Abstract
This article investigates the relationship between adolescents’ algorithm literacy and their behavior on social media platforms. Using a mixed-methods approach, including focus groups, a mobile diary study, and a representative survey of German adolescents aged 14–17, we identify three behavioral strategies in response to algorithmically curated content: (1) indifferent behavior, such as passive scrolling; (2) interactive behavior, including liking or sharing content to influence algorithmic outcomes; and (3) preventive behavior, such as blocking content or adjusting privacy settings. The results indicate that higher algorithm awareness is associated with increased indifferent behavior, whereas greater algorithm knowledge correlates with reduced interaction, possibly due to a reluctance to engage with algorithmic profiling. Adolescents’ ambivalent attitudes toward algorithms underscore these behavioral patterns. This work contributes to the understanding of algorithm literacy’s role in adolescents’ social media use and highlights the need for future research to refine the concept of “algorithm-literate” behavior.
Introduction
Adolescents’ use of social media is consistently high, with some platforms even increasing in popularity. In the German context, for example, recent research confirmed rising usage numbers for TikTok and Snapchat from 2022 to 2023, while Instagram maintained its high level of engagement (Feierabend et al., 2023). Social media platforms rely heavily on algorithmic selection processes to determine which content is displayed to users (Just and Latzer, 2017; Oeldorf-Hirsch and Neubaum, 2023a). A prominent example is TikTok’s algorithm-driven “For You” feed, where the appearance of content is primarily determined by its popularity and users’ previous interactions with certain types of content (e.g. Siles et al., 2022). These algorithms play a key role in deciding what adolescents see next on their social media feeds.
Although the relevance of algorithmically curated content continues to spark debate in both academic and public discourse, most platforms either do not disclose or only vaguely explain how such content is allocated (Oeldorf-Hirsch and Neubaum, 2023b). Thus, precise knowledge and a detailed understanding of algorithmic mechanisms remain largely inaccessible to social media users. However, from a digital literacy perspective, responsible users are still expected to have a solid understanding of these mechanisms. This includes recognizing algorithmically selected content, critically evaluating it, and responding to it appropriately.
So far, research has addressed these skills very selectively. Previous studies have primarily focused on adult users’ awareness of the existence and functioning of algorithms (e.g. Zarouali et al., 2021) or on the implicit theories users apply when encountering algorithms in their everyday lives (DeVito et al., 2018; Dogruel, 2021a). Current conceptualizations of algorithm literacy, in contrast, adopt a broader perspective. Aligning with the long-standing research tradition on media literacy, they define algorithm literacy as a holistic construct made up of various skills, including beneficial behavioral strategies (DeVito et al., 2018; Dogruel et al., 2022). However, much of the existing research has predominantly focused on samples of adults. Comparable data for adolescents is lacking. This is a serious shortcoming for the following reasons: Adolescents use social media platforms extensively. For young users in particular, social media is an important source of information about current events: usage data from Germany shows that for users aged between 18 and 24, 34% name social media as their primary source for news (Behre et al., 2025). For youth aged between 12 and 19 years, 37% state that they get to know about current events via Instagram at least several times a week (Feierabend et al., 2024). Heavily relying on the algorithmic systems in social media for informational purposes, which present a specific and fragmented excerpt of news, might influence adolescents’ perceptions and attitudes (Gagrčin et al., 2026; Shin et al., 2022). Especially because their critical reflection skills are still in the process of development (Chai et al., 2020), the relationship between adolescents’ awareness, knowledge, and behavioral strategies when dealing with algorithms in their social media usage needs further investigation.
We addressed these research gaps through a comprehensive mixed-methods study that included focus groups, a semi-standardized diary study, and a representative online survey with a sample of adolescents aged between 14 and 17 years in Germany. Using a holistic conceptualization of algorithm literacy, we analyzed how adolescents’ levels of algorithm awareness and knowledge relate to various behavioral strategies on social media platforms. By doing so, this study empirically investigated the previously neglected question of how digital literacy among adolescent social media users translates into digital literate behavior.
Conceptualizing algorithm literacy
The concept of media literacy has a long research tradition (see Potter, 2022) and is often considered to be the basis for contemporary concepts such as digital literacy (e.g. Helsper et al., 2021). In general, the emergence of digital media has been accompanied by a strong differentiation of concepts and terms regarding media literacy (e.g. news literacy, media and information literacy, ICT literacy, and social media literacy). This diversification has increased the complexity of defining, theorizing, and operationalizing media literacy (see Wuyckens et al., 2022: 169). In recent years, with the growing relevance of algorithms, the concept of algorithm literacy has emerged in scientific discourse.
Similar to media literacy (e.g. Pfaff-Rüdiger and Riesmeyer, 2016), conceptualizations of algorithm literacy adopt a holistic approach. DeVito et al. (2018), for example, describe algorithm literacy as a two-stage process involving awareness of the existence and influence of algorithmic applications and the strategic and practical handling of these applications by users. Dogruel et al. (2022: 118) offer a more detailed four-stage framework, defining algorithm literacy as “being aware of the use of algorithms in online applications, platforms, and services, knowing how algorithms work, being able to critically evaluate algorithmic decision-making as well as having the skills to cope with or even influence algorithmic operations.” On a cognitive level, it thus needs to be differentiated between users’ general awareness of the existence and functioning of algorithms, their more detailed understanding and knowledge about these technical features, and issues of legal regulation of algorithmic decision-making processes. Moreover, this cognitive dimension includes (critical) evaluation, which entails reflecting on the opportunities and challenges algorithms present at both individual and societal levels. Dogruel (2021b: 76–77) further extends this conceptualization to the behavioral level. Coping behaviors include measures such as adjusting privacy settings, counterchecking content to verify its accuracy, and engaging in reflective communication about algorithms. Moreover, programming and coding skills are also recognized as behavioral aspects of algorithm literacy (see Figure 1).
Conceptualization of algorithm literacy (Dogruel, 2021b: 84).
To date, research on algorithm literacy has mainly focused on the first sub-area of these conceptualizations, investigating users’ awareness and knowledge of algorithms. For example, Zarouali et al. (2021) developed the Algorithm Media Content Awareness Scale as a standardized measurement instrument that distinguishes four content-related sub-dimensions of awareness applicable to various online platforms: users’ awareness of (1) content filtering, (2) automated decision-making, (3) the interaction between humans and algorithms, and (4) ethical aspects. Dogruel et al. (2022) created a standardized instrument for measuring awareness and knowledge of algorithms through objective test questions. In addition to awareness and knowledge of algorithms, the attitudes of users toward algorithms and how they work were also occasionally examined. For example, it was shown that positive and negative attitudes toward the function of algorithms to filter certain content are not necessarily mutually exclusive but can coexist, especially among users with a high level of awareness. These users recognize both the opportunities and risks associated with algorithms in their daily lives (Oeldorf-Hirsch and Neubaum, 2023a). Gagrčin et al. (2026) propose another conceptualization of algorithmic media use and algorithm literacy based on their systematic literature review, in which algorithm literacy evolves as part of an experiential learning cycle starting from algorithmic media experiences, followed by reflection, abstraction, and experimentation.
Linking algorithm literacy to behavioral strategies
The approaches and instruments presented illustrate current efforts to operationalize and empirically investigate the concept of algorithm literacy. However, these studies have so far been limited to the cognitive level of the concept, while concrete behavioral strategies and their relationship with algorithm awareness and knowledge have not been systematically investigated. So far, only a few studies have integrated and empirically investigated both the cognitive and behavioral dimensions of literacy. Festl (2021), for example, showed positive relationships between adolescents’ knowledge, abilities, and motivation and different dimensions of social online behavior. For algorithm literacy, (Zhang and Liu, 2024) recently reported findings from an explorative study on older adults’ algorithm literacy in interactions with video recommendations. For usage behavior explicitly meant to influence algorithms, they proposed “folk tactics, which encompass [. . .] the actions of people as laypersons in exerting control over algorithms” (Zhang and Liu, 2024: 13). Such actions taken to deliberately influence algorithmic outcomes might consist in training or “taming” the algorithm as a form of controlling or domesticating the technology (Jones, 2023; Simpson et al., 2022) or as well strategically maneuvering algorithmic content moderation processes by using language such as algospeak, a special form of netspeak “commonly understood as abbreviating, misspelling, or substituting specific words” (Steen et al., 2023: 1). While all these behaviors require some form of algorithm awareness and knowledge, none of the cited studies investigated all three dimensions of algorithmic literacy together.
When behavioral components of algorithm literacy are addressed, their level of intentionality or purposiveness often remains unclear. This ambiguity may stem from the fact that, on social media platforms like TikTok, algorithms are so ubiquitous that using the platform inherently involves interacting with them. This makes it challenging to disentangle the relationship between algorithm awareness and intentional or nonintentional behaviors. While Dogruel (2021b: 84) describes the behavioral dimension of algorithm literacy as “coping behaviors,” we propose reframing this dimension as behavioral strategies. Drawing on Krämer’s (2013: 202) definition in the context of media use, a strategy can be understood as “a pattern of action that is explicitly or, in most cases, implicitly directed toward certain aims, has potentially valuable outcomes, is adapted to a situation, and that is to a certain degree abstract and transferable between different situations.” When exposed to algorithmically curated content on social media, users’ behavioral strategies can be diverse and situation-dependent. These strategies may not always explicitly follow a certain aim (e.g. actively adjusting ad settings to enhance privacy). Instead, they might implicitly address a goal and yield valuable outcomes (e.g. swiping away a post to avoid receiving similar content in the future).
Surprisingly, research in this area has largely overlooked adolescents – a target group that, due to their intensive social media usage, frequently interacts with algorithms and is particularly susceptible to pre-selected content and information (Hasebrink et al., 2021). Both knowledge about the world and critical reflection skills are still emerging in the developmental stage of adolescence (Steinberg, 2005; Weil et al., 2013), which suggests it might be important to take a closer look at algorithm literacy and its different dimensions, particularly among adolescents. Initial studies in related areas, such as news literacy, indicated that high levels of literacy do not necessarily relate to more competent usage behavior (see Vraga et al., 2021). Based on an explorative study on the news literacy of young users in algorithmically curated information environments, Swart (2021) concluded that they do have an intuitive understanding of the existence and functioning of algorithms in their use of news on social media based on their own experiences, but that this understanding does not have an impact on their usage behavior. Based on a representative survey of the Swiss population, Kappeler et al. (2023) reported that only a few Internet users apply strategies against surveillance resulting from algorithmic selection processes, while ignoring automated recommendations seems to be a more common strategy.
Beyond such isolated findings, however, there is still a lack of knowledge on how algorithm literacy, in terms of awareness and knowledge, translates into concrete behavioral strategies when encountering algorithmically curated content. Therefore, the first research gap addressed by the present article is to analyze how algorithm awareness and knowledge are linked to different behavioral strategies when encountering such content on popular social media platforms. A second research gap is addressed by focusing on adolescents as a target group that is particularly vulnerable due to their intense use of social media platforms, such as TikTok, whose functionality is especially aligned to algorithmically selected content.
A mixed-methods approach to adolescents’ algorithm literacy and behavioral strategies on social media
This article presents a thorough and systematic empirical investigation of different aspects of adolescents’ algorithm literacy and their relationship with behavioral strategies in an innovative mixed-methods design. First, focus groups were conducted to explore adolescents’ basic understanding of the perceived content, characteristics, and authorship of algorithms in social media (Study 1). Adolescents were then invited to participate in a semi-standardized mobile diary study aimed at an in-depth investigation of their perception of algorithms in their everyday lives, as well as their attitudes and behavioral strategies (Study 2). Finally, based on the findings of Studies 1 and 2, we conducted a representative online survey assessing adolescents’ awareness, knowledge, attitudes, and behavioral strategies on social media (Study 3). The central research interest guiding our work was to disentangle how different aspects of algorithm literacy, namely awareness and knowledge, relate to different dimensions of behavioral strategies when encountering algorithmically curated content. Based on previous findings highlighting the role of attitudes in that context, especially among users with a high level of algorithm awareness, we also aimed to explore the role of attitudes, both in relation to the general perception of algorithms as well as a control variable, when it comes to concrete behavioral strategies when encountering algorithmically curated content. Therefore, we first asked about adolescents’ perceptions of and attitudes toward algorithms in their everyday social media usage (RQ1). In the next step, and as a prerequisite to investigating the relationship between different aspects of algorithm literacy and concrete behavioral strategies, we aimed to identify different dimensions of behavioral strategies when adolescents encounter algorithmically curated content online (RQ2). Finally, we asked how adolescents’ algorithm awareness (RQ3) and algorithm knowledge (RQ4) relate to different kinds of behavioral strategies when encountering algorithms on social media platforms.
To answer our research questions, we used a mixed-methods approach combining focus groups (Study 1) with a follow-up mobile diary study (Study 2) and a representative online survey of adolescents between 14 and 17 years of age in Germany (Study 3). All three studies were fully approved by the faculty’s ethics committee. Adolescents’ participation in all studies was completely voluntary, and nonparticipation had no disadvantages. Before participation, the adolescents and their legal guardians were informed about the study content, objectives, methodological modules, and data processing. Participation was only possible when adolescents and, where necessary, 1 legal guardians consented to the study regulations. 2
Study 1: Focus groups
In the first step, six focus groups, each comprising three to five adolescents, were conducted in October and November 2023 to investigate their basic understanding of algorithms.
Procedure and sample
Adolescents were recruited via WhatsApp groups, sports clubs, and a youth parliament in and around a major German city. Following the theoretical sampling approach, the groups were built according to the gender and age of the interviewees. In total, six focus group discussions were conducted, involving 24 adolescents (11 males and 13 females). Participants ranged in age from 14 to 17 years, with the following distribution: 14 years (n = 1), 15 years (n = 8), 16 years (n = 11), and 17 years (n = 4). A significant portion of the participants attended higher education schools in Germany (n = 21; for an overview of the sample, see Table 1). The focus group discussions lasted between 45 and 83 minutes. All sessions were recorded, transcribed verbatim, and fully pseudonymized to ensure confidentiality. Transcription was performed automatically using f4x software, with all transcripts being manually reviewed by a trained student assistant to ensure correct speaker differentiation. During this step, all personal details of the adolescents were removed (e.g. school name, place of residence) to ensure anonymity. All participants were given a new first name as a pseudonym, whereby the pseudonyms of the adolescents who took part in the same focus group began with the same letter (e.g. focus group A, pseudonyms Alina, Antonia, Amaya, Alisia).
Sample characteristics of Study 1 (focus groups) and Study 2 (diary study).
Note. FG = focus groups, DS = diary study.
Measures and analytical strategy
All focus groups were based on a semi-standardized interview guideline. All pseudonymized interview transcripts were analyzed based on theory using MAXQDA. The category system was derived from the current state of research and included the use of social media (e.g. intensity, habitualization, privacy); knowledge of algorithms (everyday relevance, concept, authorship, avoidance strategies, probability of contact); algorithm evaluation (attitudes, advantages, disadvantages, perceived usefulness, perceived ease of use); and handling of algorithms (behavioral strategies). It was inductively supplemented with categories derived from the transcripts (e.g. differentiation between social media platforms used, possible effects of algorithms, emotions they elicited). The resulting category system formed the code system in MAXQDA. Each transcript was imported into MAXQDA and read several times by the authors. Participants’ statements were assigned to the relevant (sub)categories. Possible ambiguities were discussed within the team of authors. The analysis aimed to identify similarities and differences between the participants to understand their perceptions and assessments, find explanations, and contextualize results. Anchor statements were selected to substantiate the results.
Results
Addressing RQ1, we first explored adolescents’ perceptions of and attitudes toward algorithms in social media. The focus groups confirmed that social media platforms are an essential part of adolescents’ everyday lives, where they encounter content they believe is based on algorithms (e.g. personalized advertising). They often feel that the algorithm knows their interests precisely; as Thomas (15) said, “it knows me.” This results in adolescents being exposed to content they are not explicitly searching for that aligns with their interests – such as related sports content for someone interested in sports. However, it can also lead to content being suggested that is irrelevant, rejected, or even unsettling to them (e.g. “misogynistic videos”, as noted by Marie, 15). On multiple occasions, without being asked, adolescents mentioned political content that contradicted their views, such as videos glorifying violence, which they considered to be dangerous, but sometimes also trivialized it. This type of content was most commonly encountered on TikTok.
The focus groups further showed that the adolescents mentioned various usage situations in which they encounter algorithms. However, it also became clear that they were not always sure about the concept and authorship of algorithms. Instead, they tried to paraphrase the concept during the focus group interviews. Oskar (17), for example, explained that “they register what you do, what you like, and then it is stored, and then it is offered to you or displayed to you so that it is more fun. However, I cannot explain that right now.” This half-knowledge is also evident in phrases such as “I think” (Otto, 16; Nora, 16), “I do not know” (Olaf, 16; Levi, 16; Miriam, 14), and “it is difficult to describe” (Alisia, 17). They also characterized algorithms using features that they attribute to the concept (e.g. hashtags, analysis of their search terms, connections between different platforms, and listening to their conversations).
Overall, adolescents’ attitudes toward algorithms were ambivalent, and they named both benefits and drawbacks associated with using algorithm-based technologies. Thomas (15) talked about the Spotify algorithm he deemed as “actually very positive, because it helps enormously to get to know new music [. . .] if I didn’t have an algorithm, it would be difficult to find new bands.” In addition to such positive evaluations, adolescents reflected possible negative outcomes associated with algorithms. For example, they talked about “getting into some kind of ‘rabbit hole,’” when encountering algorithmically curated content on social media: “you’re right in the middle of it and it’s tough to get out of it” (Miriam, 14). As these examples show, adolescents not only see advantages or disadvantages in algorithms but also weigh up both sides.
Addressing RQ2, we next focused on adolescents’ behavioral strategies when encountering algorithmically curated content on social media. In the focus groups, adolescents reported engaging in tentative, interactive, and preventive behaviors when encountering algorithms in social media. Adolescents discussed tentative behaviors as actions they could imagine to be helpful in training algorithms in social media. To ensure that the algorithm learns what corresponds to their interests, they would, for example, “comment on something, take part in a competition, or save something” (Oskar, 17). Sharing content with others, reposting, and republishing posts can also help train the algorithm. Adolescents report applying the same procedures, also working in the other direction, aiming to ensure that the algorithm learns what content not to display. Both variants of tentative behavior made it clear that the adolescents are unsure about the extent to which algorithms can be circumvented, as Nora (16) notes: “But I do not think you can circumvent it,” a statement also confirmed by Niklas (17) and Nina (16). Thus, the only way to avoid the algorithm is not to use the respective applications.
The adolescents reported applying interactive behaviors specifically to train the algorithm. These behaviors included liking content they enjoyed, sharing posts with friends, and even taking explicit actions to influence the algorithm’s recommendations. Olaf (16) explained that he likes specific videos on TikTok to encourage the algorithm to show him more similar content. “I just like videos that I enjoy, hoping to get more of them. That is the principle of a social media platform like this” (Oli, 16). In addition, adolescents reported reloading the “For You” page on TikTok and watching videos several times to ensure new content relevant to their interests is displayed. This could even involve intentional actions such as searching for preferred videos and letting them play on repeat. As one participant noted, Then I let the video play, put my smartphone away and just quickly do other things so that the video is played through very often, so that the algorithm thinks I’m watching it 20 times because I think it’s really cool. (Oli, 16)
According to their explanations, adolescents apply preventive behavior to circumvent algorithms, while tentative and interactive behaviors are applied to (try to) train the algorithm to display more content corresponding to adolescents’ interests or bypass this content. Adolescents apply preventive behavior by clicking on “I am not interested” (Otto, 16; Melina, 15; Miriam, 14; Alina, 16), blocking accounts (Olaf, 16; Mia, 15), not posting comments (Oskar, 17), closing the app, or even switching off their mobile phone. Through these actions, they try to avoid content selected by algorithms.
Study 2: Mobile diary study
Adolescents participating in the focus groups were invited to participate in a semi-standardized mobile diary study that took place immediately after the group interviews. The diary study aimed to conduct an in-depth survey of adolescents’ perceptions of algorithms and their concrete behavioral strategies when encountering algorithmically curated content directly in their everyday use.
Procedure and sample
A total of 22 adolescents followed our invitation, downloaded the scientifically validated MeTag App (Mascheroni and Zaffaroni, 2022) on their smartphones, and filled out a semi-standardized diary entry each evening on seven consecutive days in November 2023. A daily reminder to complete the diary was sent via the app. 3
Measures and analytical strategy
At the beginning of each diary entry, participants should first state which social media platform they used most that day. Each diary entry was then structured with three questions: (1) Do you think that an algorithm at least once played a role in your usage of this platform today? (yes/no/don’t know); (2) If yes, how did you recognize the algorithm? (open-ended); and (3) How did you react to it? (open-ended). Adolescents were able to edit and delete the entries in their media diaries throughout the study period. Data was analyzed at the level of usage situations, that is, the concrete occasions adolescents perceived an algorithm to play a role in their social media usage on that day. A total of 143 usage situations were gathered, where Instagram (n = 38), TikTok (n = 31), and WhatsApp (n = 20) were mentioned as the most frequently used platforms per day. Based on the diary questionnaire, we developed three broad categories for analysis: use of social media, perception of algorithms, and reaction to algorithms. The “reaction to algorithms” category was further divided into two subcategories – attitudinal reactions and behavioral strategies – derived inductively from the data material.
Results
In the diary study, we analyzed a total of 143 usage situations of adolescents’ everyday social media use. For 101 of these usage situations, adolescents reported having perceived an algorithm; in 38 situations, no algorithm was noted, and in 4 situations, the respective adolescents were unsure. By far, the most frequent platforms where young people took note of algorithms were TikTok and Instagram. On Instagram (n = 92%) and TikTok (n = 100%), adolescents almost always perceived algorithmically selected content, while this was only seldom the case for WhatsApp (10%). Overall, algorithms were rated positively as useful and valuable recommendations for encountering exciting content, as the following quote from Alina’s (16) diary shows: It shows that the app is also geared towards the content posted by the user. I like this because the aesthetics of the feed match my content. I automatically use the app more because I know I can rely on getting customized suggestions.
However, adolescents also reported more negative attitudes of being “surprised and also a bit scared” (Alisia, 17) when perceiving algorithmically selected content that was precisely tailored to their interests. Participants also indicated that – over time – algorithmic recommendations make them sluggish and – in some cases – even reinforce negative moods. Occasionally, we saw contradictory statements in the diary entries, suggesting that adolescents struggle in their time management due to the algorithmic-based recommendations, meaning that they often spend more time on social media than intended: “The TikTok algorithm is extremely good, but it is also very addictive, which is why I want to use TikTok less in the long term” (Thomas, 15).
In the diary study, patterns identified in the focus groups were further confirmed in adolescents’ everyday social media use. The adolescents reported various behavioral patterns when encountering algorithmically curated content, ranging from just scrolling down, liking, or sharing content to even closing the app (when confronted with negative content). Nina (16) described a situation where content shown to her aligned perfectly with the topics she was focused on that day, and her reaction was, “I noticed it and thought it was strange, but kept watching.” The findings further suggested that adolescents often deliberately tried to “steer” the algorithm by liking certain content more, less, or not at all. Olaf (16), for example, mentioned the following kind of interactive behavior regarding TikTok: “If I don’t like them, I like other pictures with different aesthetics to get more of them suggested.”
Study 3: Online survey
Building on the findings of Studies 1 and 2, we conducted a representative, standardized online survey of adolescents aged between 14 and 17 years in Germany. The aim of Study 3 was to empirically investigate adolescents’ exposure to algorithms, their awareness and knowledge of them, their attitudes toward algorithms, and their specific behavioral strategies when engaging with algorithmically curated content on social media.
Procedure and sample
The survey was conducted in cooperation with a market research institute specialized in research with children and adolescents in December 2023. Recruitment initially took place via the parents, who were informed about the study and provided information on certain characteristics of their child (or children). Once a target child had been selected according to the quota criteria (gender, age, education, location, and migration background) and the parents of children under the age of 16 had actively consented to the child’s participation, the child itself was brought to the device used. The adolescents were also informed about the study and were asked for their active consent. In total, 610 adolescents (50.3% male), aged between 14 and 17, completed the standardized online questionnaire. The intended equal distribution between the age groups of 14-, 15-, 16-, and 17-year-olds (25% each) was met (M = 15.5, SD = 1.12). The formal education of the respondents was recorded via the school-leaving qualification they were aiming for or, in the case of a qualification already achieved, via their highest school-leaving qualification. In total, 17.4% of adolescents were recorded as having a vocational qualification, 34.0% as having an intermediate school-leaving certificate, and 48.9% as having a vocational or university entrance qualification. This roughly corresponds to Germany’s target rates for lower, intermediate, and higher school-leaving qualifications. Moreover, 70.5% of the adolescents surveyed lived in smaller towns (fewer than 100,000 inhabitants), and 15.1% had a migration background.
Measures and analytical strategy
The online survey comprised questions on adolescents’ general use of social media as well as questions related to algorithms in social media (perception, awareness, knowledge, attitudes, and behavioral strategies). 4 Finally, the extent to which algorithms were discussed in different socialization contexts was measured. All analyses were calculated using the lavaan package (version 0.6-19) in R (Rosseel, 2012). Maximum likelihood estimation (ML) was employed for all model estimations. The fit of each model was evaluated using standard fit indices, including CFI, RMSEA, and SRMR. Both global and specific fit indices were evaluated in line with Byrne (2010). 5
First, adolescents were asked about their frequency of use of different social media platforms on a 6-point scale ranging from “never” (0) to “several times a day” (5) (Hasebrink et al., 2021). Social media platforms most frequently used by adolescents were YouTube (M = 3.83, SD = 1.25), Instagram (M = 3.71, SD = 1.79), and TikTok (M = 3.69, SD = 1.82).
The main part of the questionnaire covered adolescents’ perceptions, awareness, knowledge, attitudes, and behavioral strategies when encountering algorithms in social media. 6 To assess their perception of algorithms, we asked, “How often do you think you are shown content on the following platforms because an algorithm has recommended it?” Adolescents answered for each platform they previously indicated to use on a 5-point scale (0 = “never” to 4 = “very often”). This item was self-developed based on our procedure in the online diary (Study 2).
For algorithm awareness, we used a scale proposed by Zarouali et al. (2021), covering 13 items, such as “Algorithms are used to recommend content to me on social media” or “The content that algorithms recommend to me on social media depends on my online behavioral data.” Responses were recorded on a 5-point scale (1 = “do not agree at all” to 5 = “fully agree”). 7 A confirmatory factor analysis (CFA) resulted in a one-factor solution (α = .88) of 12 items 8 with an acceptable model fit: χ2(54) = 92.36, p = .001, CFI = .982, TLI = .978, RMSEA = .034, and SRMR = .027.
Furthermore, we tested adolescents’ knowledge of algorithms in social media, asking them to rate eight statements as either true or false (e.g. “I can influence algorithms with my internet usage behavior,” “When searching things online, results displayed may vary from person to person despite the same search entry”; see Dogruel et al., 2022). 9 A sum score of correct answers was built ranging from 0 to 8 (M = 4.80, SD = 1.95).
Regarding their behavioral strategies when encountering algorithmically curated content on social media, we asked adolescents how often they apply different behaviors (e.g. “go on scrolling/swiping,” “sharing the content with others,” and “reporting or blocking content” on a 5-point scale ranging from 0 = “never” to 4 = “very often”). 10 These 10 items were derived both from prior research and theorizing (Dogruel, 2021b; Kappeler et al., 2023) and were developed based on our findings in Studies 1 and 2.
As a control variable, attitudes toward algorithms were assessed directly on a semantic differential (e.g. “bad–good,” “harmful–beneficial”) from 1 to 5 via the question “How do you feel about algorithms being used to display content on social media?”. In addition, attitudes were also measured indirectly via statements such as “Algorithms are fair” or “Algorithms are helpful to me” on a 5-point scale, ranging from 1 (“do not agree at all”) to 5 (“fully agree”; Silva et al., 2022). 11 A confirmatory factor analysis (CFA) confirmed a two-factor structure of positive (α = .87) and negative indirect attitudes (α = .70). The CFA confirmed an acceptable model fit: χ2(26) = 93.87, p < .001, CFI = .964, TLI = .951, RMSEA = .066, and SRMR = .036.
Results
In the representative online survey, we found that adolescents noted algorithms most frequently on TikTok (M = 3.38, SD = 0.79), followed by Instagram (M = 3.22, SD = 0.81) and YouTube (M = 3.15, SD = 0.90). Overall, adolescents showed rather positive direct attitudes toward algorithms in social media, deeming them as more funny than sad (M = 3.37, SD = 0.92), rather useful than harmful (M = 3.31, SD = 1.10), and rather good than bad (M = 3.21, SD = 1.14). Notably, their ratings regarding the aspects of dangerous versus safe (M = 3.03, SD = 1.10) and helpful versus annoying (M = 3.04, SD = 1.25) were a bit less, but still rather positive (see Table 2). Regarding indirect attitudes, adolescents showed somewhat stronger negative (M = 3.37, SD = 0.88) than positive attitudes (M = 2.94, SD = 0.87).
Adolescents’ direct attitudes toward algorithms.
Note. Items were answered on semantic differentials, ranging between 1 and 5.
Building on the findings of Studies 1 and 2, we tried to identify dimensions of behavioral strategies when encountering algorithmically curated content based on our representative online survey. Thus, we conducted a PCA on 10 items assessing different dimensions of behavioral strategies that revealed three factors (KMO = .807, χ2(45) = 1752.63, p < .001), which accounted for 65.80% of the total variance: indifferent behavior in terms of just go on scrolling/swiping, interactive behavior in terms of sharing, liking, saving, and looking at/trying/buying content, and critical behavior in terms of reporting or blocking content, unfollowing accounts, changing privacy or ad settings, and counterchecking content. A subsequent CFA showed an acceptable model fit for the identified three-factor structure: χ2(30) = 144.37, p < .001, CFI = .941, TLI = .912, RMSEA = .079, SRMR = .050. 12
To explain the relationship between different behavioral strategies and algorithm awareness (RQ3) and knowledge (RQ4), we applied a structural equation model (SEM). The latent construct algorithm awareness was represented by 12 manifest items, while algorithm knowledge was represented by a sum score and behavioral strategies by the factors identified in the CFA (see Figure 2). In addition, we controlled for positive and negative indirect attitudes toward algorithms, age, gender, and formal education. Overall, the model showed acceptable fit values: χ2(517) = 1032.68, p < .001, CFI = .926, TLI = .914, RMSEA = .040, and SRMR = .055.
Research model.
First, by examining the relationships between the different dimensions of behavioral strategies, we found no correlations between indifferent and interactive behavior (cov = .052, p = .356) and between indifferent and critical behavior (cov = −.057, p = .273), but a strong correlation for interactive and critical behavior (cov = .415, p < .001). In addition, there were strong correlations between algorithm awareness and knowledge (cov = .428, p < .001) as well as awareness and negative attitudes (cov = .460, p < .001), and a weak correlation for awareness and formal education (cov = .129, p = .003). Knowledge about algorithms was correlated with negative attitudes (cov = .160, p = .002). Positive attitudes were negatively related to higher formal education (cov = −.257, p < .001), and negative attitudes correlated positively with gender (cov = .098, p = .040), indicating that female adolescents had somewhat stronger negative attitudes toward algorithms compared to male adolescents.
Regarding the relationship between adolescents’ algorithm awareness and different dimensions of behavioral strategies (RQ3), we observed a positive relationship for indifferent behavior (β = .126, p = .028) but no significant effects for interactive (β = −.045, p = .446) or critical behavior (β = −.103, p = .148). This suggests that the more aware adolescents are of algorithms, the more likely they are to engage in indifferent behaviors, such as passively scrolling or swiping through content. Looking at the relationship between knowledge about algorithms and different dimensions of behavioral strategies (RQ4), on the contrary, we found no relationship for indifferent behavior (β = −.065, p = .124) and negative relationships for interactive (β = −.100, p = .022) and critical behavior (β = −.105, p = .047). This indicates that as adolescents become more knowledgeable about algorithms, they are less likely to engage in interactive behaviors, such as sharing, liking, or saving content, and are also less likely to adopt critical behaviors, such as adjusting ad or privacy settings or reporting and blocking content.
Age, gender, and formal education did not have an effect on the different dimensions of behavioral strategies (see Table 3). However, attitudes toward algorithms did influence these behaviors. For indifferent behavior, a negative effect of positive attitudes (β = −.227, p < .001) and a positive effect of negative attitudes (β = .252, p < .001) were observed. This means that the more negative adolescents’ attitudes toward algorithms, the more likely they were to scroll or swipe past content. For interactive behavior, both positive (β = .821, p < .001) and negative attitudes (β = .247, p = .001) had positive effects. Similarly, for critical behavior, both positive (β = .407, p < .001) and negative attitudes (β = .547, p < .001) had positive effects. These findings indicate that adolescents’ attitudes, whether positive or negative, influence their interactive and critical behaviors when encountering algorithms on social media, with both types of attitudes driving engagement in these behaviors.
Explaining the relationship between adolescents’ algorithm literacy and behavioral strategies.
Note. N = 610. Standardized path coefficients (β) indicated.
<.01. ***<.001.
Discussion
The use of social media platforms such as TikTok, Instagram, and Snapchat is an integral part of adolescents’ lives today. These platforms rely heavily on algorithms to feature, filter, and arrange content. Such algorithmic selection decisions are influenced by the popularity of content, as well as users’ previous interactions with similar content. In public and scientific debate, one main issue in this context is how to foster users’ reflected and responsible handling of such algorithmically curated content. Previous research on algorithm literacy mainly addressed the cognitive level of the concept, investigating (mostly) adult users’ awareness and knowledge of algorithms. In contrast, behavioral strategies when being exposed to such content, as well as the question of how algorithm literacy translates into different behaviors, have not yet been systematically investigated. In the present work, we explicitly addressed these points by applying a mixed-methods approach, integrating focus groups (Study 1) with a mobile diary study (Study 2) and combining this qualitative approach with a representative online survey among adolescents aged 14–17 years in Germany (Study 3). This innovative approach allowed us to comprehensively analyze the relationship between algorithm literacy and various behavioral strategies for engaging with algorithmically curated content. By integrating in-depth qualitative insights with broad, representative quantitative data, this study advances existing research (e.g. by Siles et al., 2022) and contributes to a highly relevant and socially significant field.
First, we asked how adolescents perceive algorithms in their everyday social media usage and how their attitudes toward algorithms can be described (RQ1). Our findings showed that adolescents frequently identify the presence of algorithms in their daily interactions with social media. While they seem to have a good basic understanding of their existence and role in social media platforms (e.g. for featuring personalized content), they struggle when it comes to clear-cut definitions or questions about the authorship of algorithms. Across all three studies, adolescents consistently recognized algorithms, particularly on TikTok and Instagram. Their attitudes toward algorithms may best be described as ambiguous: on the one hand, they find it positive to see content that matches their interests without actively searching for it. On the other hand, they also hold negative attitudes toward some aspects of algorithms, for example, when the tailoring of personalized content is perceived as creepy, annoying, or maybe even dangerous (see also Oeldorf-Hirsch and Neubaum, 2023a).
Looking at concrete behavioral strategies when encountering algorithms in social media (RQ2), a complex picture emerged. Synthesizing the results from all three studies, we could differentiate three dimensions of adolescents’ behavioral strategies. First, there is usage behavior that can be described as tentative or indifferent, and that is mainly characterized by adolescents’ uncertainty about how algorithms can be influenced by them as users at all. As a result of that uncertainty, indifferent behavioral strategies were evident in, for example, just scrolling through or swiping on social media. The other two dimensions we identified contrast with that indifference, in that they relate to more active behavior. On the one hand, we observed behavioral strategies that can be described as interactive, where adolescents actively engage with algorithmically curated content, for example, by sharing, liking, or saving it. This behavior reflects their (conscious or unconscious) efforts to train the algorithm to show more or less of specific types of content (see Jones, 2023; Simpson et al., 2022). On the other hand, we identified behavioral strategies that can be described as preventive or critical. These behavioral strategies aim at actively circumventing algorithms, for example, by blocking content, unfollowing accounts, or changing ad and privacy settings. Since, in the context of algorithm literacy, previous research has mainly focused on the level of cognitive skills (e.g. DeVito et al., 2018; Dogruel, 2021b; Zarouali et al., 2021) but neglected their implementation in concrete behavioral strategies, identifying different dimensions of adolescents’ behavioral strategies when encountering algorithms in social media is an insight in itself. The dimensions of behavioral strategies identified reflected various degrees of intentionality: while tentative or indifferent behavior comes with comparably low intentions, interactive as well as preventive or critical behavior are clearly geared toward actively influencing the process of algorithmic selection decisions. These findings stress the applied concept of behavioral strategies (see Krämer, 2013) in the context of “literate behavior” that can be very diverse depending on the respective usage situations and that might vary between very explicit strategies following clear behavioral goals and implicit strategies characterized by habitual usage behavior.
The key question guiding our research was how these different kinds of behavioral strategies relate to algorithm awareness (RQ3) and algorithm knowledge (RQ4). Here, too, the findings are complex: while algorithm awareness positively influenced indifferent behavior, it did not significantly affect interactive or critical behaviors. Interestingly, greater algorithm knowledge was negatively associated with both interactive and critical behaviors. Specifically, adolescents who were more aware of algorithms in their social media use were more likely to continue scrolling or swiping without engaging further. Given that platforms like TikTok heavily rely on algorithms and encourage rapid scrolling through content, this finding is perhaps unsurprising. While algorithm awareness is widespread among users, it does not necessarily lead to explicit behavioral strategies, such as changing ad or privacy settings or engaging critically with content. For algorithm knowledge, another crucial component of algorithm literacy, we observed an intriguing pattern: greater knowledge was linked to less interactive and less critical behavior. This suggests that adolescents with more algorithm knowledge may deliberately avoid interacting with algorithmically curated content – possibly because they do not want to “feed” the algorithm. This points in the same direction as the results for awareness, namely that literate behavior might not only be about applying but also about not applying certain kinds of behavioral strategies.
More surprisingly, more knowledge of algorithms was also related to less preventive or critical behavior. One possible explanation here could be that adolescents with high knowledge about algorithms are well aware that algorithmic decision-making is at the core of social media platforms like TikTok and, therefore, apply rather strict ad or privacy settings in the first place when registering for the platform, so they no longer need to take preventive or critical action in single usage situations.
Ambiguities also became apparent when controlling for the role of attitudes toward algorithms in this context: for all three dimensions of behavioral strategies identified, both positive and negative attitudes played an important role. The more negative adolescents’ attitudes toward algorithms were, the more likely the corresponding content was scrolled or swiped onward (indifferent behavior). For interactive and preventive/critical behavior, the results of the online survey (Study 3) reflected ambivalences that already emerged in the focus groups (Study 1). Both positive and negative attitudes toward algorithms were positively related to interactive behaviors, such as sharing, liking, or saving algorithmically recommended content, as well as critical (preventive) behaviors, such as reporting or blocking content, unfollowing accounts, or changing privacy or ad settings. Thus, adolescents who were undecided about their evaluation of algorithms on social media also behaved ambiguously when encountering such content. On the one hand, they highly engaged with this content when it well fitted their interests. On the other hand, they quickly tried to prevent content when it was out of their preferences.
Taken together, our findings provide the first holistic insight into the relationship between adolescents’ algorithm literacy and their behavioral strategies on social media. The approach pursued offered several advantages: adolescents could freely express their experiences with algorithms in the focus groups in great detail (Study 1). The focus groups then informed the setup of the mobile diary study (Study 2), which allowed us to investigate adolescents’ perceptions of algorithms and their behavioral strategies directly in their everyday lives. Finally, both the focus groups and the mobile diary study informed the representative online survey (Study 3), which served as an initial quantitative validation of the trends identified in the two qualitative studies.
Limitations and directions for future research
We consider the work presented as an essential starting point for further research on algorithm literacy and its relationship to behavioral strategies. However, there are several limitations that must be considered when interpreting the findings. In both the focus groups (Study 1) and the mobile diary study (Study 2), adolescents from lower-education backgrounds were underrepresented compared to those from higher-education backgrounds. This limitation was addressed in the online survey (Study 3), where adolescents from lower-education backgrounds were recruited offline to ensure a sample more representative of the general adolescent population in Germany. As a second limitation, the online survey was based on cross-sectional data that did not allow for any indications of whether more algorithm literacy would be related to different behavioral strategies of adolescents over time. This is where we see a promising avenue for future longitudinal studies that could relate algorithm literacy at one time with respective behavioral strategies at later times. Third, we could not fully reproduce the model of algorithm literacy proposed by Dogruel (2021b). Although we relied on validated measures for algorithm awareness (Zarouali et al., 2021) and knowledge (Dogruel et al., 2022), we had to slightly adapt them to use them in a survey with adolescents. Furthermore, we could not investigate the role of critical evaluation of algorithms, as proposed by Dogruel et al. (2022), in the context of behavioral strategies. Still, some indications on this issue might be derivable from the findings on adolescents’ indirect attitudes toward algorithms on social media.
Conclusion
The overarching question guiding this work was if and how being algorithm literate relates to actually behaving literate. To the best of our knowledge, the present studies are the first to empirically analyze the relationship between adolescents’ algorithm literacy and actual behavioral strategies in the context of social media, where algorithms are omnipresent. By applying an innovative mixed-methods approach, we provide an integrated, holistic perspective on adolescents’ algorithm literacy in terms of their perceptions of and attitudes toward algorithms as well as their behavioral strategies in their everyday lives.
Our findings suggest that algorithm literacy is related to different dimensions of implicit and explicit behavioral strategies. While this is an essential first step in exploring the role of algorithm literacy in the context of social media use, we still need to further investigate and discuss what behaving literate actually means. In the present data, we observed that adolescents with higher algorithm awareness – as one aspect of algorithm literacy – tended to apply more indifferent behavior, such as just scrolling through or swiping when encountering algorithmically curated content in social media. Thus, the question needs further consideration of what precisely denotes literate behavior – and if doing “nothing,” in terms of neither taking interactive nor restrictive action, can be considered a literate behavioral strategy as well, at least in the context of social media usage.
This is also an important question for the practical promotion of algorithm literacy. While explicit behavioral strategies, such as preventive and critical actions, can be more easily implemented within measures to promote algorithm literacy, implicit strategies – effective in certain situations – seem to be more individualized, tied to a person’s habitual usage behavior, and therefore harder to influence or train. However, the significant relationship between algorithm knowledge and both interactive and critical behaviors suggests that the transfer of knowledge about algorithms plays an important role in fostering algorithm-literate behavior. Especially for adolescents, school is one place where such knowledge can be imparted, independent of a person’s socio-economic background or existing skills. Although in such formats, there should also be space for discussions about ethical, legal, and other implications associated with the use of algorithms, strengthening a reflective and nuanced perspective on this complex of issues. This can be achieved by formally including such measures in school curricula as well as the appropriate training of teachers. In addition, parents can be sensitized by information events or materials to openly talk with their children about their everyday experiences with algorithms in social media and possible behavioral strategies. Holistically interlocking the perspectives of teachers, parents, and adolescents themselves might be most fruitful in systematically fostering the algorithm literacy of young users.
Supplemental Material
sj-docx-1-nms-10.1177_14614448261446589 – Supplemental material for Being literate, behaving literate? A mixed-methods approach to adolescents’ algorithm literacy and behavioral strategies on social media
Supplemental material, sj-docx-1-nms-10.1177_14614448261446589 for Being literate, behaving literate? A mixed-methods approach to adolescents’ algorithm literacy and behavioral strategies on social media by Larissa Leonhard, Ruth Wendt and Claudia Riesmeyer in New Media & Society
Footnotes
Ethical considerations
All three studies presented in this paper were fully approved by the Research Ethics Committee of the Faculty of Social Sciences at LMU Munich (GZ 23-09).
Consent to Participate
Adolescents’ participation in all studies was completely voluntary, and nonparticipation had no disadvantages. Before participation, the adolescents and their legal guardians were informed about the study content, objectives, methodological modules, and data processing. Participation was only possible when adolescents and – where necessary – legal guardians consented in written form to the study regulations.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by the Bavarian Regulatory Authority for New Media (Bayerische Landeszentrale für neue Medien, BLM).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
The material of all three studies and the data of the online survey (Study 3) are available on the Open Science Framework (OSF; 
). Data of Studies 1 and 2 cannot be shared publicly due to the privacy of individuals who participated in the focus groups and online diary. The data will be shared on reasonable request to the corresponding author.
Supplemental material
Supplemental material for this article is available online.
Notes
Author biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
