Learning From the Mess: What Happens in Social Media Elicitation Interviews

Abstract

This article examines what happened during 50 semi-structured interviews that were designed to focus on the visual content of interviewees’ social media feeds. During these encounters, interviewees were invited to talk about content that appeared on their social media apps while showing their phone screens to the interviewer. Rather unexpectedly, the researchers experienced this sharing of screens as “messy” and disruptive to the interview. From that reflection, we argue that this mess was generated by the friction between a distracted disposition toward social media and the conventions of semi-structured interviews—a friction that is itself a form of research evidence.

Introduction

Social media is moving toward increasingly visual, multimedia formats that combine image, text, video, graphics and sound, and the vast mass of visual content invites both dispersed attention across the whole range of media and focused attention to specific pieces of content that capture the user’s attention. This seemingly ever-increasing sensory and semiotic complexity of social media presents new avenues of inquiry but also challenges to researchers who are interested in the relationships between lived experience and technology. This article critically examines interactions with visual social media in semi-structured interview settings and how this relates to broader debates about visuality and visual culture on social media. To do this, we draw on 50 interviews conducted as one part of an international collaborative research project “Trust and Visuality – Everyday Digital Practices” (TRAVIS), which aims to explore how young adults express and experience trust in social media imagery relating to everyday health and well-being. The project adopted a range of qualitative methods (for a discussion of student autoethnographies, see Tiidenberg et al., 2025); however, the focus of this article is specifically on in-person interviews with social media users in Estonia, Finland, and the United Kingdom.

An ever-growing body of empirical work demonstrates the ubiquity of images on the internet (Hand, 2017). From early dating sites to recent image-based apps, in static profile pictures or memetic short-video cultures, visual content and communication serves a central function in online meaning-making (Ellison et al., 2006; Garde-Hansen, 2020; Schreiber, 2017; Tiidenberg, 2018; Van Dijck, 2008; Vizcaíno-Verdú & Abidin, 2022). This shift provides the empirical focus of our research project, which specifically considers how trust is evaluated and enacted through and with socially mediated visual communication. This article focuses on the methodological impact of incorporating glimpses of that very same socially mediated visual communication into interview situations, when interviewees were asked to show and talk about the social media content visible on their phone screens during the interview.

In our study, inviting participants to reflect on social media content related to everyday health and well-being gave them the ability to ground their reflections on visual trust within a particular realm of lived experience. The prerequisite for taking part in the interview was that the participant had seen health or well-being-related content on any social media platform. We were open to participants’ own interpretations of what counted as a social media platform and as everyday health and well-being content, meaning that participants defined this in a way that was most relevant to their own lived experience. Interviews therefore referred to a wide variety of health topics that were of personal or general interest to the participant, including (but not limited to) nutrition and diet, skincare, fitness, mental health, sexual health, specific illnesses, public health messaging (Covid-19), vaccination and medical science. The main social media platforms that participants talked about and engaged with in the interview setting were Instagram, YouTube, TikTok, and Facebook, although others were mentioned. Semi-structured interviews were chosen as a method that could accommodate this participant-led approach to defining the relevant field of research. Without doubt, the interviews generated richly suggestive materials in relation to our interest in visual trust. This article thus describes the method but also interrogates some of the “messiness” that this method entailed.

Attending to Images on Social Media

While much social media uses text and audio, imagery as a mode of communication stands out in a variety of ways. Images have been shown to be more efficient at generating affective responses (Joffe, 2008), attention (Li & Xie, 2020), persuading, and mobilizing (Davidjants & Tiidenberg, 2022; Schreiber, 2023). This has profound methodological implications (Rose, 2022). Incorporating visual material into spoken conversation in interview settings—in photo-elicitation interviews, for example—has been shown to prompt participants to give more complex and layered accounts of their experiences and allow researchers more nuanced interpretation (Cleland & MacLeod, 2021; Marston, 2019; Märtsin, 2018; Niemelä-Nyrhinen & Seppänen, 2023; Rose, 2022; Woodward, 2019).

Social media generate a specific context for image elicitation interviews. Images on social media are embedded in technocultural practices of circulation since they are shared among users, which “complicates units of analysis and concepts established in classic visual studies” including authorship and ownership (Tiidenberg et al., 2025). The meaning of visual media also shifts and changes according to the specific interpretive community. Social media content varies depending on the time of day, one’s location, the platforms involved, and because of who happens to engage with it using which profile. The circulation of visual media and content is also profoundly shaped by platform algorithms.

Today, algorithms act as “a key logic governing the flows of information” on social media (Gillespie, 2014, p. 167). This shift affects how content is prioritized and presented, which is based on user behavior, preferences, and engagement patterns. What matters for platforms “is not necessarily to create a flow of new content but of engaging content” instead (Bucher, 2020, p. 1709). This means that content curation is increasingly guided by “right-time” rather than real-time delivery. However, the “right-time” is also highly context-dependent; what might seem right for the platform may not feel right to the individual social media user (Bucher, 2020). Furthermore, algorithmic curation results in content being increasingly inconsistent (Merten, 2021). Since algorithms prioritize engagement over continuity, feeds often mix unrelated topics in various formats. Not only is it challenging to keep track of varied content but the retrieval of specific posts has also become challenging due to the constantly updating flow of content (Lupinacci, 2024). In the context of people’s everyday social media use, a range of techniques are adopted for saving or fixing content that users want to revisit, like taking screenshots, saving posts, messaging links to themselves, or commenting on posts to receive thread updates (“commenting to follow”).

The unpredictability of algorithmically curated content becomes highly relevant in interview settings. In our interviews, we wanted to engage with the users’ experiences of social media imagery. We therefore adapted some existing methods for researching social media practices, which we describe in the next section by inserting the experience of encountering social media app imagery in real-time into the interview situation: this is what we call a social media elicitation interview, which we explain in the section entitled “Doing Social Media Elicitation Interviews: Looking at Social Media Together.” However, we were unprepared for the extensive and diverse forms of those encounters; we were also unprepared for what we experienced—initially—as their disruptive effects on the semi-structured interview format. Our intention here is to pay attention to this sense of messiness and explore what it tells us about researching contemporary digital visual culture on social media. Messiness in research encounters is often “othered” and made invisible in conventional academic descriptions of methods (see Law, 2003, 2004). Instead of trying to mold our research encounters or data into something familiar and orderly, we attempted to accept the mess and learn from it.

Crucially, we do not see the presence of mess as a failure of the social media elicitation interview method; rather, we propose that our experiences of mess during interviews generated important and rich insight into how people look at social media on screen as well as the sociality of this looking as it emerged through situated interview encounters. This adds a layer of conceptualization to how we make sense of and explain digital visual trust in our project. Throughout this article, we draw on excerpts from interviews ¹ and research notes to explore what happened to the felt, observed and spoken dynamics of the interview when participants’ phones—as a particular kind of digital material object—were brought into the interview encounter (Jovicic, 2020; Woodward, 2019). This gives us an opportunity to carefully consider what precisely we were looking at when we attempted to look at social media together with participants, as well as how this was articulated in talk. Here, we understand the act of looking at social media together as the situated ways in which eyes and attention are directed toward certain elements of the phone, and how the intensity of this visual attention shifts (Lehmuskallio, 2019). This means that we do not take the act of looking for granted, but as one that is mediated by those performing a specific situation (here, the interviewer and interviewee), as well as by the visual media that is used to focus attention (Grasseni & Gieser, 2019).

Social Media Interview Methods

A range of qualitative research methods have been developed that enable close engagement with participants and their digital technologies. Encountering digital environments or content and verbally narrating this encounter in the interview setting is a very particular kind of “staged” digital practice. Some methods more overtly stage this encounter by inviting participants to practice a pre-designed digital interaction. For example, the thinking aloud interview method often involves preparatory work, where the participant completes training tasks to practice how they will simultaneously navigate and verbally narrate content, and it is often followed by a more structured interview to elaborate on the thought aloud narrative (Davies et al., 2024; Freiling, 2019). Given the potentially long-term use of certain social media platforms, scrolling back over archives of personal posts and the digital traces embedded in personal “timelines” can give structure to participants’ narration of key life moments (Dubois & Ford, 2015; Robards & Lincoln, 2017). As these examples demonstrate, bringing apps, images, screenshots or visual representations of data into the interview encounter can offer generative prompts for reflection and conversation (Rose, 2022).

Close to the concerns of our research project is the media go-along interview (Jørgensen, 2016), which offers a collaborative method for the interviewer to engage with participants via a spoken co-narration of their movements through apps (Marston, 2019). There is an explicit acknowledgment in this method of the role of the interviewer in constructing meaning from their practice of “going along” with the “verbal-visual tours” that participants give through a digital media environment (Jørgensen, 2016, p. 34). Other methods have been developed specifically to grapple with the more “ephemeral ways of ‘moving and being with’ social media” (Møller & Robards, 2019, p. 95). Broadly, digital ethnographies offer ways of engaging with unstable social media content that moves, updates, disappears, and reappears in sometimes unexpected ways (Boccia Artieri et al., 2021). An attention to the minutia of these mediated mobilities is possible through wayfaring methods (Hjorth & Pink, 2014) that bring together hands-on interactions with a media artifact, such as a phone, and their connections to experiences and knowledge of the wider media environment. Various iterations of wayfaring and walking methods pay attention to how mobile technologies are used in ways that move along with the moving nature of online/offline actions and environments (Luhtakallio & Meriluoto, 2022; Wilmott, 2020). Also attending to these movements, Moores (2014) emphasizes the “habitual movement of human hands, involving deft movements of the fingers or digits on keyboards and various touch-sensitive devices” (p. 205) and how these “small-scale media-body interactions” generate particular pathways through media environments (Møller & Robards, 2019, p. 96, also Hamper & Nash, 2021).

How social media content moves, disappears and reappears is shaped by recommendation algorithms and different platforms’ algorithms are experienced as varyingly distinct, helpful, intrusive and adaptive (Ruckenstein, 2023). TikTok’s algorithm, for example, is described in popular discourse and user imaginaries (Bucher, 2016) as speedy, reactive and particularly clever in how it responds to user preferences communicated via a variety of haptic gestures like swiping and scrolling (Siles et al., 2024). For the argument at hand, it is important to point out that algorithms can also affect the research process (Braun & Mateus, 2024; Duguay, Gold-Apel, 2023; Zhao, 2024). While interview situations have not been the focus of existing work on the methodological implications of algorithms, we note that if the interview involves scrolling along, walking through or wayfaring social media with participants, then that experience will be inextricably entangled with the algorithmic surfacing of content.

To explore everyday interactions with social media imagery, our social media elicitation method shares similarities with existing methods but also extends these in new directions in response to a changing social media environment. Inspired by the “digital wayfarer” and the distributed, non-media-centric qualities of social media practice, we wanted to focus on participants’ social media environment or content feed rather than one app; yet we also wanted to pay close attention to participants’ interactions with the complexity of their phone screen’s content. We therefore decided to conduct interviews in one-to-one situations where we could not only talk but also see and—hopefully—record those interactions. The next section describes our method in more detail.

Doing Social Media Elicitation Interviews: Looking at Social Media Together

In order to pay attention to ways of looking at, talking about and showing social media imagery during our research, we invited participants to engage with questions of visual trust in different ways. Firstly, they were given a task before arriving to our meeting, which was to find or imagine examples of good and bad social media content in relation to everyday health and well-being. What precisely constituted good or bad in this context was left open to the participants, who interpreted this prompt in a variety of ways and sometimes raised the issue of ambiguity in these terms. Sometimes, unique pieces of content could be good in some ways but bad in others, giving us an insight into how different aspects of the content carried different weight in participants’ evaluations of trustworthiness; for example, an important and impactful health message might outweigh unappealing design. Second, we had prepared a semi-structured interview guide that allowed us to focus our discussion with the interviewees while giving them ample space to bring up in speech different reasons and argumentations for their choices and actions. Third, we asked them to show examples of visual social media on their phone during the research interview, generating a set of situated interactions that we could focus on in our analysis and shifting from broader reflections on social media use to more specific evaluations of unique pieces of content.

We piloted the interview guide and had multiple conversations about the ways visual social media was part of the interview experience. While we did not make significant changes to the planned thematic blocs that structured the interview, we did opt for flexibility in how each interview interacted with visual content. This was a decision informed by early experiences where some interviewees forgot to preselect examples or felt unsure about what to select, which prompted us to integrate both preselected—occasionally present—and in-the-moment viewings of content into interviews and discussions of trust.

We interviewed 10 male and 40 female participants ² who were aged between 18 and 30. In terms of primary occupation, 37 participants were students and 13 were either unemployed or in formal employment. Our interview method was separately reviewed and approved by the appropriate research ethics committees at each of our three institutions. As part of the interview consent process, all participants were asked if they felt comfortable looking at “live” content on their personal social media accounts or feeds and sharing this with the interviewer, and if they consented to this material being photographed or filmed. While most participants were willing to show the interviewer content on their social media accounts, they demonstrated different degrees of willingness or openness in showing personal content. Some expressed hesitance due to concerns about the appearance of “embarrassing” content or personal details, which we describe later in our discussion of access control.

During interviews, the participants used their own phone to access content and only very few times did they hand over their phones or invite the interviewer to swipe through content. Looking at social media on participants’ phones, rather than tablets or laptops, reflected the most common device used by participants for accessing social media and, as we describe in the section entitled “The Mess of ‘Doing’ Social Media in the Interview,” phones have the benefit of being easy for participants to manipulate in terms of selecting what they want to share with the interviewer. However, our ability to see what was on the phone screen was often limited, meaning that what we were actually looking at in interviews was participants’ attention to and tactile interactions with their phones; the taps, swipes, and scrolls of their fingers on the touch-screen; the flicks or twists of their wrists; where they placed their phones during our conversation and their facial reactions to what they were seeing on their screens. Both our look at visual social media content, as well as that of our research participants was heavily mediated via the situated research setting (Favero & Lehmuskallio, 2025; Lehmuskallio, 2019).

All the interviews were audio-recorded. The participants did not look at and show us their phones in a stable manner throughout the interview, rather, there were some periods that mostly consisted of dialogic conversation, and others that centered on the phone screen. In this article, we focus in particular on the latter. Some participants “thought aloud” (Davies et al., 2024) while navigating through their social media by explaining their navigation choices and how they related to what was shown to them while scrolling through their feeds. Other participants scrolled in silence or privately until they found something that they wanted to share with the interviewer. We asked prompting questions while the participant was moving through their social media and we followed up with additional questions about the observed content afterwards.

The recording of participants’ engagements with social media during the interview varied between the interviewers, where we each adopted different methods for photographing and filming participants’ screens: for interviews in Estonia, visual content was recorded by the interviewer on a smartphone; in the United Kingdom, the interviewer used a small digital camera; however, for interviews in Finland, participants were invited to take screenshots themselves during the interview and share these later with the researcher. The Finnish interviews were also filmed in their entirety, with a camera directed toward participants’ hands in a way that gave them control over what they wanted to show. The screenshots, content links, screen recordings and photographs collected from interviews played a key role in our research, which at a later stage included a focused analysis of this visual content for stylistic, genre based, aesthetic and visual-rhetorical “markers” of trustworthiness. These visual traces also enriched the interview transcripts, aiding our memory of the interviews and our interpretation of interview text.

We understand that “any research situation will be dynamic, ongoing, subject-to-change and never as fixed and settled as plans or guidelines might presuppose” (Horton, 2008, p. 376). The next section elaborates the various difficulties, absences, glitches—mess, in short—that pervaded all aspects of the interviews and which as researchers we at first struggled with.

The Mess of “Doing” Social Media in the Interview

In this section, we pay detailed attention to how engaging with participants’ social media and smartphones unsettled interview situations. During the interview, the participants’ onscreen social media emerged as an uncontrollable factor, a constant and often unpredictable eruption into the interview situation. These disruptive moments affected, for example, our ability to control time, follow the interview guide, follow our interviewees’ movements or train of thought—all classic hallmarks of leading a qualitative interview. It is noteworthy that this messiness was not only felt by the interviewers, but participants also indicated how they felt self-conscious about the clarity of their contributions to the research:

I don’t know, somehow, I feel, that my answers are like a messy stream of consciousness. But hopefully you can get something out of them. (Aleksi, 26, male, Finland)

I don’t know if this stuff makes any sense at all, because I’m really bad at telling these things. (Emilia, 24, female, Finland)

Well at the moment I was specifically looking for these foods, so I would probably look for some recipes now. But then I would go to Pinterest [. . .] I don’t know, was that okay? (Angelika, 22, female, Estonia)

Our research participants could not engage in the same stream of social media they had encountered earlier because content does not “hold still” like a printed newspaper page would. Thus, finding content was messy for research participants too, and as researchers we at times needed to use our imagination to understand exactly what we were pointed toward. The interview format then struggles to engage closely and continuously with onscreen visual social media content. What we were able to observe much better, however, was participants’ interactions with their phones and phone screens during the interview (Moores, 2014). The following subsections describe these mess-generating challenges in terms of four interactions between interviewees and their phone screens: fast scrolling, slow scrolling, access control and searching. While we present these four interviewer-participant-screen interactions separately, they are closely related, and participants would shift seamlessly and intuitively between fast and slow scrolling, techniques of access control and searching.

Fast Scrolling

Looking at social media content is fast-paced (Siles et al., 2024). One of the primary challenges of engaging with participants and social media visuals in the interview setting was the speed at which participants scrolled through their social media feeds and clicked between multiple apps and platforms. Participants would move from one platform to another, between different feeds, and scroll backwards and forwards or up and down, in a way that made sense within a particular moment of the interview conversation. As interviewers, though, we could not always follow how participants found their way from one setting to another, nor why they chose to move in one direction instead of another.

The opacity of an interviewee’s “lines” or “paths” (Moores, 2014) was a major cause of what we have described as a messy feel of interviews. In looking at social media together with participants, there often did not seem to be any neat trajectory to be followed but rather a much more fleeting interaction with apps that was shaped by the participants’ ability to work with the functionalities of the apps and browsers, their understanding of the thematic focus of the interview, and their feelings about what they were willing to share with the research project and interviewer. While scrolling, participants were able to recognize something about the content (for instance, a genre, subject, person, or connection to other content), which would prompt short pauses. These were situated and momentary evaluations that, as interviewers, we had very partial insight into. Rather than moving through a social media environment or app together, looking at social media with participants felt like a series of sudden and sometimes inexplicable stops, starts, and leaps across platforms. The participants were enacting a familiar practice of fast scrolling and flipping through apps; staples in their everyday experiences with their phones and social media, which we were able to observe but not necessarily follow.

A sense of disorientation is evident in the following interview excerpts, where the interviewer lost track of what the participant was showing on their phone and attempted to (re)locate the conversation:

Q:

So where are we now? Is this—

A:

So I searched “what I eat to lose weight.”

Q:

Oh okay, yeah, yeah.

(Lily, 18, female, UK)

A:

[Referring to a content creator]—this is terrible, terrible, at least in my opinion.

Q:

So you don’t like him but you—he pops up occasionally on your –

A:

Yeah, yeah.

Q:

What was this, sorry?

A:

That was Snapchat.

Q:

Okay.

(Theo, 21, male, UK)

A:

So you can see there’s some—

Q:

Where are we now? We’re on Instagram?

A:

Instagram and this is my “For You.”

(Amelia, 28, female, UK)

Q:

So, did you go to the previous video’s . . .?

A:

Yes, I opened the video from here and then looked what is here in the recommended section of that video.

(Eetu, 21, male, Finland)

In addition to the fast movement through and across vast masses of content, unique “pieces” of social media content can also be fast, and in interviews we often struggled to keep up with content and missed opportunities to engage with it before it disappeared. This was especially the case when encountering TikTok videos, which are based on a quick sequence of video or image frames (often lasting less than a second each) in addition to text, iconography, spoken audio and music. Unlike looking back over familiar content or archives of personal memories on platforms like Facebook (Robards & Lincoln, 2017), where the speed of movement through content is largely determined by the speed of the scroll, the ephemerality and speed of TikTok content is designed to provoke momentary impressions rather than sustained and extended attention over time (Zhao, 2024), and was particularly challenging for us to engage with in interviews. Fast scrolling and complex content, combined with the small screen and imagery, presented a “concentration of details” (Beugnet, 2014, p. 204). This temporal and spatial concentration made content harder to observe from a distance, which in our case was usually only from the other side of a table but far enough not to be able to see content easily.

Fast scrolling during interviews not only posed difficulties for the interviewer keeping up with the conversation and seeing the images being discussed, but it also posed difficulties for our attempts to record content that the participant showed. The interviewer who prompted participants to take screenshots during the interview at times voiced their request too late to the interviewee, who had already moved on, causing situations where the interviewee had to go backwards or otherwise search the content in question again. For the two interviewers who used phone and digital cameras to capture pieces or sections of content, managing recording equipment and the imperative to capture content on camera created another messy dynamic or “irruptive moment” (Nordstrom, 2015, p. 389) where the conversation, encounter with social media and camera device did not work together smoothly. For example, in the following interview excerpt, the interviewer missed a particular interaction that they wanted to capture, so they asked the participant to repeat this certain action again to facilitate recording:

Q:

Okay. Could I—this is really awkward, but could I get you to do that again whilst I film it?

A:

Okay.

Q:

Just so that I can catch what you’re doing as well. Okay, that’s fine.

[Interviewer takes video recording in silence].

A:

Okay?

Q:

Thank you. Sorry, these are really awkward, like disruptive, I feel like, to ask—

A:

No, it’s fine.

(Olivia, 26, female, UK)

Similarly, in the following exchange, the interviewer again missed a particular engagement with a social media post due to the slowness of preparing the phone camera (picking up the phone, opening the camera app, selecting the right setting, focusing the image and finding the right frame), which was at odds with the fast-paced encounter with social media. When the filming started, the participant “played along” by staging the interaction as if it was happening for the first time:

Q:

Okay, okay, can I film how you search for it?

A:

Okay, mm.

Q:

Sorry, only if you’re comfortable

A:

I’ll delete it [the search keyword] then, like it’s the first time it’s happened

Q:

I’m so slow with this phone of mine

(Silvia, 24, female, Estonia)

In cases where the whole interview encounter was recorded, these re-stagings of unrecorded engagements were not needed. The act of continuous recording eased some of the stressfulness of the fast-paced interviews for the interviewer as it created a future possibility for checking fleeting moments and contents from the recordings. The continuous presence of a recording device does inevitably shape the material space and dynamics of the interview, placing a boundary around the interaction and necessitating a critical enquiry about what is contained within the frame of the recording and what remains beyond (Nordstrom, 2015).

Slow Scrolling

In contrast to the fast scrolling through masses of social media content that we struggled to see and follow, another mode of engaging with social media in interviews felt very slow, and sometimes too slow. In situations of slow scrolling, participants moved through content slowly and they mirrored this slow movement with a considered spoken narration of content. Typically, the participant would tell the interviewer about each post that appeared on their screen as they moved toward something that they thought would be relevant to the research topic, which often meant that they were engaging with a lot of content that felt less relevant to our research focus on trust in the context of health and well-being. In the following quote an interviewee was talking aloud as she scrolled through her Instagram feed for just over 3 min, which in the context of the conversation felt like a long time both to the interviewee and the interviewer. The aim of her scrolling was to find health related content to show the interviewer, but she narrated all the other content that was appearing on her screen. We intentionally present the following two quotes at length to demonstrate the diverse range of social media content that appeared in interviews as well as how participants spoke through their scroll, addressing constantly changing content:

Here comes, well, first we can see Käärijä [Finnish singer] [laughs]. This is someone who saves money because they want to achieve, I guess, I don’t know, something like a stock portfolio of 100,000 euros. Finnair advertisement. Swimsuits, summer is coming. This is some kind of counter-media. Morning newspaper cartoon. Foodora ad. HS Nyt [newspaper supplement] comic strip and Helsingin Sanomat [newspaper]. There is a lot of newspaper content, by the way, suddenly. Again, ad for swimwear. An advertisement about some animals, (–) follows. Here is one of my schoolmates who takes a lot of photographs. Again, an advertisement for clothes. Some content from the Union of Journalists. Then a sponsored ad from social media influencers, because, oh, I do not follow that guy. Then comes a makeup ad. Well, it doesn’t look like there would be any [health related content] coming up. Then comes some graduation pics. Finnish artists. Sponsored Seta [LGBTIQ Rights in Finland] content, that one is like suggested. Clothes again. I haven’t clearly been watching health stuff for a while. What’s this? This is funny, Instagram has changed so much, now that I am scrolling, I get so many ads and those, what are they, suggested content. But . . . National Geographic. This is not, does this have something to do with health? No. This, too, is suggested. Swimwear. News in Spanish. Travel photos. This one too is suggested, political content, more summer clothes [laughs]. My former teacher’s theatre stuff. New York Times. Is this health-related somehow, some break-up stuff, I don’t know. This one in a way is because here is some kind of emotional, or that this is a therapy thing. It took quite a long time unless the comic in HS Nyt, I didn’t read or open it, it could have had something to do with health. But actually, I could go back to check it. (Jenna, 26, female, Finland)

Often, social media content is consumed individually, quickly and intuitively, which is at odds with the kind of detailed narration above. In other cases, moving through diverse social media content took the form of a dialogue between the participant and interviewer, as demonstrated in the next interview excerpt. The following exchange is taken from a 1-hr-long interview where the participant narrated a range of content but generally struggled to focus on content that was health related:

A:

That’s a bakery.

Q:

Animals, I always see animals when I do this with people [both laugh].

A:

That’s what I’d share with someone. Oh, yeah, that’s hilarious, yeah.

[Pause in silence as participant looks through phone].

Q:

Okay, so this is the running?

A:

This is, yeah, the running thing I was telling you about but it’s just a parody, yeah.

Q:

That is what you were saying.

A:

Yeah.

[Both watch short video in silence].

Q:

I see, okay, that’s quite good.

A:

I’ve been looking at these trainers [shoes], which is why they’ve come up. So yeah, I get so much of this stuff now, which I never used to get and I don’t follow them. I never used to get that kind of thing.

Q:

Targeted to you for some reason.

A:

Yeah, Sean Logan on love.

Q:

Quite a mix of . . .

A:

Suggested things [laughs].

Q:

Quite a lot of fun.

A:

That’s what’s suggested to me, so that’s a bit worrying [both laugh]. Why is that suggested to me?

Q:

It’s quite funny [both laugh].

A:

Yeah, I get this comic a lot. Jess Phillips.

Q:

What’s this one about?

A:

So this is Jess Phillips, the MP but it is about . . . meds.

Q:

Medicine. Okay, general health and illness. Age related.

A:

I’ve never looked at that before in my life, so I don’t know why I was given that. Another bakery.

Q:

Another fun one.

A:

Yeah [laughs]. I do tend to follow my local like cafes and bakeries, to see their opening times.

Q:

I think this is—there’s a lot of mix of all sorts of things on here, isn’t there really?

A:

Yeah.

Q:

Oh, so we’ve got some news now.

A:

BBC news.

Q:

So would this be quite typical of what you would see?

A:

Yeah, yeah.

Q:

So there is health stuff dotted—oh, what was that one, is that a tracker?

A:

This?

Q:

Yeah.

A:

Oh these are headphones.

Q:

Headphones.

A:

Yeah, I just bought some actually. Oh look at this, I do love the latte artworks.

(Charlie, 27, female, UK)

As interviewers, we aimed to be open to any kind of content that appeared during the interview, and for the conversation to be led by the participant. Yet at the same time, we occasionally felt challenged by this slow narration of curated content that only touched the surface of social media, when we were trying to move toward more in-depth reflections on distinct visualizations of health and well-being, such as a post, image, or video. We felt some discomfort at interrupting participants’ silent scrolling through content during the interview or their narration of what they were seeing, but sometimes we felt a need to intervene in order to prompt reflection or break longer periods of scrolling in silence. One strategy for doing this was by finding points of connection and commenting on something we found interesting that appeared on screen; however, especially when engaging with video content, these prompts often did not work. All of the interviewers had experiences of participants not responding to questions or shushing the interviewer to focus—often in quiet—on content or their scrolling. This suggests that scrolling feeds or getting stuck watching a video is both familiar and immersive, so the norms regulating these activities upstaged the less familiar norms of an interview conversation.

Access Control

In general, the unpredictability of content offered a generative prompt for getting participants to reflect without prior preparation on what they were seeing and articulate tacit perceptions, principles and criteria that they relied on in their assessments of the quality, credibility, authenticity and in general trustworthiness of social media content. As we have described, some participants comfortably narrated the suggested content that appeared on social media, which was wide-ranging and not of a sensitive nature. For others, though, showing suggested content offered an insight into their media consumption that some participants felt was too personal, embarrassing or inappropriate for sharing during the interview. To manage the risk of disclosing such content unintentionally, many participants adopted a strategy that we call access control. This primarily involved participants holding their phones at an angle where only they could see the screen while they selected pieces of content, such as a unique post or account, to show the interviewer. This offered a strategy for the participant to retain some agency over what they wanted to disclose from their personal social media accounts and necessitated flexibility and openness in how we invited participants to show us their screens. The small, hand-held and touch-activated screens of smartphones invite individualized use and potentially a more intimate user experience of apps that can be difficult to share with others (Hamper & Nash, 2021).

While trying to engage with live social media content during the interview we were mindful of how tailored content on social media can infer sensitive and personal information or preferences, and how participants could not fully anticipate what might appear on the screen during the interview. So too were our participants. One noted how “it’s such a Pandora’s box going through your social media” (Olivia, 26, female, UK). Similarly, the Explorer page on Instagram was described as “always such a wild card” (Jieun, 25, female, UK). The following participant also remarked the unpredictability of content on TikTok:

A:

Let’s see what comes up [laughs]. Hah . . .

[Participant starts scrolling through TikTok].

Q:

You can’t always quite control it.

A:

That’s so true, you really can’t.

(Laura, 27, female, Finland)

The metaphors of the wild card and Pandora’s box demonstrate an important facet of how participants made sense of algorithms and the potential to be surprised by new social media content, as well as how showing live social media is unpredictable. This kind of “material metaphor” (Woodward, 2019) is central to the talk that circulates about social media (Tiidenberg, 2020).

Differently from studies that concern user-generated content or personal timelines of events on social media (such as Møller & Robards, 2019; Niemelä-Nyrhinen & Seppänen, 2023), the primary risk of unplanned disclosure in our interviews pertained to algorithmically tailored content on the Explore page of Instagram, the For You page of TikTok, the newsfeed of Facebook and the Shorts section of YouTube. These pages offer recommended content based on the account owner’s recorded interests or previous browsing habits. The unpredictable nature of engaging with tailored content in interviews created situations where we had to be sensitive and responsive to the appearance of uncomfortable content. Indeed, this content is often not subtle but designed to interest—possibly shock and surprise—but at least grab the viewer’s attention (Bucher, 2020).

Our experience was that participants were also sensitive to the risk of unintentional disclosure. Many commented on the personal nature of algorithmically suggested content and some suggested that their targeted content feeds revealed something intimate about their personality:

Q:

What was it that made you think that the algorithm knows what you . . .?

A:

Well, this feels very intimate to tell.

Q:

Tell only as much as you want.

(Anna, 26, female, Finland)

Another participant noted how the For You page on Instagram gave an insight into the kind of content she consumed, which she described as being at odds with how she presented herself, especially in a professional capacity:

Q:

Okay, I mean, would you be happy to have a look at some content on your phone or would you prefer not to? It’s really up to you.

A:

I’m happy to.

Q:

You’re happy to, okay.

A:

I just like, you know, I don’t have anything to hide, necessarily.

Q:

No.

A:

It’s not something I’m proud of, because, you know, it’s one of those things I feel quite off brand of me, I’ve engaged with so much of this content. But I have done it for many, many years, which is kind of sad when you think about it a little bit.

(Jieun, 25, female, UK)

The opaque workings of algorithmic content curation thus had significant impacts on the interpersonal dynamics of the interviews. Moreover, the unpredictable element that shaped interviews was not exclusive to social media but extended to participants’ phones more generally. Occasionally, participants received messages or phone calls that were displayed on their phone screens in the middle of the interview:

A:

[A message appears on the phone screen]. Oops, oops. Is it a problem that I just got a message?

Q:

No, absolutely not, when someone sends you something—

A:

Yeah, that can be left out afterwards.

(Piret, 22, female, Estonia)

There were many other examples of tactics for not disclosing potentially intimate content. When finding specific content to show the interviewer on YouTube, one participant noted how they were using a browser on their phone to present a more generic platform and avoid their personal account in the YouTube app. Others chose to show content after the “formal” interview had ended and recording devices had been notably and visibly switched off. A male participant chose not to show any social media content in the interview, explaining that his Instagram feed contained a large amount of soft pornographic content: “Another really important facet of the meaninglessness of the posts that came up is somehow, or for some reason, um, soft core porn invaded these accounts, all of them, all of them” (Sam, 24, male, UK). Another participant, while showing the interviewer a video on Instagram, quickly intervened by swiping his phone screen as a fully nude male content creator appeared in the video. Others joked about the awkwardness of showing the interviewer photos of women in “skimpy bikinis.” We suspect that the fast scrolling that we described in the previous section could also at times have been a strategy for dealing with potentially embarrassing content by limiting the interviewer’s ability to see what was appearing on screen and enabling the participant to select content that they perceived as “safe” to show the interviewer. Both fast and private scrolling could offer strategies for the participant to protect themselves from embarrassment as well as protect the interviewer from seeing explicit content that we were not prepared for.

All this presented ethical challenges pertaining to participants’ practice of informed consent (Møller & Robards, 2019). While our consent agreement covered sharing a screen, sharing screenshots and having their screen recorded, and highlighted that consent to any of these could be withdrawn at any moment, evaluating when and how to look at social media together involved on-the-spot ethical dilemmas and decisions about the boundaries of disclosure. In practice, we dealt with this by reminding participants that looking at their social media was an optional part of the interview as well as facilitating the kind of interaction that participants seemed to prefer, whether this was looking at social media together or facilitating the participants’ private scrolling. Furthermore, looking at a small screen together necessitates a degree of closeness between the interviewer and participant, and this physical boundary also had to be negotiated carefully to avoid feeling intrusive. For example, we would avoid leaning over to the screen when we had not been invited to look or touching the participant’s phone when we had not been invited to touch.

Searching

Whereas the premise of our social media elicitation method was to ask participants if they could show us content on their social media channels, the previous discussion demonstrated that algorithms significantly determined the direction of our conversations. Here, we further reflect on the impact of the algorithmic feed on our planned interview process, focusing on how the content that appeared during the interview was unlikely to be content that the participant had encountered before the interview. While all participants had encountered health-related content on social media in the past, which was one of the prerequisites for taking part in the interview, and some participants had prepared exemplary posts they wanted to discuss, there was a general agreement that this content is “just there” but not something they actively sought on social media. This meant that participants often had to—rather unusually—search for content that aligned with the interview’s thematic focus on health and well-being, which emphasizes how interviews were nudging them toward encounters with social media that were outside of their ordinary use contexts (Woodward, 2019). The following quotes show two participants’ attempts to find health related topics:

Of course, I might stop to look at some of these [posts on screen] if I were just by myself. But now I’ll try to find something health-related [laughs]. (Sofia, 21, female, Finland)

Can you give me an example of what problem I have, what help I’m looking for? I don’t know how [. . .] Okay, wait, wait-wait-wait. Okay, some kind of health issue. I have to come up with some kind of a health issue. (Angelika, 22, female, Estonia)

Other times, participants would spend a lot of time looking for particular content to show the interviewer but not being able to find it. It was common that participants would have a post in mind but could not remember the name of the account or content creator:

A:

It was something with the letter K, I remember that [. . .]

Q:

But the recommendation came from YouTube?

A:

Yeah on YouTube and then later it was like on that. Ah, now it has like disappeared from the face of the earth.

(Piret, 22, female, Estonia)

Participants would sometimes be able to find the content through searching with keywords, or going through their direct messages, favorites or saved posts, but often they were not able to find it in the moment of the interview and would sometimes become frustrated or apologetic. Many noted how they were never able to find particular content when they wanted it: “Where are all those posts when you need them? [Pause for 9 s while participant scrolls]. Nothing like that comes across here now” (Emilia, 24, female, Finland). Similarly, during the following exchange, the participant was clicking through posts to find the content that she wanted to talk about. The interviewer could not see the content that the participant was scrolling through:

Q:

Yeah, yeah, okay, that’s really helpful to have an example. Are you able to pull up any of these content [sic]?

A:

Yeah, sure.

Q:

Posts, sorry, on your phone.

A:

Yeah, let me have a look. I’m trying to think where they will –

Q:

If you don’t—if you can’t find them, or you don’t remember, then that’s absolutely fine.

A:

No.

Q:

It’s not a—not an issue –

A:

Right, let me –

Q:

—if it doesn’t work out.

A:

Is it this? No, it’s not that one. Right, just give me one second, sorry.

Q:

No worries.

A:

I’ll try and find—

Q:

It’s actually quite difficult to find things on the spot.

A:

Yeah. No, exactly.

Q:

It’s the kind of thing that you see all the time and it’s on—you know it’s on there but actually—

A:

Actually, yeah.

Q:

Finding it when you want it.

A:

Finding it’s very different—difficult. Right, let me try and see if I can find—sorry about this.

Q:

Don’t worry, it’s not—I didn’t warn you that I was going to be asking.

A:

Right, some of them should be down at the bottom. So this is one person I quite like.

(Nova, 20, female, UK)

A tension exists therefore between the platforms’ aim to present users with engaging, ever-changing content (Bucher, 2020) and the desire of researchers and research participants to recall specific content according to our research topics. Participants’ searching through content was influenced and disrupted by algorithmic curation, which shifted the direction of the conversation not only by introducing unrelated subjects but also by making it difficult to retrieve older posts (Lupinacci, 2024). Hence more unpredictable mess in our interviews.

Discussion: Interrupted Talk and Distracted Visuality

Researchers interested in users’ experience of social media have frequently used interviews to talk about what users do, why they do certain things and how this makes them feel. Researchers have also adapted the standard semi-structured interview in various ways to engage with the complexities of digital practices. They have responded to arguments about the agency of the apps and their platforms by carefully itemizing apps’ technical affordances as well as how their users deploy those affordances as in, for example, the walkthrough method (Light et al., 2018) and media go-along interview (Jørgensen, 2016; Møller & Robards, 2019). They have used personal archives or digital traces of activity on social media to prompt talk about life histories (Dubois & Ford, 2015; Robards & Lincoln, 2017). They have taken the interview out on the road to accompany digital wayfarers as they chat, scroll and navigate in urban environments (Hjorth & Pink, 2014; Luhtakallio & Meriluoto, 2022; Wilmott, 2020).

All of these methods have generated rich and nuanced talk about social media use, evidenced perhaps in the current popularity of the notion of digital imaginaries. Bucher’s (2016) use of the term in relation to algorithms has been particularly influential, defining the algorithmic imaginary as “the way in which people imagine, perceive and experience algorithms and what these imaginations make possible” (p. 31). This allows us to understand that “[w]hat the algorithm does is not necessarily ‘in’ the algorithm as such [. . .] Rather, we may begin to understand the performance of algorithms through the ways in which they are being articulated, experienced and contested in the public domain” (Bucher, 2016, p. 40; see also Ruckenstein, 2023). In such work, the emphasis is on the human meaning and significance through which digital technologies are made sense of and deployed, and interviews are the most frequently used method to explore such user imaginaries.

Thus, we too were inspired to adapt the semi-structured interview method to explore social media imaginaries, and specifically why some social media images were trusted by young adults in relation to their everyday health and well-being. We did this by asking our research participants to share with us certain kinds of posts, images or their phone screens as we talked. The result was 50 rich and nuanced conversations about social media, imagery, trust, health and well-being—conversations that were also littered with what we term “mess.” In his discussion of mess in social science research, Law (2004) describes how the mess of the world is “vague, diffuse or unspecific, slippery, emotional, ephemeral, elusive or indistinct, changes like a kaleidoscope, or doesn’t really have much of a pattern at all,” and argues that conventional social science methods—among which is the semi-structured interview—are “poor at catching these” (p. 2). Law’s provocation is in some ways banal: every interviewer knows that interviews can contain unpredictable, emotional or baffling moments, that thematic coding can erase these in its search for meaningful codes, and also that the mess of the research process rarely appears in published research findings. We found that images on social media apps appear to have a specific generative force when placed in the situation of a research interview. Or, better perhaps, the conventions of the research interview throw that force into sharp relief.

In her recent discussion of digital photography, Zylinska (2023) discusses “distraction” as a mode of perception. Rather than dismissing distraction as a trivializing disposition, she says, “it would be interesting to probe instead what the audiences are being distracted toward, what new orientations and bindings their attention now engages” (p. 122). We now propose that the “mess” in our interviews is best understood as enactments of at least some of the orientations and bindings by which social media platforms configure distraction. As Paasonen (2021) has argued, distraction is as much a constitutive orientation of social media as attention. Distraction concurrently holds excitement, boredom, interest, frustration, and pleasure. These distracted ways of seeing are not impossible to talk about, as the quotations in this article demonstrate. But the talk is in a particular relation to the apps: the visual content of the app often eludes the talk: too quick, unrepeatable, unshareable with a stranger.

Conclusion

We propose that the forms of mess identified in the fifth section of this article merit close attention. Our engagement with participants’ social media content in interviews was messy because it made its appearance too quickly to be recorded, was embedded in all sorts of material irrelevant to our project, risked being too private, and/or was unfindable. We explored these traits in terms of fast scrolling, slow scrolling, access control, and searching. These sociotechnical configurations inflect and even interrupt the “normativities” of the interview format (Law, 2004, p. 4). In other words, mess-generating interactions between interviewees and their phone screens challenged our expectations of what should happen in a research interview. Yet we gradually came to see how the difficult moments in interviews were generative and gave us valuable insights into the workings of contemporary social media, especially how platforms dispose their users toward distraction. Furthermore, it prompted us to analyze ways of looking and ways of engaging with social media content as part of our participants’ trust repertoires and thus as something that we need to account for in our explanatory framework of visual digital trust. Researchers planning to use social media elicitation interviews should be attentive to the generative potential of mess, but—where possible—prepare for and be mindful of the ethical issues that such mess may raise.