online and offline group memberships, as well as the implications of this psychological process for the well-being and performance of both individuals and groups. In outlining this theory, we aim to extend theorizing around social identity and digital identity by integrating these fields within a framework that recognizes and helps us better understand the merging of online and offline life.
The online world is not identical to the offline, but it is entirely real. (Golder & Macy, 2014, p. 143) There is a real opportunity to connect more of us with groups that will be meaningful social infrastructure in our lives. —Mark Zuckerberg, open letter to Facebook community, February 17, 2017
We live in a world in which the majority of people (4.9 billion as of 2023) have a social media account, on average using six to seven different digital platforms (Wong, 2023). Moreover, many people choose or are required to use digital technology to access essential services such as welfare, banking, or health care (Anand, 2021; Castelo et al., 2015; A. Rieger et al., 2022). This means that, for most people, answering the question “Who am I?” will involve some consideration of their online activities. We refer to these parts of a person’s identity that come from their use of digital platforms as their “digital identity.” Put simply, digital identity is “who a person is” online. For example, a person’s digital identity may be constituted by their use of social media profiles, logging in to a digital portal used to access government services, or even choosing what music to listen to on a streaming service. Digital identity may correspond directly to offline identity (e.g., when accessing services that involve identity verification), may be very different (e.g., when embodying an avatar for a role-playing game), or anywhere in between (e.g., when using a platform such as Facebook that involves some level of verification but allows for some degree of flexibility in self-presentation). Importantly, as with offline identity (Roccas & Brewer, 2002; Turner & Oakes, 1997), digital identity is socially structured (Boyd & Ellison, 2007; Ellis, 2020; Kavakci & Kraeplin, 2017) and multifaceted (V. Cho & Jimerson, 2017; Fieseler et al., 2015; Ollier-Malaterre et al., 2013).
Digital identity is as “real” in psychological terms as offline identity (Golder & Macy, 2014). For example, using social media can increase (Gentile et al., 2012; Gonzales & Hancock, 2011) or decrease (Kalpidou et al., 2011; Mehdizadeh, 2010) self-esteem, people with more connections on social media have higher self-efficacy (Kramer & Lewicki, 2010), and social media authenticity is both a predictor and outcome of affective well-being (Reinecke & Trepte, 2014). Online behavior is also shaped by offline identity. For example, people with low self-esteem are eager to self-disclose on social media (Forest & Wood, 2012) and use fewer privacy settings on Facebook (Christofides et al., 2009). Moreover, people are motivated to look at their own social media accounts after a blow to their self-esteem (Toma & Hancock, 2013), in part because it can affirm social connections. Furthermore, people from vulnerable groups (e.g., refugees) are sometimes reluctant to enroll in digital identity systems that require them to disclose personal information because of fear of persecution (Shoemaker et al., 2019). In other words, people act online and offline in ways that indicate that both kinds of existence contribute to their sense of self and its expression.
However, there are important differences between social life online and offline that have implications for the way identity is shaped by, and enacted in, the social environment. One unique feature of online social interaction is that it can operate over a different time scale to in-person interaction, with asynchronous communication potentially taking place over weeks, months, or years (H. Li et al., 2021). At the same time, online conversations can leave a permanent digital footprint, potentially being accessed at some point in the future by unknown third parties (Davis, 2016; Erickson, 1999). Another novel feature of online social interaction is that it can be anonymous or pseudonymous depending on the specific platform used (Golder & Macy, 2014). Furthermore, social networks on online platforms can be symmetric (as with Facebook or LinkedIn, where social connection must be mutually agreed on) or asymmetric (as with X, formerly Twitter, where social connection can be one-way; Paul & Friginal, 2019). Moreover, social connections and interactions are increasingly mediated through algorithms that infer and utilize information about people’s identities (C. H. Smith, 2020). Accordingly, it is not simply the case that digital identity is more flexible or controllable than offline identity (Davis, 2016) but rather that self-perception and self-expression are shaped by the features of the specific digital platforms that people use (Laurent et al., 2015).
Existing Theories of Digital Identity
Digital identity, like identity more broadly, can be analyzed from a number of different perspectives (Feher, 2015). For example, some approaches to understanding digital identity have conceptualized it without reference to people’s subjective experience. In particular, there is an important body of research that has evaluated digital identity from a data-management perspective—that is, as constituted by people’s objective digital footprints (Beck, 2015; Camp, 2004; Grassi et al., 2017). Indeed, from this perspective any entity can have a digital identity, not just a person (Grassi et al., 2017). This work is important for understanding issues such as online privacy, cybersecurity, and effective system architecture. However, in the current article we approach digital identity from a social-science perspective in which people’s subjective experience is the central phenomenon to be understood. In this regard, Stryker and Burke (2000) identified three distinct uses of the term “identity” within the analytic framework of social science. The first approach seeks to understand identity by analyzing the culture of a people, that is, identity as ethnicity (e.g., Calhoun, 1994). The second approach sees identity in terms of common identification with a collective or social category (also described as the “social cognition” approach; see Howard, 2000). A third way of understanding identity is the “symbolic interactionist” approach (Howard, 2000), in which identity is understood as the “parts of a self composed of the meanings that persons attach to the multiple roles they typically play in highly differentiated contemporary societies” (Stryker & Burke, 2000, p. 284). All three of these perspectives on digital identity are important to consider. However, theorizing and research concerning digital identity has been for the most part dominated by the symbolic interactionist approach.
A key theory within the symbolic interactionist paradigm is identity theory (Stryker & Burke, 2000). This theory outlines how people internalize social structures into their sense of self, which in turn affects how they read social situations and perform on this basis. According to identity theory, people perceive standards associated with different identities and are motivated to live up to these standards, acting to verify or confirm their identity (Davis, 2016). When a person is unable to verify a salient identity, this produces negative affect and motivates them to change their behavior (or their construal of the current situation) to allow for verification (Stets & Burke, 2005, 2014). Because identity theory focuses on identity standards and verification, it has been applied primarily to understand how people’s psychology, behavior, and well-being are shaped by the social roles that are demanded of them (Stryker & Burke, 2000). Furthermore, identity theory emphasizes the importance of commitment to others in shaping the salience of identities and the activation of associated roles such that being highly committed to a particular identity increases the salience of that identity, making the activation of roles associated with that identity more likely (Burke & Reitzes, 1991; Stryker, 2001). In other words, identity theory accounts for the ways in which social structures shape identity and how identity motivates cognition and action to maintain a positive self-concept (Stryker & Burke, 2000).
There is an extensive body of research that has used a symbolic interactionist approach to understand digital identity, in particular how online contexts shape self-presentation demands. For example, Cho and Jimerson (2017) found that school administrators who used social media platforms had to negotiate competing roles demanded by their identities as instructional leaders and school public-relations facilitators, which led to behaviors such as compartmentalization and self-censorship. Similarly, Kavakci and Kraeplin (2017) investigated how “hijabistas” (Muslim women who work as online influencers for a broad audience while adhering to the rules governing apparel that coincide with Islamic dress code) navigate these potentially conflicting demands in their online activities. Other research has looked at how people manage professional versus personal identities online (Fieseler et al., 2015; Ollier-Malaterre et al., 2013). Furthermore, numerous authors have investigated the relationship between the online and offline self from a self-presentation perspective (Baker, 2009; Boellstorff, 2015; Hogan, 2010; Vaast, 2007; Waggoner, 2009).
Several authors have drawn on this empirical research in proposing theoretical accounts of digital identity. For example, Bullingham and Vasconcelos (2013) conceptualized digital identity from a self-presentation perspective, using Goffman’s (1959) framework of identity and self-presentation to analyze the relationship between the online and offline self. From this perspective, digital identity is played out on a public stage, with offline identity being the “backstage” that allows for true self-presentation (particularly when the online context is anonymous or pseudonymous). Bullingham and Vasconcelos’s empirical findings show that, when given the choice, people often recreate curated versions of their offline identities online.
A more structured account was provided by Davis (2016), who outlined an agenda for an identity theory of digital identity but stopped short of formalizing this in terms of specific propositions. According to Davis (2016), social media accounts are “proximate social structures” that structure interactions within social networks and “call forth specific role, group, and person identities” (p. 141). Social media also sometimes causes “context collapse” (Marwick & Boyd, 2011), in which multiple (potentially conflicting) identities are demanded simultaneously by the situation, leading to behavioral expectations that may be contradictory (Davis & Jurgenson, 2014). One strategy to deal with context collapse is a “lowest common denominator” approach, such as only posting content on social media that is acceptable to the most sensitive members of the network (Hogan, 2010). These theoretical frameworks help to explain and predict how people will act in accordance with relevant social roles online.
Another way to understand identity is the social cognition approach, which focuses primarily on how identity is shaped by belonging to groups and other social categories (Howard, 2000; Stryker & Burke, 2000). This perspective is best represented by work informed by the social identity approach (Tajfel & Turner, 1979; Turner et al., 1987). There are similarities between the social identity approach and identity theory—for example, they both emphasize the importance of the social environment in shaping identity. However, prominent theorists from both traditions acknowledge key differences (Hogg et al., 1995; Stets & Burke, 2000). In particular, the social identity approach focuses more on how the immediate social environment shapes a person’s identity in a dynamic sense and on psychological group membership rather than interpersonal ties and social structures. The ability of the social identity approach to explain not only how the social environment shapes individual identity but also how identity facilitates social life by making collective psychology possible (Turner, 1982) means that it is well placed both (a) to account for the effects of online group membership on individuals and (b) to explain phenomena such as collective online behavior (Hopthrow et al., 2020; Thomas et al., 2015). Indeed, a social identity approach to digital identity is not incompatible with the empirical evidence or existing theoretical frameworks that follow an identity theory tradition. Rather, such an approach has the capacity to unlock the potential of digital identity to explain a wider range of social phenomena.
The social identity approach has frequently been used to understand how people think, feel, and behave as group members (Bentley et al., 2017; Kuppens & Yzerbyt, 2012; Tajfel & Turner, 1979; Van Bavel & Cunningham, 2012; Voci, 2006); how a person’s previous experience and expectations interact with features of their social environment to make different identities more or less salient (Oakes et al., 1991; Turner et al., 1994); and the implications of these psychological processes for personal and collective functioning (Greenaway et al., 2015; C. Haslam et al., 2018; S. A. Haslam, 2004; Jetten et al., 2012; van Knippenberg, 2000). However, the vast majority of social identity research and theorizing concerns offline identities. As a result, there is currently no formalized, comprehensive account of how the unique features of digital social interaction might shape people’s social identities and, through such an account, their collective psychology. Moreover, existing social identity accounts, such as those that have been applied to disparate research areas such as leadership (S. A. Haslam et al., 2020; Steffens et al., 2020), conformity (S. A. Haslam et al., 2014; S. Reicher & Haslam, 2006), trust (Cruwys et al., 2021; Voci, 2006), well-being (S. A. Haslam et al., 2022), and political partisanship (Van Bavel & Pereira, 2018) do not differentiate between online and offline identities. As a result, it is assumed that the online context does not uniquely shape social identity processes and outcomes in these areas. However, as previously discussed, there are several features of the digital world that are likely to affect how social interaction unfolds and is experienced. Accordingly, just as digital identity research lacks a social identity perspective, the social identity approach likewise lacks an account of digital identity.
Despite the lack of a formalized social identity theory (SIT) of digital identity, some researchers have applied this theoretical approach to study specific issues surrounding digital identity. For example, in a study of massively multiplayer online players, Kaye et al. (2017) found that gamer identity was associated with increased self-esteem and social competence, as well as reduced loneliness. These effects align with those found for offline social identities (Bat-Chava, 1994; Cooper et al., 2017; C. Haslam et al., 2019). Similar results were found for players of online football games (Kaye et al., 2019), particularly when these games also fostered offline relationships. Moreover, online social identity has been found to determine whether online social comparison translates into benign or malicious envy, enhancing the former and reducing the latter because online identification makes social comparison less threatening (Latif et al., 2021). However, online identity can also have negative consequences, as shown by Pegg et al. (2018), who found that Australian adolescents who identified strongly with their online social network showed a stronger relationship between exposure to alcohol-related online content and actual alcohol use. The authors proposed that this effect resulted from more highly identified adolescents aligning their behavior with psychologically relevant group norms regarding alcohol use. Similarly, Ridout et al. (2012) found that university students who posted more alcohol-related identity content on social media (i.e., photos of alcohol consumption, logos of alcohol brands, textual posts about alcohol-related interests) were more likely to engage in problematic alcohol use.
Moreover, there are several theoretical accounts that use the social identity approach to understand a particular aspect of digital identity. Lüders et al. (2022) discussed how collective selfhood, formed via social identity, is shaped by the affordances of social media platforms. Specifically, social media platforms facilitate and enhance large-scale community building based on shared attitudes through features such as user tags and hashtags that can be used to signify social identities. Social media platforms further shape social identity by allowing for group discussions involving direct interactions, thereby allowing groups to negotiate the content of their shared identity. Furthermore, “affective affordances” such as emojis contribute to social identity formation by facilitating shared emotions. Relatedly, Code and Zaparyniuk (2010) argued that a social identity perspective can be useful for understanding how online groups form and develop. Together, these various articles constitute an initial exploration of how social identity processes might operate in an online context.
However, several key questions remain unanswered by these theoretical accounts that have focused primarily on group formation. In particular, existing accounts of digital identity (whether following identity theory or SIT) do not spell out (a) what it means psychologically to see oneself in terms of social rather than personal identity in an online context; (b) how different social identities might be made more or less salient by features of the online social environment, including technical aspects of different digital platforms; (c) the dynamics of the relationship between online and offline identity; or (d) the implications of identity salience for the performance and well-being of both individuals and groups. In the current article we aim to address these questions by outlining a formalized, comprehensive theory integrating social identity and digital identity: social digital identity theory (SDIT). We outline SDIT from the “ground up,” starting by explaining the social identity approach, applying this to the online context, and making predictions relating to online and offline identity on the basis of this analysis.
Social Digital Identity Theory
SDIT is constituted by formal propositions concerning social identity processes (personal vs. social identity, online vs. offline identity), individual outcomes, intragroup outcomes, intergroup outcomes, and contextual factors. We outline each of these propositions in turn in the following sections.
Personal versus social identity
The social identity approach that we draw on for SDIT comprises two closely related theories: SIT (Tajfel & Turner, 1979) and self-categorization theory (SCT; Turner et al., 1987). Both of these theories posit that in addition to being able to define themselves in terms of their unique personal identity as individuals (i.e., as “I” and “me”), a person can also define themselves in terms of social identities that they share with other members of a relevant in-group (i.e., as “we” and “us”). And the more that they identify with any given group, the more likely they are to use this as a basis for self-definition. In addition, the more that a person identifies with a given group (e.g., as “us Catholics” or “us Australians”), the more they will perceive, think, feel, and act as a group member rather than as an individual. For example, a person who identifies with a group is more likely to perceive information about that group as self-relevant (Bentley et al., 2017; Coppin et al., 2016; Xiao et al., 2016); to hold attitudes that reflect group norms (Hogg & Reid, 2006; Terry & Hogg, 1996); to feel group-based emotions such as collective guilt, anger, or pride (Iyer & Leach, 2008; Kuppens & Yzerbyt, 2012; Yzerbyt et al., 2003); and to act in line with group goals (Ellemers et al., 2004) as well as on behalf of other group members (Levine et al., 2005; Levine & Manning, 2013).
SDIT predicts that these same psychological effects will occur in an online context. For example, it predicts that a person who identifies as a member of a given group will be more attentive and receptive to information provided by people who are perceived to be in-group members rather than out-group members. This is relevant for processes such as staying safe from cybersecurity threats—for example, our theory predicts that in-group members will be trusted both as sources of information and as trustworthy agents (for better or worse). The theory also predicts that in online contexts people will internalize the emotions and attitudes of other members of groups that they identify with (i.e., in-group members) in ways that affect group functioning (Massa, 2017; Neff et al., 2013) but that can also fuel misinformation (Pereira et al., 2023) and polarization (Van Bavel et al., 2021). Furthermore, these thoughts and feelings are predicted to translate into action such that people will be motivated to respond more positively to in-group members than to out-group members. Among other things, then, they are more likely to engage in acts of group-based kindness toward in-group members at the same time that they engage in identity-based trolling of out-group members (Nekmat & Lee, 2018; Ortiz, 2020). These ideas can be formalized in the following propositions:
P1: When a person sees themselves as a group member, their digital identity (who they are online) will be defined by social identity (their sense of “we” and “us”) rather than by personal identity (“I” and “me”).
P2: When a person’s digital identity is defined by social identity, they will perceive, think, feel, and behave online as a group member rather than as an individual.
People not only define themselves along a continuum from “I” to “we” but also belong to many different groups and social categories (Roccas & Brewer, 2002). But which of these social identities will determine a person’s digital identity in a given online context? Addressing this question, SCT expands on social identity theory by accounting for the way in which social identities become more or less salient (i.e., psychologically active) in different social contexts through a cognitive process of self-categorization. Specifically, SCT posits that a person will self-categorize in terms of a given identity (a) when their previous experiences and current psychological state predispose them to think of themselves as a member of that social category (perceiver readiness); (b) when there is a low level of perceived difference between that person and other category members in the current social environment, as well as a high level of perceived difference between that person and other people who are members of other comparison categories (comparative fit); and (c) when the content of those perceived differences is aligned with category expectations (normative fit). It is important to note that a person can self-categorize as an individual or group member or as belonging to a broader social category (e.g., “Black,” “female”) because all these kinds of social categories can be self-defining (Onorato & Turner, 2004; Skorich et al., 2021). These insights can be formalized as follows:
P3: The extent to which a person will define their digital identity in terms of a particular personal or social identity in a given context will be determined by features of both that person and their social environment. Specifically, a person will be more likely to define themselves in terms of a given social identity when (P3a) their previous experiences and current psychological state predispose them to do so (the principle of perceiver readiness), (P3b) the differences between themselves and other in-group members are perceived to be smaller than the differences between themselves and comparison out-group members (the principle of comparative fit), or (P3c) the content of the perceived category differences is aligned with category expectations (the principle of normative fit).
In SCT, perceiver readiness (P3a) is considered to arise from a combination of what perceivers know about the situation from past experience and knowledge, what they expect in the current situation, and their motives, goals, and needs (Reynolds et al., 2003, p. 292). In this regard, a person’s readiness to self-categorize in terms of a particular social identity is often conceptualized in terms of psychological identification with that category—“the centrality and evaluative importance of a group membership in self-definition” (Turner et al., 1987, p. 55). Emerging research suggests that people increasingly use the internet and other technologies to outsource or “externalize” cognition. For example, reliable access to online information leads to less factual information being stored in memory, with cognitive resources being redistributed to accessing this external information (i.e., the “Google effect”; Gong & Yang, 2024; Heersmink, 2016; Sparrow et al., 2011). This can be thought of as a form of extended cognition (Clark & Chalmers, 1998) and, more specifically, part of the web-extended mind: “an extended cognitive system whose processes supervene on a set of constituent material elements that includes one or more Web resources” (Smart, 2017, p. 362). We propose that identification with a social category (and thus perceiver readiness) may also be externalized, both online and offline, when cues to the centrality and evaluative importance of that identity are encoded in the physical or virtual environment. Specifically, this encoding would need to fulfill Clark’s (2010) criteria for external resources constituting extended cognition, namely that (a) “the resource be reliably available and typically invoked” (availability criterion), (b) “any information thus retrieved [from the resource] . . . should be deemed about as trustworthy as something retrieved clearly from biological memory” (trust criterion), and (c) “information contained in the resource should be easily accessible as and when required” (accessibility criterion; p. 46). These ideas can be formalized as follows:
P4: Perceiver readiness is “externalized” when information about the centrality and evaluative importance of a social category membership is stored in a person’s physical or digital environment rather than in their biological memory. The extent to which this information will be externalized depends on whether (P4a) external information about the centrality and evaluative importance of a social category membership is reliably available and typically invoked (availability criterion), (P4b) this external information is deemed to be about as trustworthy as information stored in biological memory (trust criterion), and (P4c) this external information is easily accessible as and when required (accessibility criterion).
To illustrate how this concept of “externalization” differs from (internalized) identification and contextual fit, imagine a person who finds themselves in what would clearly be considered an intergroup context, such as a sporting contest with two teams in which the person is temporarily playing for the other team rather than their usual team (perhaps because of an insufficient number of players for a fair competition). Despite current similarities with the other team and differences from the usual team (e.g., in terms of one’s current jersey and one’s current purpose) and an understanding that there is a normative context to these differences (the game requires you to play for the other team), there is a “stickiness” to the person’s usual team caused by ongoing, internalized identification that should lead the person to have mixed loyalties. In this context, externalization could be the presence of the person’s own name written on the clubhouse championship board, or the sight of a favorite seat overlooking the field, or any other external information that tells the person “you belong to your usual team.” All of these factors (i.e., externalized and internalized perceiver readiness, comparative and normative fit) together determine the identity that will be most salient in the present context.
Another variable that we propose will shape social identity salience both online and offline is embodiment. An emerging body of research has found that “sensorimotor experiences selectively evoke particular psychological content” (O’Connor, 2017, p. 5). For example, anger-related conceptual knowledge is more available in heated environments, and anger-related primes produce higher estimates of temperature (Wilkowski et al., 2009). Similarly, people feel that the room temperature is colder after experiencing social rejection (Zhong & Leonardelli, 2008) and experience keeping a secret as a physical burden (Slepian et al., 2012). Physical actions such as clenching one’s fist (Schubert, 2004) or using smiling muscles (Soussignan, 2002) activate analogously related concepts and experiences. Building on this research, we propose that one way in which sensorimotor experiences might shape social identity salience is via linguistic metaphors (Gibbs, 2005; Lakoff, 2012; Landau et al., 2010) such that these experiences make metaphorically appropriate identities more salient. In other words, if particular sensorimotor experiences are associated with certain kinds of social relations, their presence should contribute to a consideration of which social categories provide the best “fit” to the situation. For example, a high-status group identity may be more likely to become salient when a person is in a physically high (vs. low) position; or a positively valued group identity may become more likely to be salient when one is using smiling muscles (vs. frowning muscles).
Another way in which embodiment might shape salience is through shared embodiment. This can be understood as physical bodies contributing to the fit of stimuli with social categories, whether this is normative fit (e.g., “I perceive that bodies, including my own, are acting and reacting in ways that match my understanding of a particular set of social categories”) or comparative fit (e.g., “The way bodies, including my own, are acting and reacting make them appear to form two distinct groups pursuing different ends”). Along these lines, physical actions and experiences such as synchronous action (Good et al., 2017; Wiltermuth & Heath, 2009), coordinated action (Krishna & Götz, 2024), and shared emotions (Livingstone et al., 2011; Páez et al., 2015) have been found to increase shared social identity. Moreover, physical copresence can promote physiological synchrony and, through this, shared social identity (Baranowski-Pinto et al., 2022). Similarly, cooperative action can create a sense of shared agency (“we did it”; Bolt et al., 2016).
We also propose that embodiment is likely to shape online as well as offline social identity. The idea that online identity is entirely “disembodied” has been criticized (e.g., Wynn & Katz, 1997). For example, researchers have found that people’s physical bodies and their goals shape their choice of online avatars (Freeman & Maloney, 2021; Park & Kim, 2022; Vasalou et al., 2008). Moreover, the kinds of virtual bodies inhabited online also affect experiences and behavior both online and offline, a finding known as the “Proteus effect” (Y. Liu, 2025; Praetorius & Görlich, 2020; Yee & Bailenson, 2007). Furthermore, these avatars are sometimes used to perform, “physically,” who a user wants to be offline (Schultze, 2014). Offline social identities such as gender also persist in virtual reality (VR) contexts, and these identities moderate the extent to which virtual avatars are experienced as embodied (Peck & Good, 2024). Accordingly, rather than digital identity being inherently disembodied, it is better to think of embodiment as a spectrum that exists both online and offline (Crone & Kallen, 2024). Supporting this idea, the degree of disembodiment in online chats negatively predicts well-being outcomes (Kang, 2007). Moreover, building on our previous discussion of how shared embodiment might shape social identity, we also propose here that shared disembodiment might also shape social identity such that this shared experience may intensify or even be sufficient for a feeling of shared group membership.
Accordingly, building on this recent research into embodied cognition (Shapiro & Spaulding, 2024), we propose that thinking and perceiving (offline and online) are both enabled and constrained by physical bodies and their environments. Drawing on the literature discussed above, we make the following predictions regarding embodiment and social identity salience:
P5: Social identity salience will be shaped by embodiment. Specifically, (P5a) a person’s sensorimotor experiences will make metaphorically appropriate self-categories more likely to become salient, (P5b) sensorimotor experiences will contribute to comparative fit by shaping perceptions of differences between people, and (P5c) sensorimotor experiences will contribute to normative fit in terms of how well these experiences match category expectations.
There are several unique features of digital platforms that are likely to shape how these self-categorization processes will function in relation to digital identity specifically. In particular, the potential for anonymous and pseudonymous interactions online (Golder & Macy, 2014), the symmetry or asymmetry of online social engagement (Boyd & Ellison, 2007), the extent to which interactions are private versus publicly accessible (Bingley et al., 2021), the extent to which the platforms use algorithms that treat people differently on the basis of their group membership (Bingley et al., 2023), the permanence of digital footprints (Davis, 2016; Erickson, 1999), and the extent to which the user’s (and other users’) engagement with the platform is embodied (Gerhard et al., 2004)—as well as the nature of this embodiment (Dourish, 2001)—are factors that are predicted to shape digital self-categorization.
Some digital platforms allow for anonymous or pseudonymous interactions (e.g., in which interactions are tied to a username; D. Cho et al., 2012; Tsikerdekis, 2012). A social identity perspective predicts that this anonymity and pseudonymity will help to determine identity salience. Specifically, as outlined by the social identity model of deindividuation effects (SIDE; Postmes et al., 1998; Reicher et al., 1995), anonymity should make personal identity less salient while making available group identities more salient by comparison. In other words, when cues to a person’s individual identity are lost (as in the case of anonymity), cues to social identity become more relevant to understanding oneself in the present context. This means that a person who is anonymous is less likely to see themselves as an individual, but at the same time they are more likely to see themselves in terms of whatever social identity is most salient (e.g. “us Redditors” rather than “me as an individual”). The key point here is that the self is not simply “lost” in the anonymous crowd—rather, it is more likely to be “found” in terms of social identity. Because a pseudonym can either identify a person as an individual or as a group member (such as when online gaming usernames highlight personal individuality or clan membership, respectively), pseudonymous online interactions are predicted to make either personal or social identity more salient depending on the extent to which pseudonyms and the digital platform more broadly convey cues relating to these identities (Spears & Postmes, 2015).
The structure of online social networks may also shape social identity salience. Some social media platforms require symmetric connections, where social distance between users is mutually equivalent (e.g., with Facebook friends or LinkedIn connections), whereas others allow for asymmetric connections, where social distance between individuals is variable or unclear (e.g., X followers; Paul & Friginal, 2019). We propose that because symmetric social networks are formed through interpersonal ties, symmetric digital networks will tend to make personal identity more salient (i.e., people will see themselves more as individuals interacting with other individuals rather than as group members interacting in an intergroup context). Moreover, groups formed and maintained on symmetric digital networks will tend to consist of overlapping interpersonal ties rather than shared social categorization (Ren et al., 2012) and are therefore more likely to lack the key ingredient of shared social identity that transforms a series of interconnected relationships into a psychologically meaningful group (Turner, 1982; Turner & Oakes, 1997). In contrast, asymmetric networks have an enhanced potential to facilitate broader social identities such as those based on common interest to the extent that cues to these are available (in a similar way to anonymity).
Likewise, when communication on a platform is interpersonal by design (allowing for individuals to interact only with other individuals), engaging with such platforms is predicted to make personal identity more salient compared with platforms that facilitate collective interactions (e.g., “groups” on social network sites). Furthermore, when group communication online is structured as a series of interpersonal interactions combined together (as in a group video call in which typically only one person can speak at a time, participants’ bodies are hidden, and there is no sense of spatial orientation), the formation of shared social identity will be impaired relative to a platform that affords elements of collective communication such as nonverbal cues and turn-taking (Nguyen & Canny, 2007). However, when communication is more collective by design there will be more potential for shared social identities to be made salient (as in the case of X’s mission to create a “global town square,” facilitated by features such as hashtags that allow for large-scale discussions; Burgess, 2022).
Patterns of collective secrecy afforded by online platforms also have the potential to shape social identities. For example, some social media platforms (e.g., Facebook, Instagram) allow for private groups or chat rooms. Moreover, in the case of platforms such as WhatsApp and Telegram, extra layers of privacy are provided via encryption. In contrast, other social media platforms such as Reddit operate as public forums. The social identity theory of information access regulation (Bingley et al., 2021) predicts that preventing outsiders from accessing information (e.g., through collective privacy or secrecy) will increase shared social identity among those given access. In part this is because information access regulation accentuates the similarity between those who have access as well as the difference from those who do not have access (i.e., through comparative fit; P3b). Accordingly, digital platforms that allow groups to maintain privacy, secrecy, or confidentiality are likely to foster the social identity of those groups.
Another feature of digital platforms that may shape identity salience is the extent to which these platforms use algorithms that treat people differently on the basis of their group membership. Artificial intelligence (AI), and particularly AI based on neural networks and large language models, can be used by digital platforms to classify users based on the social groups that they belong to (AL-Qawasmeh et al., 2022; Conover et al., 2011; Gichoya et al., 2022). For example, algorithmic recidivism systems (Isaac, 2017; Lum & Isaac, 2016; Richardson et al., 2019) and health-care resource allocation systems (Obermeyer et al., 2019) have been found to treat Black people differently from White people in the United States. This classification is not necessarily intentional on the part of the system designers because cues to group membership are embedded in, and can be inextricable from, other social data (Gichoya et al., 2022; Matz et al., 2019, 2020). Moreover, some automated systems work better for people from particular groups and worse for others, such as airport security screening systems (Costanza-Chock, 2018), facial recognition systems (Buolamwini & Gebru, 2018), and natural language processing systems (Blodgett & O’Connor, 2017)—often because these groups are either not well represented in the data that were used to train the system (Eisenstein, 2013; Hovy & Søgaard, 2015) or not sufficiently considered during system design. According to the social self-determination model (SSDM) of AI system impact (Bingley et al., 2023), when such systems treat users or other people as group members (i.e., as systematically different from others on the basis of their group membership), those group memberships are predicted to become salient through the SCT principles of fit. Similarly, we predict that identity will be more likely to become salient when a digital platform treats a user as a group member (either intentionally or unintentionally).
Digital footprints are “the trails and artifacts that people leave behind when interacting in a digital setting,” such as online click data, credit card transactions, emails, and social media content (Haimson et al., 2016, p. 2895). These digital footprints can be persistent or permanent (Davis, 2016; Erickson, 1999) and often encode information about not only personality (Lambiotte & Kosinski, 2014; Vazire & Gosling, 2004) but also group membership, as demonstrated by a systematic review that found that demographic details such as age, gender, and political orientation can be inferred from these footprints (Hinds & Joinson, 2018). This kind of information can shape a person’s social identity when interacting with a digital platform. For example, a study of trans peoples’ experiences with identity change online found that persistent digital footprints in the form of social media photos and comments, profile names, gender markers, friendship networks, and accounts often posed a challenge to this process, giving “voice to a past that a person wishes to forget” (Haimson et al., 2016, p. 2903). Accordingly, we propose that persistent digital footprints can constitute a form of externalization (P4), acting as a kind of perceiver readiness that is activated whenever the user is confronted with these footprints while using a digital system. Following from P4, this externalization should make social identities associated with digital footprints (e.g., through forum posts, instant messaging histories, or social media content) more salient when the user is engaging with the system on which these footprints are stored.
As discussed in relation to P5, we predict that social identity salience will be shaped by embodiment in several ways—through linguistic metaphor, through shared patterns of experience, and through expectations about the kinds of experiences associated with particular identities. Building on these ideas, we further predict that features of digital platforms will shape social identity salience through embodiment. Interactions with digital systems involve various degrees (Gerhard et al., 2004) and kinds (Dourish, 2001) of embodiment. For example, interaction with a digital system might involve the use of gestures (Kopp & Wachsmuth, 2010; Uba & Jurewicz, 2024), affective touch (van Erp & Toet, 2015), or speech (Derboven et al., 2014). These modalities can be combined, such as in immersive VR interfaces (Fuchs et al., 2011; Guttentag, 2010; Hudson et al., 2019). Moreover, interactions with digital systems involving intelligent virtual agents can involve embodiment in terms of the artificial agent itself, the human, and objects and actions in the environment (Pustejovsky & Krishnaswamy, 2021). Interactions with other people using digital systems also have varying degrees and kinds of embodiment, ranging from immersive, copresent VR interactions (Rogers et al., 2022) to video teleconferences with varying degrees of physical reality (Kauff & Schreer, 2002; O’Conaill et al., 1993).
We make two predictions about digital system embodiment and social identity salience. First, sensorimotor experiences in the context of system interaction will provide information for metaphor-based identity salience (following P5a). For example, social interactions through VR systems that increase one’s perceived height from the ground and/or other users would be predicted to make identities that subjectively fit this embodied experience more salient—such as high-status identities. Along these lines, research has found that using an older avatar in VR affects walking speed offline (Reinhard et al., 2020), that is, a Proteus effect based on social category information, operating similarly to classic priming studies in social psychology (Banfield et al., 2003; Bargh et al., 1996)—and perhaps reflecting the activation and interpretation of a relevant social identity (Coesel et al., 2024; Yee & Bailenson, 2007). Second, perceived similarity between the embodied experiences of the user and other users of the same system should increase feelings of shared social identity between these people (following the principle of comparative fit; see P3b and P5b). Indeed, sharing sensorimotor experiences in VR can lead to a sense of social closeness (Järvelä et al., 2021; Piumsomboon et al., 2017). Moreover, the extent to which the degree of embodiment is shared between users should also increase feelings of shared identity (following P5c). Miller et al. (2019) showed that participants using an augmented reality (AR) headset (vs. no headset) felt less social connection with others not using a headset, consistent with this assertion.
To summarize, features specific to digital platforms (anonymity and pseudonymity, symmetry, the structuring of communication, information access regulation, algorithmic social categorization, persistent digital footprints, and embodiment) mean that social identity salience is predicted to operate differently in online (vs. offline) contexts. It is not simply that social identity is inherently more or less likely to be salient than personal identity online but rather that different digital platforms will have affordance for particular social identities to a greater or lesser extent. We can formalize these ideas as follows:
P6: Social identity salience will be shaped by features of digital platforms. Specifically, a person will be more likely to see themselves in terms of a particular social identity in a given online context when the digital platform they are using (P6a) makes personal identity less salient through anonymity or pseudonymity (so long as there are available cues relating to the social identity in question); (P6b) structures social interactions and connections collectively rather than interpersonally; (P6c) allows the group associated with this social identity to maintain collective privacy, secrecy, and/or confidentiality; (P6d) treats them as a member of the social category associated with this social identity; (P6e) reinforces how they have acted previously in accordance with this social identity (i.e., the identity is externalized on the platform); (P6f) involves an embodied context that is metaphorically appropriate to this social identity; and (P6g) involves an embodied context in which this social identity provides good comparative and normative fit.
Online versus offline identity
Up to this point we have discussed identity in terms of personal versus social identity. Another identity continuum created by the online context is offline to online identity. Personal identity can have both online and offline components (Hongladarom, 2011). Indeed, the individual is not necessarily the atom of identity—self-categories can be intraindividual (e.g., “my online self” vs. “my offline self”; see Turner & Onorato, 1999). Along these lines, Rodogno (2012) argued that “who I am online” can be considered to be part of the same continuous self as “who I am offline”; and for some people their online identity may even be more important than their offline identity. Moreover, in addition to having online versus offline personal identities, SDIT proposes that people may also experience a given social identity differently depending on whether they see themselves in terms of the online or offline version of this social identity. This is because people’s groups can be offline, online, or exist in both domains (Lehdonvirta & Räsänen, 2011). For example, a person who belongs to a religious group both online (e.g., on an online forum) and offline (e.g., at a place of worship) may feel that these represent two quite different groups, with different norms, attitudes, and emotions; and they may think, feel, and behave differently when they see themselves in terms of these identities. As Figure 1 suggests, considering people’s online versus offline identities turns the personal-to-social identity continuum into a two-dimensional identity space.
Personal/social and offline/online identity.
As with the personal-to-social identity continuum, the offline-to-online identity continuum is predicted to be shaped by the principles of fit and perceiver readiness (per P3). Specifically, (a) if a person generally sees themselves in terms of their online personal or social identities, they should be more likely to see themselves in terms of those identities, even when offline (i.e., these identities should be “sticky” or resistant to context); (b) if they perceive more similarity between their digital self and the digital selves of others than between their offline self and the offline selves of others, they should be more likely to see themselves in terms of their digital identities, even when offline; and (c) if people online are acting in terms of how they understand that people should act online, whereas people offline are acting in terms of how they understand that people should act offline, they should be more likely to categorize their social world (including themselves) in terms of online versus offline identity. The salience of these identities is also predicted to be shaped by externalization (P4), embodiment (P5), and features of digital platforms (P6).
This picture of online versus offline identity is complicated by the increasing integration of online and offline spaces (Spottswood & Wohn, 2020). This hybrid online/offline existence is facilitated not only by social media and smartphones but also emerging technologies such as the Internet of things, which involves embedding online capability in physical devices (S. Li et al., 2015); AR and VR (Rauschnabel et al., 2022; Xiong et al., 2021); and combinations of these such as the metaverse (K. Li et al., 2023). These new technologies have enabled increasing continuity between online and offline groups and activities (e.g., university courses that offer blended learning; O’Byrne & Pytash, 2015; organizations with flexible work from home arrangements; Shifrin & and Michel, 2022; and e-government; Malodia et al., 2021) as well as within personal identities (e.g., through offline identity verification on social media; Moore, 2018). This blurring of the line between online and offline life suggests a third kind of identity in this personal/social/online/offline identity space—hybrid identities. In particular, people may define themselves in a given context as someone who belongs to a personal or social identity that exists both online and offline—as opposed to an exclusively online or exclusively offline version of the same identity. Indeed, there are people who live their lives almost exclusively offline (König & Seifert, 2020; Lythreatis et al., 2022) or exclusively online (Tateno et al., 2019), which in certain contexts may become an intergroup distinction for those who share their existence across these domains. Together, the online/offline identity continuum, the personal/social identity continuum, and the existence of hybrid identities form a conceptual space that we refer to as the “social/digital identity space” (Fig. 2).
The social/digital identity space.
Although the six types of identity identified within this social/digital identity space are predicted by SDIT to operate in a psychological sense in the same way as any other social category (e.g., in terms of identity salience), they are likely to differ in terms of their outcomes. For example, offline social identities are more likely to create barriers both to engagement with technology and “technology people”—which has negative implications for groups that are more likely to be on this side of the digital divide (e.g., older adults; Friemel, 2016)—particularly as society itself becomes more “online.” Similarly, online social identities may create barriers to participation to the extent that society is seen as not online enough. A key group to consider here is younger people, who increasingly live their lives online (Szymkowiak et al., 2021) and may feel that they are not included in broader society as a result. Furthermore, hybrid identities, both personal and social, have an increased potential to result in self-presentation challenges such as context collapse (Boyd & Ellison, 2007) to the extent that these identities come with incompatibilities between online and offline versions of the same group or role. For example, a teacher who feels part of a school community both online and offline may find it harder to navigate the different self-representation requirements of these domains than a teacher who sees themselves clearly as part of either the “offline” or “online” school community (V. Cho & Jimerson, 2017). This is because a hybrid identity has a potentially wider range of relevant social norms and roles given that it includes both online and offline domains. However, hybrid social identities also have the capacity to be a particularly strong source of social resources for individuals because they can potentially draw on the group for help both online and offline (see P12a).
The salience of these different kinds of identity is predicted to depend on context and therefore likely to change to the extent that people move through different social environments. In addition to the predictors discussed in P3 through P6, another consideration in the equation of salience is the idea of “compatibility” between these identities that we introduced in the previous paragraph. Identities vary in the degree to which they are compatible with each other (Crisp & Hewstone, 2007; C. Haslam et al., 2021; Roccas & Brewer, 2002) such that some social identities subjectively preclude membership of others whereas others can be held simultaneously. The original formulation of SCT (Turner et al., 1987, p. 49) proposed that identities are functionally antagonistic such that the salience of one level of self-definition makes other subordinate or superordinate identities less salient. However, this idea has subsequently been contradicted by findings that personal and social identity can be mutually reinforcing under conditions of “identity fusion” (e.g., Baray et al., 2009; Gómez et al., 2011; Swann et al., 2009). Consequently, SDIT proposes that identities can be functionally antagonistic but only to the extent that the content of these identities is perceived as incompatible (e.g., when being a unique individual is seen to be incompatible with being a group member; or when being an “online person” is seen to be incompatible with being an “offline person”). Greater perceived incompatibility is predicted to lead to greater functional antagonism such that more strongly held and/or more situationally appropriate identities will “crowd out” less meaningful, incompatible identities. In contrast, when identities become not only compatible but also functionally equivalent (as in the case of identity fusion; Swann et al., 2009), the salience of one identity is predicted to make the other more salient. Accordingly, personal online, offline, or hybrid identity is not necessarily predicted to be a barrier to social identity salience because social identity salience depends on the compatibility of the content of these identities and their integration within the self.
The specific predictions of SDIT relating to the digital/social identity space are as follows:
P7: A person can define themselves in terms of their online versus offline versus hybrid identities, personal or social, leading to the following distinct types of identity: • Online personal identity—in which the self is seen primarily in terms of one’s own personal online idiosyncrasies • Offline personal identity—in which the self is seen primarily in terms of one’s own personal offline idiosyncrasies • Online social identity—in which the self is seen primarily in terms of online group or social category membership • Offline social identity—in which the self is seen primarily in terms of offline group or social category membership • Hybrid personal identity—in which the self is seen primarily in terms of one’s own personal idiosyncrasies both online and offline • Hybrid social identity—in which the self is seen primarily in terms of membership of a group or social category that exists both online and offline
P8: These different kinds of identities are predicted to have the following unique outcomes: (P8a) Salient online, offline, and hybrid social identities are more likely to cause barriers to participation in mismatched social contexts (e.g., online identity/offline context) versus matched social contexts (e.g., online identity/online context); (P8b) compared with online or offline identities, hybrid identities are more likely to result in self-presentation challenges; and (P8c) hybrid social identities (vs. online or offline identities) have an increased potential to provide individuals with social resources (see P12a).
P9: As with personal versus social identity (P3–P6), in any given context a perceiver will define themselves in terms of their online, offline, or hybrid identities depending on features of themselves, their environment, and (when online) the digital platform that is being used.
P10: The perceived compatibility of online, offline, and hybrid identities (personal and social) will contribute to their relative salience. Specifically, (P10a) the more that these identities are perceived as incompatible in terms of their content, the more their salience will be functionally antagonistic such that identities that are more strongly internalized and situationally appropriate will become more salient and those that are less strongly internalized and situationally inappropriate will become less salient; and (P10b) when identities are seen to be functionally equivalent in terms of their content (i.e., they are “fused”), the salience of one identity will increase the salience of the other.
P11: As with personal versus social identity (P1 and P2), online versus offline versus hybrid identity salience will shape a person’s perception, cognition, emotion, and behavior in line with the content of the salient identity.
Individual outcomes
Social identity theorizing of this form can be used to make predictions not only about the extent to which social contexts and digital platforms shape people’s online and offline identities but also about how this might in turn affect the well-being and performance of individuals and groups. Beginning with implications of digital social identity for individuals, there is a large body of empirical research that has demonstrated that social identity (internalized group membership and group identification) has a range of well-being benefits (C. Haslam et al., 2018; S. A. Haslam et al., 2018; Jetten et al., 2012). Importantly, these benefits are derived via different mechanisms. In particular, one benefit of social identity is that it fulfills the fundamental human need for social connection, thereby increasing psychological well-being (C. Haslam et al., 2018). Along these same lines, online social identities have been linked to psychological well-being (Kaye et al., 2019; Naserianhanzaei & Koschate-Reis, 2021; Pendry & Salvatore, 2015). Another benefit of social identity is that it can improve physical-health outcomes, in part because groups facilitate access to social resources in times of need (Best et al., 2018; Bliuc et al., 2017; Godard & Holtzman, 2024; S. A. Haslam et al., 2018). It is in this domain that the benefits of digital identities may depend on the extent to which they correspond to offline resources. Indeed, as Kaye et al. (2019) found, online social identities are particularly beneficial when they facilitate offline social identities. As previously mentioned, hybrid online/offline social identities may be particularly potent in this regard.
Another mechanism by which digital (social) identity might affect individuals concerns the influence of online group norms. A key finding in the social identity literature is that the impact of social norms (i.e., written and unwritten rules about how people should behave) on individual behavior depends on the extent to which a person is identified with the group that the norms relate to (Hogg & Reid, 2006; Terry & Hogg, 1996). What this means is that when a person identifies with a group, they are guided by the group’s norms—for example, being highly identified with a group with health-conscious norms is likely to motivate health-conscious behavior (J. Liu et al., 2019), whereas identifying with a group with risk-taking norms is likely to lead to risk-taking behavior (Cruwys et al., 2021). These predictions help to explain why people who identify with online groups that promote alcohol consumption are more likely to engage in problematic alcohol use (Pegg et al., 2018; Ridout et al., 2012). However, SDIT also predicts that identifying with online groups that have positive and desirable norms will motivate behavior in line with these norms in ways that can be beneficial for individual group members. Similar predictions might follow from an identity theory analysis, with salient roles influencing behavior in a similar way to norms. However, beyond this, SDIT predicts that the influence of norms or roles on a person’s behavior will be moderated by group identification such that they will be impactful only when that person is highly identified with the group the norms or roles are associated with.
Furthermore, the effects of digital group membership and identification on individual well-being outcomes are likely to be affected by the extent to which these processes allow group members to be connected to other groups and to society more broadly. Research has found that belonging to multiple groups predicts increased well-being outcomes (Cruwys et al., 2016; C. Haslam et al., 2016; Sani et al., 2015). Moreover, the extent to which a group is positively valued by other groups and society is also an important moderator of the impact of social identity on well-being (DeMarco & Newheiser, 2019; Jetten et al., 2017)—indeed, even the benefits of belonging to multiple groups are reduced if these groups are devalued by others (Sønderlund et al., 2017). The implication of these findings is that if digital social identity leads a person to define themselves in terms of a group membership that excludes them from broader society, or that cuts them off from other existing group memberships, this will result in a “social curse” rather than a “social cure” (C. Haslam et al., 2018; Wakefield et al., 2019). As we have discussed previously, online- or offline-only identities (vs. hybrid identities) may be problematic in this regard.
We can summarize these predicted impacts of digital social identity on well-being as follows:
P12: Online group membership and social (vs. personal) online identification will be associated with improved psychological and physical well-being to the extent that they (P12a) facilitate offline group membership and social identification, (P12b) promote healthy rather than unhealthy group norms, and (P12c) allow for integration with rather than exclusion from other groups and the broader society.
Intragroup outcomes
SDIT predicts that digital identities will affect not only individuals but also intragroup and intergroup dynamics. Regarding intragroup dynamics, research has shown that social identity functions as a form of “social glue” that makes group life possible (Ellemers et al., 2004; Turner & Oakes, 1997; van Vugt & Hart, 2004). Specifically, group identification motivates group members to work on behalf of the group (van Knippenberg, 2000), increases trust in other in-group members (Cruwys et al., 2021; Voci, 2006), improves communication within the group (Greenaway et al., 2015), facilitates shared group emotions (Kuppens & Yzerbyt, 2012; Yzerbyt et al., 2003), and is a basis for people to see the group as a group (Yzerbyt et al., 2000). These dynamics support an improved capacity for organization and performance within groups whose members share a strong sense of identity (Ellemers et al., 2004; S. A. Haslam, 2004; van Knippenberg, 2000). SDIT predicts that the same effects will also be present for digital identities such that shared identity within an online group will make that group more cohesive and functional. These predictions align with existing social identity accounts of online group formation (Code & Zaparyniuk, 2010; Lüders et al., 2022). Moreover, these predictions represent a novel element of SDIT that differentiate it from identity theory primarily in outlining not only how online group membership shapes the psychology of group members (e.g., through providing social roles) but also how this collective psychology affects the functioning of the group. Specifically, our theory predicts the following:
P13: Groups (online, offline, or hybrid) in which group members have a stronger sense of shared social identity will be more cohesive and functional.
However, these positive effects of shared identity on group dynamics are predicted to depend on the extent to which the shared identity corresponds to that specific group. There are situations in which shared social identity can undermine group dynamics, specifically when this shared identity focuses on subgroups (Chrobot-Mason et al., 2009). In these cases, shared social identity is predicted to lead to more cohesion within the subgroups but can lead to reduced cohesion in the group as a whole if it creates an intergroup context within the main group. This is particularly relevant for groups that exist both online and offline because shared social identity within an online version of a group may not necessarily flow through to the offline version of that same group if the relevant distinction is between “us online” versus “them offline” rather than between “us (online and offline)” and “them (online and offline).”
This is where the principles of online identity salience (P3–P6, P9, and P10) are crucial in determining the impact of shared online identity on group performance. For example, if there are many differences between group members who tend to exist online and those who tend to exist offline, they may perceive two groups (an online group and an offline group), potentially leading to an intergroup context in which these groups see each other as “us” and “them.” Such an intergroup context is predicted to lead to competition between the online and offline versions of the group and thus reduced performance of the group as a whole. However, in a situation in which group members both offline and online feel that these identities form part of a single “superordinate” identity (Gaertner & Dovidio, 2000; Gaertner et al., 1993), shared identity is predicted to have the positive effects outlined in P13. Fostering hybrid social identities may be an effective way to bridge such subgroups, but there is still the potential to create fault lines between hybrid and nonhybrid (entirely online or offline) subgroups. The important part here is that such hybrid identities should be superordinate and shared within the whole group:
P14: Groups that exist both online and offline will be more cohesive and functional when group members identify with the superordinate (online and offline) group but less cohesive and functional when they identify with either the online or offline subgroup.
An important contribution of the social identity perspective has been to provide a novel perspective on leadership (S. A. Haslam et al., 2020; Hogg, 2001). In contrast to approaches that focus on individual difference traits or skills of leaders (the “zombie leadership” metatheory; S. A. Haslam et al., 2024), the social identity approach identifies several contextual factors that make leaders influential within a group. Specifically, to be influential, leaders must (a) be seen to be prototypical such that they best represent how the group sees itself, (b) be seen to be advancing the cause of the group rather than their own interests or the interests of other groups, (c) be able to articulate a shared vision for what the group is through identity entrepreneurship, and (d) deliver concrete outcomes for the group that make the group matter in the world through identity impresarioship (Steffens et al., 2014). SDIT builds on the social identity approach to leadership by highlighting domain consideration (online, offline, or hybrid) as important for shaping the influence of leaders. Specifically, we propose that the social identity influence of a leader will depend on whether the group is offline, online, or hybrid and how the leader is positioned in relation to this. For example, the prototype of the group online may not be the same as the prototype of the group offline—a leader who utilizes technology to communicate may seem more prototypical to online group members and less prototypical to offline group members. Moreover, representing the cause of the group online may not support the group offline and could even impair it if these subgroups are opposed or if supporting the group online means the group cannot be supported offline. More formally, SDIT makes the following predictions in relation to identity leadership:
P15: A leader of an online, offline, or hybrid group will be influential to the extent that they (P15a) are perceived as prototypical of that specific group as a whole (rather than of a different group or a subgroup); (P15b) are seen to be advancing that specific group’s interests (rather than their own personal interests, the interests of another group, or the interests of a subgroup); (P15c) engage in effective identity entrepreneurship that increases the coherence and cohesiveness of that specific group (rather than that of a different group or a subgroup); and (P15d) engage in effective identity impresarioship that makes that specific group matter (not another group or a subgroup).
Intergroup outcomes
SDIT can also be used to make predictions about intergroup relations online. Indeed, social identity theory was originally developed as a theory of intergroup relations (Tajfel & Turner, 1979). The central prediction here is that people identifying with a group will generally be more positively disposed toward fellow in-group members (i.e., will display in-group favoritism; Brewer, 1999; Tajfel et al., 1971). An exception is when an in-group member is perceived to be working against the group or its norms, in which case they will be seen less positively (i.e., the “black sheep” effect; Marques et al., 1988). Moreover, when there is a context in which the group that a person identifies with is competing with another out-group, they will tend to be less positively disposed toward members of that out-group—although the nature of this reaction (e.g., whether it takes the form of out-group derogation; Branscombe et al., 1999) will also be conditioned by in-group norms (McGarty, 2001). These psychological effects also flow through to behavior such that people tend to interact more positively with in-group members than with out-group members—although again this should vary as a function of both their identification with the group in question and its norms (Hewstone et al., 2002). We predict that these effects will also be found in an online context such that people will treat in-group members more positively than out-group members to the extent that the in-group is one they identify with. This prediction can be formalized as follows:
P16: To the extent that they identify with a given in-group, people will treat in-group members more positively than out-group members online unless they are perceived to be working against the group or its norms. The nature of this treatment will be conditioned by group norms.
Moreover, building on P16, we can use a social identity perspective to make predictions about the conditions under which online intergroup contexts will be seen as competitive versus cooperative. In particular, increasing the salience of a superordinate identity (i.e., one that encompasses both the in-group and out-group) can reduce intergroup tension by shifting the comparative context (Gaertner & Dovidio, 2000; Gaertner et al., 1993). Leadership is key to this process because leaders’ identity entrepreneurship can help to bridge the gap between the in-group and out-group by cultivating a sense of superordinate identity and respectful norms (S. A. Haslam et al., 2020). For example, Barack Obama successfully appealed to shared American narratives to appeal to voters across racial boundaries in his 2008 election campaign (Hammer, 2010). Focusing on shared beliefs or common fate can also help to reinforce a sense of superordinate identity (Batalha & Reynolds, 2012; Gaertner et al., 1993). We predict that these effects will also occur in an online context (for a similar discussion, see Lüders et al., 2022) such that efforts to build and promote superordinate identity by highlighting shared opinions, narratives, and fate will dampen the appetite for intergroup conflict between online groups (and between online and offline versions of the same group; Simchon et al., 2022; Van Bavel & Packer, 2021). More formally, then, we predict:
P17: Emphasizing a superordinate group identity through shared opinions, narratives, and fate will reduce tension between online groups as well as between online and offline versions of the same group.
In P6 we introduced the idea of social identity affordances regarding identity salience. Because social identity affects intragroup and intergroup dynamics, features of digital platforms are also predicted to shape these outcomes through social identity. In particular, the growing extent to which social interactions and belonging exist on algorithmic social media has implications for these outcomes. For example, research shows that people tend to interact with others who share the same views (Brady et al., 2017) and demographics (Monti et al., 2023; Usher et al., 2018) on social media platforms. Algorithmic feeds amplify this effect (Levy, 2021). Specifically, by siloing people on the basis of differences in opinions and demographics, algorithmic social media contributes to polarization (Van Bavel et al., 2021). Indeed, both X’s and Facebook’s algorithms have been linked to polarization (Levy, 2021; Yarchi et al., 2021)—with the latter being found to causally increase political polarization (Allcott et al., 2020). This effect occurs because AI-enabled algorithms can identify the social categories that users belong to (AL-Qawasmeh et al., 2022; Conover et al., 2011; Gichoya et al., 2022) and are designed to provide them with (collectively) self-relevant and affectively charged content (Milli et al., 2025; F. Smith et al., 2025). From a social identity perspective, designing such algorithms to optimize engagement with the platform will therefore tend to “supercharge” polarized social identities. This has the potential to be problematic because humans are highly sensitive to group concerns such as exclusion (Williams, 2007), social comparison (Gilbert, 2001; Gilbert et al., 1995), morality (Axelrod & Hamilton, 1981), and intergroup conflict (Rathje et al., 2021). Accordingly, algorithms optimized for engagement with the platform will deliver content to users that is relevant to these evolved instincts—activating polarized social identities and constructing polarized intergroup contexts that may be harmful for society (Van Bavel et al., 2024).
Indeed, this effect of engagement-based algorithms on social identity has implications for a variety of important group processes. For example, algorithmic social media use has been identified as a catalyst for extremism and radicalization (Risius et al., 2024) partly because it provides engaging, polarizing, collectively motivating content. Similarly, affective polarization on algorithmic social media facilitates the development of conspiracy theories (Cinelli et al., 2022; Grandinetti & and Bruinsma, 2023) and the spread of misinformation (McLoughlin et al., 2024). The increasing prevalence of online vigilantism, collective moral outrage, and “cancel culture,” in which groups of online users target perceived wrongdoers, has also been linked to algorithms on social media (Mihailov et al., 2023; Sekuler, 2024; Walls et al., 2025). Specifically, these outcomes arise when affectively polarizing content relating to a particular issue or target is rapidly shared, allowing novel social movements to form and persist. Alongside algorithms, social identity has been identified as playing a key role in these processes (Leach et al., 2025; Mihailov et al., 2023; Walls et al., 2025).
Accordingly, SDIT predicts that engagement with social media platforms that use algorithms optimized for engagement with the platform will lead to increased polarization, extremism, radicalization, conspiracy beliefs, online vigilantism, collective moral outrage, and normative policing. However, we also predict that these intragroup and intergroup dynamics operate primarily through social identity such that these outcomes are far less likely to materialize if using these platforms does not make social identities salient (P6) and does not create a sense that there is a hostile “them” threatening “us” (P17).
P18: Social media algorithms that are designed for engagement with the platform will make social identities more salient and lead users to perceive these identities as under threat. This increased sense of both social identity and group threat will in turn tend to increased polarization, moral outrage, extremism, radicalization, conspiracy beliefs, misinformation, and online vigilantism.
Contextual factors
Similar to the initial formulations of SIT and SCT, the general psychological model of the perceiver and the context (including technology) coming together to predict outcomes outlined in SDIT is assumed to operate more or less universally. However, in recent years the field of social identity has been enriched by research and theorizing about the ways in which cultural differences moderate both (a) the inputs to this model and (b) the way that this model operates on these inputs.
Regarding the first point, patterns of technology use differ between cultures (Bandyopadhyay & Fraccastoro, 2007; K. J. Calhoun et al., 2008). For example, collectivist cultures tend to have higher levels of social media addiction than individualistic cultures (Cheng et al., 2021). Similarly, individual differences in collectivism predict intentions to use social media in online and blended learning (Trivedi et al., 2024). Digital technologies are also used differently in collectivist versus individualistic cultures (Hong & Na, 2018)—for example, social media networks tend to be less egocentric in collectivist than in individualistic countries (Na et al., 2015, p. 201) such that users in individualistic countries have larger networks that are less based on family, friends, and groups (Jackson & Wang, 2013). SDIT predicts that in cultures in which digital technologies are more prevalent, and particularly those in which these technologies are used for social connection, online and hybrid identities will be more prevalent.
Moreover, to the extent that social identities tend to be more online-based throughout a culture, online group membership is predicted to have a more positive impact on individual well-being because important social resources in this cultural context are more likely to be organized via digital technologies. For example, in a culture in which social connections are generally maintained online, it may be easier to receive concrete help from one’s groups through online platforms. This idea is supported by research showing that internet users had more social capital than non-internet users in Singapore, which has a technologically developed and relatively collectivist society (Skoric et al., 2009). These differences in the prevalence and purpose of digital technology are also predicted to shape the social meaning of this behavior by affecting social identity processes (i.e., normative fit; P3c). Specifically, when technology use is normative for a social identity (which is particularly likely for online and hybrid identities), that identity should be more self-relevant in the context of technology use.
Regarding the second point, research has shown that culture shapes key social identity processes. For example, groups are generally seen as more “entitative” (as distinct entities) in collectivist cultures (Spencer-Rodgers et al., 2007), a judgment that is related to comparative fit (P3b). Similarly, in making these judgments, collectivist cultures tend to focus more on interconnectivity within groups than on group-level traits, on social structure more than individual needs, and on agentic rather than essence properties (Kurebayashi et al., 2012). Social identities tend to be constructed more through interpersonal ties than through social category membership in collectivist cultures (Brewer & Yuki, 2007; Yuki, 2003), in line with these findings. However, there are also more specific cultural differences—some collectivistic societies (e.g., Japan; Kurebayashi et al., 2012) focus more on interconnectivity versus group-level traits than others (e.g., China; Brewer et al., 2004). Drawing on these findings, SDIT proposes that digital technologies that shape interpersonal connections will have a greater effect on social identity in collectivist than in individualistic cultures—particularly in collectivist cultures that emphasize the importance of relational ties. The moderating effects of culture on the processes of SDIT are summarized as follows:
P19: Culture is predicted to moderate the processes of SDIT in the following ways: (P19a) People in cultures in which digital technologies are more widely used, used more for social connection, and are normatively associated with social identities are predicted to have, on average, more meaningful online and hybrid identities (in such cultures, online group membership is predicted to have a greater impact on well-being; see P12); and (P19b) digital technologies that shape interpersonal connections will have a greater effect on social identity salience in collectivist cultures, particularly those that emphasize relational ties.
To the extent that digital technologies are integrated into everyday life, SDIT predicts that social identity processes will be shaped and mediated by these technologies (P6). Accordingly, social identity is increasingly beholden to these technologies and is therefore influenced by the political context surrounding them. For example, the Chinese government (both directly and through technology companies such as Tencent) has engaged in censorship of content online and exerts control over influential social media users because it recognizes the importance of online platforms for “opinion leaders’ ability to organize individuals around counter-hegemonic ideas” (Gallagher & Miller, 2021, p. 1018). Similarly, in 2017 Ukraine blocked access to the Russian social media platform VKontakte to reduce the influence of Russian propaganda among its citizens (Golovchenko, 2022). Both of these cases involved states limiting access to social media platforms or controlling the content on these platforms to shape social dynamics. The actions of governments can also motivate the companies that provide digital platforms to engage in regulation—for example, Facebook banned the far-right group Britain First after members of the group were convicted for hate crimes (Nouri et al., 2021). Moreover, beyond intentional deplatforming, the political environment can shape access to platforms for certain groups by systematically depriving them of social and technological resources (Epstein et al., 2011; Vassilakopoulou & Hustad, 2023). Drawing on the principles we have already outlined, SDIT predicts that the political environment regarding digital technologies will affect individual and collective outcomes via social identity. This can be formalized as follows:
P20: Political environments that support the ability of individuals to connect to psychologically meaningful groups through digital technologies will increase (P20a) the well-being of these individuals (to the extent that belonging to these groups constitutes a social cure rather than a curse; see P12) and (P20b) the potential of these groups to act collectively (see P13).
Another way that the cultural and political context can shape the processes and outcomes of SDIT relates to differences between the online and offline world. For example, although we proposed that social media algorithms designed for engagement are likely to increase social identity variables that predict polarization (see P18), this may depend on the alternative offline culture in which the person is situated. For example, Asimovic et al. (2021) found that ethnic polarization increased when Bosnian participants deleted Facebook during genocide remembrance week—potentially because the offline context was more polarized than the online context in this case. Moreover, the political environment can shape culture, as demonstrated by findings that collectivistic expressions in China increased after the COVID-19 epidemic (Han et al., 2021). This relates to a broader point, which is that although the core psychological mechanisms of SDIT (P1–P5, P10, and P12–17) are assumed to be relatively stable in terms of how they operate, the components specifically relating to digital technologies (P6–P9, P11, and P18) and their social context (P19 and P20) may change rapidly as these technologies, their use, and the societies in which they are embedded develop. In other words, the psychological elements of SDIT will change at the speed of biology, whereas the other elements will change at the speed of culture (Perreault, 2012).
SDIT explains how social identities can be—and often are—an important part of a person’s sense of who they are online (i.e., their digital identity). At the same time, SDIT highlights the importance of accounting for digital identity when trying to understand social identity, both online and offline. Furthermore, it articulates the conditions under which particular identities (personal vs. social, online vs. offline, or hybrid identities) will become psychologically self-relevant. And last, this theory outlines some of the key impacts of social identity on individuals and groups, both online and offline. The formal propositions of this theory are set out in Table 1.
Propositions of SDIT
Note: SDIT = social digital identity theory.
Applying SDIT
SDIT can be applied to understand the antecedents and consequences of social identity for a range of stakeholders in digital contexts, including individuals, groups, societies, and those who design and implement technology.
Implications for individuals
There are several ways in which social identity might affect individuals in a digital context. Notably, because group membership and social identification (whether online or offline) are predicted to improve well-being outcomes so long as they do not promote unhealthy norms or lead to broader social exclusion, it follows that individuals wishing to improve their psychological and physical health are likely to benefit from engaging with digital platforms that promote, and help people to build, positively valued and integrated social identities. Along these lines, social identity has been proposed as a solution to the loneliness epidemic (S. A. Haslam et al., 2022), and SDIT suggests that online social identity formation may be a key mediating variable between internet use and decreasing loneliness in people who find it easier to engage socially online (Deters & Mehl, 2013). Moreover, because offline groups can be sources of material resources and assistance whereas online groups may provide resources to group members that offline versions of the group cannot provide (such as when the group is highly stigmatized offline or the group member is physically isolated), groups that have both an online and offline component (hybrid groups) are likely to provide more substantial benefits than those that are online- or offline-only (e.g., Kaye et al., 2017, 2019).
However, there are also several ways in which social digital identities might lead to negative outcomes for individuals. As outlined in P12b, one follows from the observation that identifying with a group that has unhealthy social norms (e.g., those that are dangerous or antisocial) is predicted to motivate individuals to act in line with those norms. Another relates to identification with online groups that increases trust in those groups and their members. Although important for the intragroup functioning outlined in P13, this increased group trust may lead group members to be more likely to engage in risky behavior when they are thinking as group members (Cruwys et al., 2021). For example, we would predict that strongly identified group members would be less concerned about cybersecurity-related issues such as data breaches when these issues are caused by in-group members (Bingley, 2021). In such contexts, however, the specific content of digital social identities (and norms around such things as trust) will be important in determining their specific effects on individuals.
Following from P12c, another way in which digital social identities might negatively affect individuals is when connection to an online group leads to feelings of exclusion from other groups or from society more broadly. For example, a qualitative study of young people living in rural areas in the United Kingdom (Awan & Gauntlett, 2013) found that increased online inclusion was associated with feelings of cultural exclusion. In particular, some young people in this study felt that their online selves were more authentic than their offline selves and that this led them to feel disconnected from people in their local area. In such contexts, it is vital to ensure that digital identities are compatible with offline identities (e.g., through social identity mapping; Bentley et al., 2020; Cruwys et al., 2016) and that efforts are made to avoid stigmatizing people who gain self-meaning from online life.
Implications for groups
SDIT makes predictions about social identity and intragroup dynamics online that can be applied to help understand and improve the functioning of groups. For example, according to P13 shared social identity is the key ingredient underpinning group functioning in both online and offline contexts. In combination with the predicted effects of digital platform features on identity salience outlined in P6, we can analyze how social networking platforms might facilitate online group functioning. Specifically, we would predict that groups operating on a platform that (a) forces digital identities to be linked to offline identities and (b) structures social interactions in an interpersonal (one-to-one) rather than a collective way, (c) does not allow groups to interact privately, (d) treats group members algorithmically as members of different groups, (e) does not “save” records of previous group-related behavior, and (f) creates mismatches between the embodied contexts of group members and the group as a whole will be less cohesive and functional than groups formed on a platform that (a) allows for anonymous or pseudonymous interaction (as long as cues to the group identity are still salient), (b) affords group interactions, (c) allows groups to keep interactions private from outsiders, (d) treats group members algorithmically as members of that group, (e) compiles and makes salient records of previous group-related behavior, and (f) matches the embodied contexts of group members and the group as a whole. Following from this, we would suggest that those looking to form and maintain online groups should use platforms that resemble the latter model rather than the former. A related point is that leaders of groups should seek to shape and harness social identity (P15) to exert social influence in a digital context.
In line with P14, research into “fault lines” has found that social identity within subgroups can undermine superordinate group cohesion (Voida et al., 2012). However, researchers have not considered the possibility of identity fault lines based on domain (online vs. offline vs. hybrid). Following P14, if an organization has some online and some offline workers (as many do following the move to working from home prompted by the COVID-19 pandemic; Bick et al., 2023), this would be sufficient grounds for subgroup formation even in the absence of other potential fault lines. However, this would be particularly fertile ground for subgroup formation (and potentially organizational disruption) if other attributes are shared by those working online versus offline. This is because, consistent with the principle of comparative fit (P3b; Oakes et al., 1994), multiple overlapping fault lines have been found to intensify these effects by making intergroup comparisons between subgroups more salient (Zanutto et al., 2011). For example, an organization in which only employees who are based overseas work from home is more likely to have fault lines than one in which some overseas employees work in an office whereas some local employees work from home.
Implications for society
SDIT can be used to analyze the psychology of intergroup dynamics online and the implications of these dynamics for society. In particular, because identification with a group leads a person to treat others as in-group or out-group members (P2 and P16), the same digital platforms that foster shared social identity have the potential to be hubs for intergroup conflict. For example, the social media platforms X and Telegram both have features that SDIT identifies as potentially facilitating shared social identities—the former uses asymmetric social ties between users, allows for pseudonymity, and encourages collective interactions via hashtags, whereas the latter also allows for pseudonymity alongside large-scale group communication with collective privacy through encryption. Accordingly, SDIT predicts that these platforms are ripe for the development of both in-group solidarity and intergroup conflict. Indeed, a study of Telegram messages found that many of these contained identity-based content, specifically in-group favoritism and out-group derogation (Schlette et al., 2023). A further notable finding was that 35% of messages voicing in-group support also contained out-group criticism, implying that the group-friendly structure of Telegram carries with it the risk of fomenting intergroup conflict. Similarly, X has been found to encourage polarization (Hong & Kim, 2016). More generally, there is clearly potential for people to condone and engage in more extreme behavior when they are thinking and acting as group members (rather than individuals) online (D. Rieger et al., 2020). In particular, if people come—or are led—to identify with extremist online political groups then they will be much more open to acting in ways that help to realize those groups’ objectives (e.g., through acts of violence; Gaudette et al., 2021; Strindberg, 2020).
Developing this point, appreciating the double-edged sword of shared social identity online (which makes it good for groups but often bad for relations between groups) is vital for understanding the rise of social problems such as hate groups, conspiracy theories, and political polarization—all of which are facilitated by online platforms (Bliuc et al., 2018; Cinelli et al., 2022; Van Bavel et al., 2021). From the perspective of SDIT, just as fault lines between subgroups can destabilize organizations, so too the superordinate group of society will become less cohesive the more that people identify with polarized ideological or demographic-based online groups. However, by the same token, efforts to make the superordinate group salient and meaningful and to promote respectful intergroup relations have the potential to help reduce conflict between subgroups (P12; Gaertner & Dovidio, 2000). In particular, SDIT predicts that if online platforms can highlight shared opinions, narratives, and fate, they can lessen or even reverse the societal fragmentation they are currently driving. That said, this is unlikely to happen so long as structural factors continue to incentivize this fragmentation (Ghosh, 2020; Lauer, 2021). Indeed, following from P18, a purely commercial logic (Thornton et al., 2012) of technology design is fundamentally at odds with social cohesion.
From a different perspective, the subgroup conflict described in the previous paragraphs can be seen as a struggle for social influence within society—or indeed as resistance. There are two key ways in which those seeking to resist might make use of social identity online. The first relates to harnessing the power of groups, which requires identity leadership in a digital context that helps to build social identity. For example, hashtags on Twitter were identified as a key piece of the technological infrastructure that allowed the Arab Spring movements to unite people in resistance online (Lotan et al., 2011), partly because these hashtags facilitated the creation of opinion-based group identities (McGarty et al., 2014). In movements such as the Arab Spring, the lack of formal hierarchical structures online can create an opportunity for identity leadership to emerge from within and to be exercised by those who are “prominent nodes in the digital networks” (Arafa & Armstrong, 2016, p. 76).
The second way that resistance can harness social identity online is outlined by the political solidarity model of social change (Subašić et al., 2008). This model proposes that minority groups can exert social influence by positioning themselves to be seen as more representative than authority groups of a superordinate identity that they share with the majority. In other words, if majority group members see a minority group as more like “us” than the authorities, that minority group will have more power to challenge those authorities. An example of this kind of online social influence was seen when far-right groups used hashtags such as #TruckersForFreedom2022 to position themselves as sharing an identity with a working-class majority (i.e., “us”), in contrast to left-wing leaders who were positioned by hashtags such as #COVID1984 as forming a global elite that was seeking to oppress the world’s populations (i.e., “them”; Farokhi, 2022). Indeed, it is by using digital platforms to shape and harness social identity in precisely this way that far-right groups have experienced a surge in popularity, mainstream acceptance, and influence (Winter, 2019).
Implications for digital system design and implementation
SDIT also has implications for those who create and implement digital systems and for governance and legislation related to their use. For example, designers should consider the extent to which their systems have affordance for social identities through specific features that affect the processes of identity salience outlined in P6. In particular, systems that do not allow people to see themselves in terms of social identity online, or that facilitate subgroup at the expense of superordinate social identity, will obstruct particular group dynamics. This may be desirable or undesirable depending on the goals of the designer, but it is nevertheless important to consider social identity processes in design.
Along these lines, following P20, designers, developers, and those who implement and regulate systems (such as governments) should be aware that they can empower or disempower groups through the choices they make. For example, systems that systematically treat people from certain groups poorly and at the same time do not allow them to access the benefits of group membership (e.g., digital welfare systems that are not only difficult for older adults to use but also allow them to engage only as individual users) risk widening the various digital divides that are opening up across societies around the world (Olsson & Viscovi, 2023). Moreover, private organizations that maintain digital platforms that can facilitate social identities have immense power to influence online groups—and by extension the offline world. In particular, a company such as X or Meta that provides a platform for groups to exist online is also able to harm those groups by making it hard for them to use their digital infrastructure or removing it altogether. Indeed, Facebook has previously banned particular groups on the left and right of politics (Biddle, 2021). Although these bans were ostensibly intended to prevent armed groups from organizing, this power can also be wielded against groups that might seek to challenge social media companies themselves (e.g., unions).
Another implication of our theory for designers and developers is that social identity needs to be accounted for as a vital part of human-centered design because digital systems can affect people’s well-being via identity processes (Bingley et al., 2023). Specifically, as outlined in P6d, a digital system that treats people as group members is predicted to make those identities salient. According to Bingley et al. (2023) “social self-determination” model of AI system impact, this salience in combination with the way the group is perceived as being treated by the system or society is predicted to affect self-determination needs and, by extension, well-being. This is because social identity salience increases the importance of group-based needs for autonomy, competence, and relatedness—and at the same time the way in which the system treats the user provides information about how the user’s group is treated by society, affecting those group-based needs. Moreover, this means that a system that performs worse for certain groups than for others (e.g., Blodgett & O’Connor, 2017; Obermeyer et al., 2019) is predicted to negatively affect not only the well-being of a defined subset of users but also the well-being of their fellow group members who are affected by or become aware of that system. Moreover, drawing on P18, designers should be aware that utilizing algorithms that are optimized solely for platform engagement is likely to negatively affect social cohesion because of their impact on social identity.
In summary, according to SDIT, digital system design and implementation have profound implications for social identity, and through this for individuals, groups, and society. This means that those who design, regulate, and implement digital systems should be aware of the potential impacts of their decisions on these outcomes—and the key role of social identity as a psychological mediator.
Theoretical contributions
Building on the point with which we concluded the previous section, we can see that the key theoretical contribution of SDIT relative to existing theoretical accounts of digital identity (in particular those that are based on identity theory) is that it accounts for a much wider range of social phenomena. This is because although previous accounts based on identity theory provide a compelling account of how an individual’s digital identity is shaped by social structures (Bullingham & Vasconcelos, 2013; Davis, 2016; Marwick & Boyd, 2011), these accounts do not go further to explain how digital identity makes group life possible online (cf. S. A. Haslam et al., 2003; Turner, 1982). Reconceptualizing digital identity in terms of personal versus social identity bridges the psychological (intraindividual) and sociological (individual within groups and society) levels of analysis, providing a way to understand how individuals perceive, think, feel, and act together as group members online. Accordingly, by bringing group psychology into the digital identity domain, SDIT can account for social phenomena at multiple levels of analysis (individual, interpersonal, intragroup, and intergroup). Moreover, by connecting digital identity to the wider social identity literature, the theory generates novel predictions regarding topics as diverse as well-being, leadership, organizational team dynamics, polarization, and minority group influence.
Accordingly, although SDIT intersects with existing identity theory accounts relating to the influence of social norms on people’s online behavior (e.g., Davis, 2016), it goes further by proposing that group identification is a vital psychological mediator of this relationship. This implies that people will act to verify identities online only to the extent that these identities are made salient by features of the social context. In other words, it is not just people’s commitment to an identity that will make this identity salient and therefore meaningful—but also the fit between this identity and the social context in which they find themselves.
Another way in which our theory builds on existing identity theory accounts is in its treatment of online versus offline identity. In particular, rather than seeing digital identity as a “front stage” to the backstage of offline identity (Bullingham & Vasconcelos, 2013), SDIT treats these as theoretically interchangeable, with the relative importance and stability of each reflecting the social reality a person inhabits. Indeed, someone who derives most of their meaningful social connection online may feel that they can really “be themselves” online but that they must “perform” more in an offline setting. In this way, the theory can be integrated with identity theory accounts to provide a richer understanding of how social structures shape people’s online behavior.
SDIT also has a number of advantages over existing social identity accounts of digital identity. Although Lüders et al. (2022) and Code and Zaparyniuk (2010) covered ground that is similar to elements of our theory regarding personal versus social identity (P1 and P2), intragroup cohesion (P13), and intergroup conflict (P16), their accounts did not address (a) the impact of social identity on individuals in the digital space (P12); (b) how the social environment (including features of online platforms) shapes identity salience (P3 and P6); or (c) the finer details of intragroup and intergroup dynamics provided by P13, P15, and P17. Most importantly, these accounts were not formalized in a way that is directly testable. Much like the theories on which it is based (i.e., SIT and SCT), SDIT gathers together the threads of the existing literature regarding social identity and digital identity into a single unified framework and a set of predictions that are clearly formulated, well evidenced, and far-reaching.
Indeed, our theory makes several novel contributions to social identity theorizing. For example, although the social identity approach has been applied to various online contexts (e.g., Bliuc et al., 2018; Code & Zaparyniuk, 2010; Lehdonvirta & Räsänen, 2011; Ridout et al., 2012; Van Bavel et al., 2021), this research typically does not differentiate between online and offline versions of the same social identity. In other words, it is tacitly assumed that a group member online is the same as a group member offline. In contrast, we have proposed that there is an online–offline identity continuum that dynamically shapes a person’s psychology in both online, offline, and hybrid contexts (P7–P11).
These hypotheses represent a new direction for social identity research and, as we have seen, are a basis for novel hypotheses (e.g., the prediction that partially distributed work teams may suffer from domain-related faultlines; P14). Another new idea for social identity research resulting from SDIT pertains to the proposition that specific features of online platforms such as digital footprints and relational symmetry might shape identity salience (P6). Previous research has investigated the effects of anonymity and pseudonymity on social identity (e.g., SIDE; Postmes et al., 1998; Reicher et al., 1995), but our analysis takes this further by suggesting that digital environments contain many other such social identity affordances (for a similar discussion, see Khazraee & Novak, 2018). Indeed, we see the identification and exploration of these affordances as a particularly fertile field for future research.
Beyond extending social identity theorizing into the digital domain, SDIT also draws on more recent research into “4E cognition” psychology (Clark, 1999; Newen et al., 2018) and technology development (e.g., AR, VR, metaverse) in offering several substantial expansions to SIT and SCT as theories more generally. Specifically, through P4 and P5 SDIT expands on perceiver readiness and fit, the two central cognitive components of identity salience introduced by SCT. Regarding perceiver readiness, which is a relatively undertheorized aspect of SCT, P4 proposes that memory for group membership does not have to be internal to the perceiver but may also be externalized in the environment through physical and digital artifacts. Furthermore, this externalization differs from other aspects of the perceiver’s environment (e.g., those that relate to fit) such that the externalization of identity should shape salience beyond the contributions of comparative and normative fit (and indeed beyond the contribution of biological memory). Additionally, P5 proposes three ways in which embodiment (specifically sensorimotor experiences) might shape social identity salience through comparative and normative fit. “Externalization” and “embodiment” are novel concepts for social identity theorizing, and they are predicted to operate not only in the digital domain but also in offline and hybrid contexts.
An example of the added explanatory power of SDIT over existing theories can be seen in its ability to provide a novel integration of explanations for the Proteus effect (Yee & Bailenson, 2007). This effect describes the ways in which people tend to take on the psychological characteristics of their virtual avatars. In their review of research on this effect, Coesel et al. (2024) identified six potential psychological mechanisms: self-presentation, deindividuation, priming, cognitive dissonance, embodiment, and perspective. The authors proposed that these mechanisms are likely to work together in some sense to produce the effect but did not propose an integrated account of how this might happen. The propositions of SDIT speak to all of these processes: To the extent that the embodied context of avatar use makes a particular social identity salient (i.e., primes these identities), this should result in deindividuation in line with this identity. Deindividuation and the process of self-stereotyping that this involves (Turner et al., 1987) are predicted to motivate the perceiver to (a) self-present as a prototypical group member, (b) minimize dissonance between the actions of the avatar and their understanding of how a group member should behave, and (c) strive for positive self-evaluation in terms of this identity. Most fundamentally, SDIT predicts that the Proteus effect will be strongest when the user identifies with their avatar (i.e., feels as if they and the avatar share a sense of social identity). Supporting this idea, Latifi et al. (2025) found that the user-avatar bond longitudinally predicted the Proteus effect and that “identification” (feeling “one” with one’s avatar) was the strongest predictor of Proteus effect propensity. Additionally, SDIT goes beyond existing accounts to predict that the Proteus effect will be stronger for social identities that are accessible and provide a good fit for the context of the avatar.
Directions for Future Research
There are numerous ways in which it will be possible for future research to explore and expand SDIT. For example, it will be important to test the propositions of the theory empirically, particularly those that are more theoretically novel in social identity terms. Specifically, although P1 through P3 follow closely from SIT and SCT, P4 through P11 and P18 and P20 represent ideas that have not been tested previously. P12 through P17 also require empirical testing, but these propositions are derived from an established body of social identity literature relating to offline identities and are thus less theoretically innovative. That said, the digital domain clearly provides interesting new opportunities to demonstrate the power and utility of these elements of social identity theorizing. At the same time, the digital domain is an important new context for social identity and one that requires new theories such as SDIT.
Regarding the more theoretically novel propositions of SDIT, one particularly generative approach might be to utilize emerging technologies to test P4 through P6. For example, the concept of externalization (P4) in the context of technology design could be tested by having participants engage with custom-designed digital platforms on which interaction histories are either maintained or not maintained and on which these histories reflect personal versus group interactions. Moreover, the criteria of P4 (i.e., availability, trust, and accessibility) could be manipulated experimentally in such a paradigm. Similarly, studies might use AR and VR technologies to vary the degree of shared embodiment (P5b and P5c) in social interactions as well as the extent to which embodiment is metaphorically appropriate to particular social identities (P5a). For example, participants might interact in a VR environment in which they have an avatar or not and in which they interact with other users who have the same or different kinds and degrees of embodiment. In such a paradigm, metaphorical appropriateness might be manipulated by assigning participants to minimal groups (Tajfel et al., 1971), in which the content of these groups is either matched or mismatched with sensorimotor experiences, or by having participants engage in intergroup relations that are matched or mismatched with these experiences. Having participants interact with digital systems in which relevant features (P6) such as secrecy, algorithmic identification, and interaction structure are manipulated experimentally is another potential avenue for future research.
A key contribution of SDIT is that it does not simply apply existing social identity theorizing (e.g., SCT, SCT, SIDE) to the digital context; it also makes novel theoretical claims beyond these theories. For example, SCT (Turner et al., 1987) states that social identities are functionally antagonistic, which implies, inter alia, that as one’s identity as part of the online version of a group becomes more salient, one’s identity as part of the offline version of that group would become less salient. In contrast, SDIT builds on identity fusion research (e.g., Swann et al., 2009) to propose that the degree of functional antagonism is contingent on the degree of perceived compatibility and integration between these identities such that subjectively compatible identities are not functionally antagonistic and functionally equivalent (fused) identities make each other more salient. Accordingly, if online and offline identities are seen as fully compatible, according to SDIT they should not be functionally antagonistic. Moreover, if one’s personal online identity is fused with a particularly online group, making either of these identities salient should also make the other identity salient.
Last, an important part of testing SDIT is to connect high-level social factors (e.g., culture, political environment, design decisions) to outcomes for individuals, groups, and society through social identity processes. This mediation model is implied by propositions such as P18, P19, and P20, and we see this as an important part of the theory’s contribution beyond existing accounts. The reason that this mediation model is important to test is because if social identity is in fact the key psychological process through which emerging technologies shape well-being (individual or collective), this points to a potential solution to the many problems identified as following from recent technological developments (e.g., young people’s mental health: Haidt, 2024; increasing misogyny: Bates, 2020; social anomie: González-Bailón & Lelkes, 2023). In other words, it is important to test SDIT not only to validate the theory but also because this may provide solutions to a range of technosocial “wicked problems” facing societies.
Conclusion
Our personal and social identities are becoming increasingly digital as we spend more of our lives online. This means that the distinction between who we are as individuals, group members, users of digital platforms, and inhabitants of the material world dissolves further with every passing year. This means too that digital identities are increasingly becoming a core part of “who we really are.” No longer the stuff of science fiction, today we all have digital selves.
In the current article we have outlined a theory (SDIT) that aims to encompass all of these aspects of identity. As we have seen, this theory can be used to understand and predict a diverse range of psychological phenomena across multiple levels of analysis. Moreover, at the same time that SDIT follows the social identity tradition by seeing groups to be as psychologically real as individuals, so too it goes further by also understanding online selves to be distinct from online selves while seeing both to be equally real and valid. In this way, the theory is interested not only in discontinuities in the psychology of material and digital life but also in their intersection. In particular, it speaks to the ways in which the capacity for people to act collectively in both domains can be a vehicle for social change—so that what groups do together digitally brings about new material worlds, and what they do together socially brings about new digital worlds. If we are interested in capitalizing on these dynamics—and avoiding their dystopian potentialities—these are capacities that we need to work together on to fully understand. SDIT, we suggest, is a powerful framework for doing precisely this.
