Patterns of Surprise and Ambivalence: Studying Social Media Visuality by Way of Aggregated Autoethnography

Abstract: Visuality is central in social media experiences, but complex to research. In this paper, we introduce aggregated autoethnography for nuanced analysis of socially mediated visual practices. The approach starts from guided autoethnographies which help to empower participants to explore their own experiences and build thick descriptions, and moves through multiple levels of aggregation, integration and synthesis (from individual autoethnographies to national datasets of coded snippets, to datasets specific to arguments emerging out of multinational patterns). The aggregated autoethnography approach makes unexpected topics accessible; offers dynamic, rather than static insight; makes visible that which is routine and tacit, as well as that which is experienced as ambivalent. Further, aggregation allows synthesis of multiple perspectives, revealing patterns across contexts that are otherwise difficult to detect. The approach detailed here is used to move back and forth between the singular pieces of visual content and the flows they are part of; to remain loyal to the situational perspective that the visual communication becomes meaningful in; to capture relevant artifacts as well as people's practices; and to be mindful of the affective, embodied and material aspects of ways of seeing with social media.

4.1 Synthesis and integration: Reformulating first-level datasets into second-level datasets

4.2 Iterative re-integration of (parts of) second-level data sets for multinational arguments, insights and follow-up questions

The importance of visual content in online cultures, commerce and communities cannot be understated (HAND, 2013; MIRZOEFF, 2015). Visuality is central in social media experiences and to people's enactment of sociality (BRECKNER & MAYER, 2023; MÜLLER, 2011; SCHREIBER, 2020; TIIDENBERG, 2025). As a mode of communication, visuality has been found superior to alternatives in terms of generating attention (LIE & XIE, 2020) and affective responses (JOFFE, 2008) and hence, it is of key importance in identity formation, ideology work, emergence and dissolution of solidarities and otherwise boundary making of all kinds (DAVIDJANTS & TIIDENBERG, 2022; SCHREIBER, 2023; ZHAO & ZAPPAVIGNA, 2018). [1]

In this article we use two terms—"social media visuality" and "visual social media," relying more on the first when speaking to the reader of this text, and more on the second when quoting how we spoke to our student-participants. This choice was pragmatic; visual social media is narrower and thus easier to grasp in an already complex research and learning context. We define social media visuality as a way of seeing afforded by the confluence of platform features, functionalities, governance, vernaculars and business interests. Social media visuality is "what we think is possible to see, what we are allowed to see, made to see, what is worth seeing and what is unseen" on and with social media (TIIDENBERG, 2018, p.14). Visual social media, on the other hand, are a conceptual delineation of visual communication that happens on social media platforms. This is a narrow focus on visual content, interactions, norms and visual affordances of platforms. The two are conceptually different, yet overlapping. However, empirical research questions pertaining to both often lead to exploration of the relationships between technologies, rules, norms, practices and feelings. Thus, we propose that the methodological approach we describe here is well suited for studying both visual social media and social media visuality. [2]

That being said, visuality of social media is—for the purposes of research operationalization—profoundly messy. This messiness complicates units of analysis and concepts established in classic visual studies. Authorship and ownership are difficult to pin down: Original content is often remixed with appropriations of third-party material, reshared, endorsed or reframed (BRUNS, 2008), which does not only impact the location of the field or object of one's analysis, but how one might conceptualize methods (MARKHAM, 2013a). As visual media move through time, space and contexts, their meaning can obviously change as interpretation is always situational (MIRZOEFF, 2015). What visuals refer to might not always be entirely clear either: Representation of someone or something can be iconic and indexical, but only be accessible as such for specific interpretive communities (SCHREIBER, 2017; TIIDENBERG, 2025). There are no obvious visual, textual, technical or metadata markers that clearly categorize what is what on visual social media, a point of pivotal importance in the context of research methods. It is deeply situational. Moreover, to overcome a simplistic, causal understanding of media effects, one needs to focus on how people relate to images in different ways. This is why our own and colleagues' earlier work has argued that one needs to take seriously what people are doing, saying and feeling in, on and with visual content—e.g., their practices (RECKWITZ, 2003; SCHATZKI, 1996), focusing specifically on social media practices (TIIDENBERG, HENDRY & ABIDIN, 2021) including visual social media practices (GOMEZ CRUZ & LEHMUSKALLIO, 2016; LEAVER, HIGHFIELD & ABIDIN, 2020; MIGUEL, 2016; SCHREIBER, 2017, 2020; TIIDENBERG, 2015, 2018). Not every research design is well suited for that. [3]

We propose a large-scale ethnographic approach we call aggregated autoethnography for making sense of socially mediated visual vernaculars, practices, norms and the messy flows they are enacted and experienced in. The approach starts with empowering participants to become autoethnographers of their own experiences, guiding them through the process of building thick descriptions (GEERTZ, 1973) and then aggregating individuals' materials to generate rich datasets. The guided autoethnography model—designed in 2012 by Annette MARKHAM and developed with more than 1,500 young adults between 2012-2020—has shown repeatedly to result in

"the production of extraordinarily rich ethnographic accounts from youth, reflecting on their own digital media production and consumption, their relationships with others as mediated through digital media, their interpersonal relationships with various non-human entities, including devices, interfaces, platforms, algorithms, and companies" (MARKHAM, 2022, §6). [4]

Aggregating such nuanced and robust materials as a form of data collection and analysis (TIIDENBERG et al., 2017) enables studying social media cultures and practices, including social media visuality in ways that are sensitive to broader patterns, yet granular and nuanced. Unlike traditional autoethnography, aggregated autoethnography allows for the synthesis of multiple perspectives, revealing patterns across contexts that are otherwise difficult to detect. Drawing on the experience of using aggregated autoethnography within the research project "Trust and Visuality in Everyday Digital Practices" (TRAVIS),¹⁾ we introduce the core principles of guided and aggregated autoethnography and describe the benefits and complications of this type of methodology. [5]

The TRAVIS project is focused on visual digital trust and how it is experienced, practiced and made sense of in the context of health and wellbeing related communication on social media. Specifically, we analyze the trust experiences, practices and perceptions of young adult social media users and content-creators; we also explore what imbues some visual digital images with trustworthiness and how trust is shaped by networked digital communication technologies. [6]

We begin with a brief introduction of autoethnography and guided autoethnography (Section 2). After that, we describe how we approached guided autoethnographies in the TRAVIS project (Section 3). Next, we offer an account of how we aggregated the materials created as part of the guided autoethnographies (Section 4) and follow that up with an analysis on what kinds of data this approach generates and what its methodological limitations and affordances are (Section 5). We conclude (Section 6) with a discussion of aggregated autoethnography specifically as an approach for researching social media visuality. [7]

The so-called crisis of representation of the late 20th century disrupted positivist forms of knowledge production and research practices. It transformed qualitative methodologies as well as reflections on what inquiry practices mean, politically as well as epistemologically. Innovations in ethnography, such as autoethnography, grew out of this epistemic turn. It is a form of inquiry that, for ADAMS and HOLMAN JONES (2008), BOCHNER and ELLIS (1992), DENZIN (1989), GOODALL (2001), HOLMAN JONES (2016), SPRY (2001) and other foundational practitioners, combines ethnography and autobiography. Autoethnography is a mindset as much as a set of techniques that encourage researchers to use their own experiences and ways of knowing to study cultural phenomena. [8]

ELLIS, ADAMS and BOCHNER (2011) reminded readers that autoethnography embraces contingency, that the "truth" of a situation varies depending on who is giving the account, what has been drawn upon to build an account, for what purposes or audiences, and any number of other factors. Under this umbrella of contingency, the value of autoethnography is that it derives from the relevance as well as richness of one's own experiences. Autoethnography is one of many interpretive approaches in anthropology and sociology that highlights the importance of reflexivity (MANN, 2016; MARKHAM, 2017; MRUCK, ROTH & BREUER, 2002; ROTH, BREUER & MRUCK, 2003) as a practice for critically examining one's ways of seeing and making sense of the world. Autoethnographic exploration goes beyond simply writing autobiographically or including one's own viewpoint in the study; it is a matter of layered reflexivity sometimes described via the notion of positioning (GRESCHKE, 2023). "Autoethnographic texts are both (a) accounts of positionings in the field as they were experienced by the ethnographer and (b) retrospective self-narratives [...] and thus positionings themselves" (BOLL, 2024, p.678). Conducting fieldwork at both the direct and meta levels of observation of one's life, researchers can bring attention to the contingent factors shaping knowledge production in overt acknowledgment that the tools they use to make sense of the world will shape what is seen or understood. [9]

Whether or not they were labeled autoethnography, many accounts of the early internet drew heavily on the analysis of personal experiences (for early studies of cyberspace see MARKHAM,1998; SONDHEIM, 1996; STONE, 1995; for more recent examples of autoethnographic exploration of social media see also ARE, 2021; ESCHLER, 2016; MAVROUDIS, 2018). Autoethnography is thus well suited to recognizing and analyzing processes and relations in digitally-mediated experiences that might otherwise remain tacit or buried in infrastructural elements of digital platform interactions (MARKHAM, 2022). This operationalization of autoethnography is concerned with using experiences, glitches, blind spots, and what DENZIN (1989, p.70) would call "moments of epiphany" while using digital technologies to recognize relevant dynamics and larger issues. [10]

Autoethnographic approaches are often distinguished into analytical and evocative modalities. Risking oversimplification for the sake of clarity, we can say that in analytic autoethnography (ANDERSON, 2006), one's personal experiences become ways of observing and analyzing larger cultural phenomena, whereas within the evocative approach, researchers "view themselves as the phenomenon and write evocative narratives specifically focused on their academic, research, and personal lives" (ELLIS et al., 2011, §24). [11]

An analytic approach to autoethnography is an excellent starting position for students to draw on well-established tools for focusing closely on their personal experiences as complete members (ANDERSON, 2006) of their preferred platforms or communities of social media. While the autoethnographic stance enables students to get beyond some of the limitations of having researchers observe their activities, the guidance of instructors well-informed about the legacies and complexities of social media is very important. It nudges students towards self-reflexivity and illuminates how their being in a digital lifeworld is a temporary assemblage of digital, material, and informational elements emerging repeatedly through interactions in socio-technical environments. It is the complexity of entanglements that becomes valuable as the outcome of the autoethnographic process. [12]

The project discussed in this paper particularly draws on MARKHAM's guided autoethnography method (MARKHAM, 2022), which provides a much-needed response to popular and news media oversimplifications of how people use and rely on digital technologies, as well as research participants' tendency to repeat those same media narratives to describe their experiences (TIIDENBERG et al., 2017). The problem of participants giving surface level treatment to complex issues and relations is for MARKHAM (2013b) a methodological problem, where researchers often simply transfer typical engagement techniques from face-to-face to digital contexts, rather than adapting them to be more situationally appropriate. Take the method of interviews, for example. This remains, alongside participant observation, the most common form of qualitative engagement with participants. Because interviews focus attention on the user's recollection, the method presents inherent limitations for understanding processes or the granularity of interactions central to digital life. MARWICK and BOYD (2014) are among those who have creatively adapted interview methods for research on and with digital technologies by looking "over the shoulder" of users and having them talk through their choices. Later, and based on a similar logic of methodological innovation, DUBOIS and FORD (2015) elaborated on trace interviewing, JORGENSEN (2016) on the media go-along method, ROBARDS and LINCOLN (2017) on scroll-back interviewing, and LIGHT, BURGESS and DUGUAY (2018) on the walkthrough method. Such revisions to how traditional qualitative methods are enacted help "create better resonance with contemporary social media contexts, which would result in stronger research products" (MARKHAM, 2013b, p.444). [13]

Practically speaking, guided autoethnography involves an extended, step-by-step process building a social media user's ability to engage in iterative cycles of self-observation, self-reflexivity, and self-analysis. The practice focuses initially on observing and documenting mundane micro-behaviors in everyday situations and afterwards, analyzing these from various defamiliarized perspectives to explore how the taken-for-granted is accomplished. [14]

The TRAVIS study adapts the guided autoethnography method to work with student-participants to build nuanced and thick data that is then aggregated to find patterns across countries. We piloted the approach for studying digital visual trust in Estonia in 2022 and extended it in 2023 to other project countries. Following MARKHAM (2022), we asked participants to adopt an autoethnographic lens to defamiliarize their lived experience of platforms and devices, focused specifically on visuality and trust. For six to nine weeks, researchers guided students through sequential fieldwork and analysis activities. Students were taught to take field notes, write short qualitative code memos, produce visual maps, mood boards, or other multi-modal interpretations of their experiences. The goal was to capture and generate multiple analyzable moments of experience at both the level of the single object, and also at progressively broader levels of sociality, i.e., situations, contexts, and cultural trends. [15]

Researchers introduced these tasks by explaining to the participants that the goal was to explore how and why we trust some content and interactions on visual social media while we distrust others, and how we ourselves enact trustworthiness when we post or share. As trust is an abstract topic, we invited the participants to explore it in the context of everyday health and wellbeing, however they defined that. We reminded the participants that all fieldwork starts from deciding on the boundaries of the field which is guided by their individual research question. Dependent on the participants' study level and previous experience with research methods (we had both MA and BA students) we offered various pedagogical guidance in preparation. Students were invited to come up with their own research questions, but the following questions (Box 1) were offered as potentially helpful to get them started.

To guide the participants towards an auto-ethnographic sensibility that centers experiences of trust as well as the everyday, routine, situated practices of visual social media use, we encouraged them to plan their autoethnographic fieldwork. We reminded the participants to focus on their own experiences and avoid pre-emptive analyses of what they think "typical" social media use is. This emphasis on the subjective came from previous experiences with guided autoethnographies in classrooms where we had noticed a certain tendency of reaching for hasty and generic conclusions. It was important to remind students not to succumb to quoting grand cultural narratives but instead, to really lean into the minutiae of their experience. At this stage, we taught participants field noting (drawing on techniques described by EMERSON, FRETZ & SHAW, 2011; SPRADLEY, 1980; WOLFINGER, 2002) and asked them to prepare and pilot an observation plan (Box 2):

Giving the participants an opportunity to test the plan and get feedback on their initial foray into the field served multiple purposes, created a sense of ownership within the guided process and helped highlight the open-ended nature of (auto)ethnographic work. On a more practical level, it allowed participants to find a field site that felt genuine, which generated better material for them as well as the research team. Testing the plan also set up the necessarily repetitive process of reminding participants not to relax into self-tracking and self-quantification. Rather, participants were trained in the classic ethnographic mindset of noticing and respecting the micro-moments in single situations, diving sequentially and iteratively into different elements that influenced these situations, making the ordinary seem remarkable, and taking seriously the seemingly mundane social media experiences and the flows of their everyday practices. We will show shortly how participants often cycled around to the same moments or objects to explore them at different (micro and macro) levels. Such actions (e.g., noticing something once) and reflections (that one tends to notice it often, but react differently based on emotional response triggered) were written in participants' field notes, which we included in the aggregated data (detailed in Section 4). [18]

Following the planning and piloting of the observation plan, the participants set out on their fieldwork. In the case of TRAVIS research in 2023 students did so for a month, focusing on their experiences of trust, distrust and trustworthiness in the context of health and wellbeing-related visual social media. We directed students to take detailed field notes and document their experience in various ways, including via taking screen recordings and screenshots. Further, we taught them a free-writing "introspective-elicitation" technique that MARKHAM (2017) devised as a type of brain dump to combine aspects of GOODALL's (2001) interpretivist writing exercises with constructivist grounded theory techniques of CHARMAZ' (2014 [2006]) and CLARKE's methods of situational analysis (2005). We required the participants to write three brain dump style memos following the guidance in Box 3.

Finally, we requested that the participants reflect on the process of fieldwork and write memos on that as well. As this was a classroom experiment which served a pedagogical function beyond creating a set of materials for the researchers, the final part of the process entailed teaching the students basics of inductive coding and asking them to take first steps towards making sense of the insights created during fieldwork. Students did initial, open coding of their data, created a mood board of their experience of trust based on it, and wrote short reflections on the process of fieldwork, coding and mood boarding. [20]

Being part of such a project has several advantages for the participants: They can contribute to and gain insights into actual research, and their work becomes relevant within the scientific field. They can list their participation in their CVs or receive—as we have done in some cases within TRAVIS—a certificate to include in their professional portfolios. [21]

Further, guided autoethnographies are an excellent training ground for inexperienced qualitative researchers. As the approach invites a focus on personal experiences, the shift towards introspection leads students to unlearn former experiences or research practices. Through their fieldwork, students explicate their practices and defamiliarize tacit, unconscious routines. This strategy prevents participants from quickly drawing common sense conclusions and enforces an analysis grounded in data. [22]

The students also develop a more nuanced understanding of their digital experiences, increasing their critical digital literacy (MARKHAM, 2018), transforming everyday digital practices from seamless flows to specific affected and affecting moments that have consequences. Our participants critically reflected on both the constraining and enabling factors of digital media use. As one participant put it, "social media is not fun anymore, but I learned a lot about myself and what I am doing." Developing such awareness is crucial for research, but it also has self-developmental potential—increased awareness results in new ways of being with digital technologies (see Box 4, Caveat 1).

Aggregation is a common concept in statistics and data science, indicating the procedure whereby data from different sources or using different measures are combined. Analysis of these aggregates might be based on summaries derived from the disparate datasets, or merely indicate generating what is sometimes called a "high-level dataset" from a combination of individual data-sets. Alternatively, the data points themselves might be transformed into comparable units, a simplification that involves using a broader categorization scheme as the base for comparison. These types of aggregation are used in traditional statistics, machine learning, and predictive analytics. [24]

In qualitative inquiry and within the tradition of ethnography, aggregation is a far less common term. Scholars prefer to speak of emergence, which is apt but can inadvertently sideline certain combinatorial elements (e.g., comparing, synthesizing, aggregating, and integrating) involved in the analytic processes of interpretation. However, there is a growing body of work developing qualitative techniques that rely on or include aggregation, predominantly in nursing research (BERGDAHL, 2019; REIS, HERMONI, VAN-RAALTE, DAHAN & BORKAN, 2007 for critical overviews). Qualitative aggregation often involves tactics like interpretation, integration, iteration, combination, comparison and categorization, to identify patterns, offer insights or practical actionable evidence, build theory, or make sense of complexity. Scholars disagree about whether these combinatorial processes should be labeled aggregation or synthesis (BARROSO & POWELL-COPE, 2000; ESTABROOKS, FIELD & MORSE, 1994; PORRITT & PEARSON, 2013). These disagreements align with differences in how the concepts are framed, prioritized, and enacted in practice. [25]

We find aggregation to be a useful concept to draw on since it helps to highlight certain aspects of combining, collaborating, and synthesizing that seek to build interpretive strength into large scale qualitative studies. In TRAVIS, aggregation also enables us to more systematically reflect on what participants and researchers are doing to transform individual accounts into data that will be combined and compared at various levels (MILLER et al., 2019 [2017]), without the apt but vague label of multi-sited ethnography (MARCUS, 1998). [26]

In the TRAVIS project, aggregation occurred in three ways. First-level aggregation took place in courses wherein the guided autoethnography exercise was conducted. In each course, we combined participants' individually produced autoethnographies to generate a single dataset for the local researchers to analyze as a whole. Second, the international TRAVIS team synthesized its country-specific datasets toward themes, reformulating them according to group standards to be more usable for answering the project's research questions. Third, we aggregated iteratively since in the interpretative process different pathways towards further analysis appeared, and we combined and integrated data accordingly. These are discussed in more detail below. It is worth mentioning that students themselves conducted the initial level of aggregation. They narrowed the focus of their analysis, filtered in or out certain information in making field notes, edited what they submitted and shared with the research team; students were already comparing, combining, synthesizing, integrating, and culling the data. As researchers, we had access to some of their logics—students submitted their fieldwork plans, reflections on what worked and what didn't, notes on adjustments of the fieldwork plan, and reflections on doing the fieldwork and coding. These materials were read by the TRAVIS researchers doing the aggregation, coding and interpretation. Just like field experiences are knitted into an individual autoethnographer's interpretive apparatus, so were students' thoughts, hesitations and feelings incorporated into TRAVIS team's reflexivity guiding interpretations and argument building down the line. [27]

We created initial datasets by combining individual student autoethnographies for each course in which we conducted this exercise. We treated the whole bundle of submitted material (field notes according to the format they were taken in, notes, memos, brain dumps, links, inserted images and screengrabs) as one unit of data. The initial aggregated datasets were thus a combination of selected, donated student autoethnographies in each course, which also mostly functioned as initial country-specific datasets²⁾. However, it is important to remember that not all enrolled students completed their autoethnography tasks or chose to donate them to the project. [28]

At the moment of aggregation, a shift in analytical scope occurred. When conducting fieldwork, all participants were autoethnographers of their own experience, full members of their chosen networked publics and communities. For the TRAVIS team, however, these materials were not personally autoethnographic, but rather a form of elicited material. Although, as described in the previous section and will be discussed more in detail below, we did make a number of efforts to integrate contextual information on participants' experiences and their metacommentary into our processes of interpretation. [29]

Further, when creating the first-level data-sets, we discarded some student autoethnographies. We based inclusion and exclusion on the presumed autoethnographic quality (e.g., we excluded autoethnographies that did not seem to engage with the experience in depth and lacked reflexivity, most often because they had taken the self-quantification route). Further, all aggregation is driven by a guiding question and our questions could differ from, or at least be agnostic to the questions that had guided the students. This means that we also excluded autoethnographies based on topical misalignment. For example, one of our participants had focused on content pertaining to Russia's war in Ukraine, instead of everyday wellbeing, so we discarded it. In another case however, a student had also focused on the war, but in their notes and memos pondered the relationship between that content, its presumed trustworthiness, their own different ways of consuming that content and their (mental) wellbeing, in which case we included it in the dataset. Sometimes we had to exclude materials because the students had not paid attention to how their social media experiences intersected with their experiences of trust. However, if related social norms (e.g., authenticity, legitimacy, expertise, reliability, credibility, resonance) were evident, we still included the materials. [30]

4.2 Synthesis and integration: Reformulating first-level datasets into second-level datasets

After aggregating the first-level datasets, we coded the data relying on a project-wide, collaboratively prepared coding scheme (although additional codes were added by individual coders where needed). Our coding scheme is quite detailed, with 29 parent-codes pertaining to creators and their characteristics, platforms and their features and algorithms, practices and imaginaries, content with its perceived types, genres and aesthetics, and of course a variety of trust criteria. Here, another layer of aggregation happens, where different country-based first-level datasets (interviews³⁾ and autoethnographies) are synthetized and combined into a second level dataset. This produces a new dataset comprising coded snippets. To exemplify, we offer a selection (see Table 1) of snippets coded as "visual trust"—one of the child-codes in our "trust criteria" parent-code⁴⁾. First, we included three snippets from one particular Austrian autoethnography (all from the field notes), to showcase how the same person kept returning to the theme; then two snippets from an Estonian autoethnography (one from a field note, one from a brain dump, to showcase different formats of autoethnographic self-expression by the same person), and finally, a snippet from a UK interview to show how a similar topic emerges in autoethnographies and interviews as well as across different cultures. We anonymized the usernames and removed the screenshots, but marked where they were included in our data. Each country team was responsible for coding the local first-level data set in their native language, as well as preparing analytical memos.

By aggregating the first-level data, the autoethnographies became "searchable," and available for a different level of interpretation. Our second-level datasets were concurrently larger, yet narrower. Synthesis and reformulation removed particular cases (which the researchers can always return to), instead working with snippet data that relate to salient themes prioritized in this phase of aggregation. This illuminated repeating patterns and potentially interesting questions. At the same time, the snippets still involve links to social media posts, screenshots of visual content and various levels of meta-commentary from the autoethnographers, allowing researchers to retain some of the original context. [32]

4.3 Iterative re-integration of (parts of) second-level data sets for multinational arguments, insights and follow-up questions

At this point of writing, we have started but not yet concluded, interpreting and integrating across second-level data sets. The project team has a messaging channel for sharing observations, questions, interesting snippets, hypothetical samplings of patterns or themes. In these, a team member might elevate a fascinating observation (e.g., the tacit link between trustworthiness and aesthetic quality evident in participants' field notes noted in Table 1). If this observation resonates, it gets prioritized for a systematic integration and focused analysis exercise that we have been calling "data sessions." [33]

By now, we have conducted four team-wide "data sessions," (e.g., on body image, visual aspects of trust, ways of seeing and visualization of medical professionals, props and places) and a number of smaller ones with special interest sub-teams (e.g., on sexual health, authenticity, affordances etc.). As part of a data session, a lead team or a lead researcher identifies topics that come up in more than one country and requests specific translated sections of second-level data-sets. Each country team includes with the submitted, translated data a reflexive memo on linguistic or cultural specificities. [34]

In the past data sessions, for example, differences in how the word "trust" and "authenticity" are used or not used came up as relevant. The memos also comment on the data, and what might not be evident in these coded snippets, but should be taken into account when interpreting them. This might include the interviewer's notes on the interview-process, or first and second-level aggregator's notes on what other codes might be relevant to look at. The lead team then aggregates the data, condensing it into what we call a data session packet, including also a written memo on the criteria and logics of condensation. The data packet is shared with the international team. Everyone analyzes the provided data and each country team creates a memo of preliminary interpretations, which again includes reflexive meta-notes on included and excluded strands of interpretation and other logics the researchers followed. These are shared, discussed and taken forward by interest-based teams working towards coherent arguments. [35]

In our data sessions so far, smaller (mixed country) groups discussed their respective interpretations first and then we aggregated our findings in a plenary. This has allowed us to identify and agree upon essential elements of our research object, and to discuss focal points of further analysis. The sessions are critically important to introduce another layer of reflexivity (now of team members' interpretative positionalities), to make different cultures/habits of interpretation transparent and to reflect on the process and level of analysis we aim for. [36]

All data packets for these data sessions have included both textual and visual material. However, in the first three sessions, we prioritized particular codes, with each team pulling snippets from written field notes, transcribed interviews etc. All visual content that organically attached to that text (i.e., the screenshots inserted into field notes or interview transcripts, links to posts with videos) was included, but the orientation was to textual data. In our fourth data session we inverted the dynamic, focusing specifically on visual content (images of health professionals, props and places) which was submitted into the data package with relevant textual data, e.g., the text that surrounded and contextualized the visual content in the field notes or the sections of the interview discussing the visual content. [37]

Compared to other forms of data elicitation we have employed over our research careers, people's autoethnographic field notes have some notable characteristics, in particular when it comes to what is often described as the richness and denseness of data (BECKER 1970; LATZKO-TOTH, BONNEAU & MILLETTE, 2017). Further, aggregating those autoethnographies opens to analysis some aspects of everyday practice and meaning-making that are difficult to access otherwise. This is a result of what guided autoethnographies make possible on an experiential level for those enacting them, what they make possible as ways to create material for interpretation, and what their aggregation makes possible for researchers. In the first instance, the methodological affordances of the approach come from how both the "guided" and the "autoethnographic" aspects of the process direct participants to engage with their experience and operationalize it into an empirically observable and recordable activity. It heightens participants' reflexivity towards their technology use, the tacit assumptions and expectations built into their everyday routines, the feelings embedded in and emerging from their often-opaque relationships with platforms, algorithms, creators and themselves. [38]

In the second instance, the methodological affordances of our approach are linked to generating a data set of multiple participants' experiences that is imbued with the landmark thickness of (auto)ethnographic description, but allows for detection of patterns across individual experiences and to an extent, cultural contexts. This is because of how the data are coded (snippets with included links, screenshots, metacommentary), because of what is included as data (we coded not only field notes, but also brain dumps and memos) and due to the iterative procedure of aggregation (with built-in memos for researcher reflexivity). Of course, combining the two levels also has specific limitations. In the following discussion, we examine the limitations of the approach and elevate its strengths. [39]

Like all approaches, aggregated autoethnography has its limitations. First, it is a complex and labor-intensive process at every stage, from guiding participants through data collection, providing feedback, aggregating individual accounts in ways that still maintain nuance, to analyzing the vast quantity of varied material. The intensive scaffolding of the guided autoethnography is necessary for the type of commitment needed to generate rich materials. While it is feasible in an academic setting where incentives like grades exist, it might be very complicated in non-classroom contexts. Further, it is important to carefully consider academic traditions and power differentials when relying on data generated in the classroom, to make sure that the consent given when participants donate their material is genuine. Aggregation also creates a specific sample of data, which, while not problematic within the tradition of qualitative inquiry where we don't generalize to a population, and where samples reflect a variety of self-selection and subjectivities, is still something to be reflexive of when interpreting and making arguments. [40]

Second, the epistemic shift from studying the self to aggregating multiple self-narratives introduces a layer of complexity that also demands reflexivity which, when operationalized into research procedure (i.e., via our data sessions described above), is also time consuming and effortful. Autoethnography traditionally centers on personal experience, but aggregation transforms individual insights into a set of data. Compared to a conventional auto-ethnography where fieldwork, interpretation and analysis are done by the same person, aggregation brings with it some loss of the data thickness in terms of contextual awareness, evoked memories, or a deeply personal sense of routines. However, aggregation allows for a broader reach that, as argued throughout the text, makes possible for a different scale of patterns to emerge. Further, we argue that if compared to other methods that combine elicited information about one's social media use across a number of people (e.g., from a number of individual interviews), aggregated autoethnography still offers significant contextual and processual insight. [41]

Finally, working with aggregated autoethnographic materials precludes the opportunity for real-time follow-up. Unlike interviews, where researchers can probe further when something interesting arises, aggregated autoethnographies work with what has been recorded in whatever detail it was recorded. Intriguing but under-examined moments do not reach their full potential unless the research design allows for subsequently interviewing, and even then, the participants may not be able to recall what it was that inspired a field note. [42]

That being said, aggregated guided autoethnographies also boast a number of strengths. First, they allow access to the unexpected. The approach can offer interesting forays into topics we would not know to ask about in interviews and the interviewees would not know to bring up. This pertains both to the content and communication on social media, the experiences and evaluations it triggers, but also the practices, routines and ways of doing things. The autoethnographers are often surprised and share the unexpected tangents and twists of direction in their brain dumps and notes. As Ruby, one of our participants wrote in her field notes:

"It feels strange this realization that most of the time I spend on Instagram is actually a passive activity that still affects me in ways I probably never consciously notice. I laugh, then I am surprised, I send some videos to friends, but then I immediately go back to scrolIing. It is kind of an ‘obsessive passive' behavior." [43]

Linked to, but epistemologically different from the previous point, aggregated autoethnographies also provide insight into participants' mundane encounters, which they would not be able to notice, consider significant, or worth mentioning when recollecting their experiences during an interview, even if pre-interview prompts were to invite such thinking. Whenever the research questions are about cultures, vernacular forms of communication, norms and practices—e.g., phenomena notoriously difficult to capture with traditional social scientific methods but persistently important when trying to make sense of people, groups and societies—access to the tacit and the routinized is extremely useful. It is not in the first, but the fifth field note like: "I opened TikTok after my last lecture for the day to take my mind off uni for a bit until I have to get back to my laptop to do assignments," that Valerie, another one of our autoethnographers, herself and / or we, the research team, recognized that something is indeed a routine. [44]

The autoethnographic data are also really good at bringing out inherent contradictions in human experiences, thus allowing us access to the ambivalent, as well as to shared patterns of ambivalence. For example, in Table 1 above, one of the autoethnographers linked visual quality and professionalism to their experiences of trust. This came up across autoethnographies; participants would link quality or lack thereof both to increasing and decreasing sense of trust in contradictory ways, allowing the team to notice the ambivalence of the trustworthiness of visual quality as a recurring pattern. Comparing this to interviews, we argue that as a form of storytelling, interview accounts tend to undergo a process of narrative linearization (BRUNER, 1986; POLKINGHORNE, 1988). The data they provide is inescapably story-shaped; made to cohere in subtle ways. Guided autoethnography data offer access to a less-story-shaped rendering of the various experiences of life. On an epistemological level, these data are from a different level of the symbolic-interactionist process of sense- and self-making (TIIDENBERG et al., 2017). [45]

Finally, aggregated autoethnographies are also less of a snapshot, because participants collect the data and think about them over a period of time (we recommend at least a month), thus the approach allows access to the dynamic. The process helps foreground different aspects of participants' experience and perhaps, in particular, the dynamic nature of the experience—the daily processes and practices of experiencing visual social media and their situatedness, where some actions are routine, but the content flows in various contexts and is encountered in various personal situations allowing participants as autoethnographers and us as researchers to notice patterns pertaining to ways of looking, motivation for engaging, emotional "triggers," networked forms of verification that only emerge as noticeable over time. It is this extended engagement and the attention to shifting perceptions that allowed one of our participants to notice that:

"[I]t's almost as if the social validation from my friends' engagement adds an extra layer of credibility and appeal to the content. This validation creates a sense of connection, knowing that people I know have enjoyed or found value in what I am about to watch." [46]

Social media visuality is inherently complex. The historically and culturally shaped ways of making meaning of, on and with images converge with the non-neutrality of technologies, the economies and politics of their ownership and governance, and vernacular ways of using, refusing and misusing them. To make sense of how visuality plays out in these complex ecologies, we need to find ways of integrating concepts from different fields. Classic visual studies ideas regarding negotiated interpretations, indexicality, representation and intertextuality (BERGER, 1972; MITCHELL, 2005) continue to be relevant but intersect with notions of platform vernaculars, socio-technical affordances, attention economies and algorithmic imaginaries (BUCHER & HELMOND, 2018; GIBBS, MEESE, ARNOLD, NANSEN & CARTER, 2014; MARWICK, 2015). How this becomes a research agenda can (and should) take many different forms; we have described but one possible approach in this article. [47]

Broadly, in the call for papers for the special issue Digital Images and Visual Artifacts in Everyday Life: Changing Media—Changing Uses—Changing Methods the editors asked whether any new concepts are needed to describe the social life of images today. In the case of the TRAVIS project, we have answered the question by elevating two. Neither is new per se, but both have renewed importance in how we study visuality in the context of social media. These are concepts of practice and trust. First, we argue that the multifaceted ontologies and epistemologies of socially mediated visuality invite a focus on process; on the doings, sayings, and feelings that are enacted through and attach themselves to visual content, visual communication and their social and platformed spaces of circulation. Social media visuality, therefore, is always necessarily practiced. Second, the increasing social and political weight of visual content and communication in the context of its concurrent ubiquity, banality and persuasiveness (HAND, 2013; MILNER & WOLFF, 2023) demands centering trust both as a social norm and a relational practice. [48]

To study social media visuality as a practice in ways that allow exploring something so mundane and taken-for-granted, yet so abstract and slippery as trust, significant methodological reflection is needed. We have responded to this need with an aggregated autoethnography approach. Our hope is that sharing it will be a source of inspiration to colleagues interested in studying socially mediated visuality beyond questions of trust. Our approach guides participants through the process of becoming autoethnographers of their digitally-mediated experiences and then undertakes a layered exercise of aggregation, integration and synthesis. As an approach to data collection, guided autoethnographies provide rich insights which allow for a type of granularity and nuance difficult to attain via other established qualitative methods used in digital research. Aggregating those autoethnographies allows for the synthesis of multiple perspectives, revealing patterns across contexts that are otherwise difficult to detect. [49]

We suggest that the aggregated autoethnography approach lends interpretive richness to the analytical ambitions of large scale projects by allowing them to explore 1. the topics neither researchers nor participants would otherwise know to bring up (e.g., the unexpected), 2. the mundane and the routinized, which participants would omit when responding to direct questions (e.g., the tacit), 3. the inherent contradictions of everyday life that often get smoothed out by the linearization of narrative (e.g., the ambivalent) and 4. the shifts and changes, but also that which stays the same over a period of time (e.g., the dynamic). These four characteristics contribute towards the sort of "thick description" interpretations (GEERTZ, 1973) that can help us better understand the fine-grained interrelations of visuality, practices, algorithmic processes, sociality, and trust. Further, the aggregation then allows noticing patterns not only in what the participants are experiencing or how they are framing it, but also in their everyday embeddedness in visual social media, their routine experiences with digital technologies. We can explore the procedural, relational and practice-related patterns that imbue social media visuality with meaning and social import. [50]

The aggregated autoethnography approach moves back and forth between the focus on singular pieces of visual content or single instances of visual communication and the flows of content they are part of. The flow might consist of the many posts on a particular creator's account in one situation, an algorithmically personalized feed in another, and a hashtag public or a search result in a third. The situatedness of a particular unit of visual content in a particular flow importantly shapes its meaning—a selfie in a flow of other selfies all marked by a shared hashtag will be interpreted differently compared to a single selfie in an account where no other selfies are present. This, in turn, invites consideration of how the flow relates to a particular site of meaning making in the analytical sense—e.g., do we want to study it from the perspective of the creator of each of the selfie posts, from the perspective of a scrolling audience member, from the perspective of the circulatory logics of attention on the platform (ROSE, 2022 [2001]; SCHREIBER, 2017, 2020; TIIDENBERG, 2015, 2018)? These specific relations determine which flows we need to capture or enter as field sites, what research questions can be answered and what arguments are made. [51]

Further, the aggregated autoethnography approach also allows concurrent capture of relevant artifacts and people's practices. This makes it eminently suitable for studying cultures characterized by vernaculars of remix and produsage (BRUNS, 2008; MARKHAM, 2013a) which all socially mediated visual cultures are. The guided steps of fieldwork push participants towards a perspective mindful of the affective and embodied aspects of socially mediated visuality, thus precluding the approach from becoming divorced from the embodied, material, lived experiences and their classed, raced and gendered circumstance. Finally, on a more practical level, it offers a solution to the problem of choice researchers face in the context of overabundance of data when working with networked, digital communication. Here, those whose experiences become data make the choices of what to include themselves. [52]

1) Project TRAVIS is supported by the Estonian Research Council, Academy of Finland, Austrian Science Fund, FWF, Economic and Social Research Council under CHANSE ERA-NET Co-fund program, which has received funding from the European Union's Horizon 2020 Research and Innovation Programme under Grant Agreement no 101004509. <back>

2) In Austria we ran the autoethnography exercises at two universities, so the dataset emerged from two instances of aggregation. Both Austrian courses followed the same instructions and task timelines, and the participants even met up to exchange their experiences. All included data were stored on the same server and labeled and pseudonymized together. <back>

3) During the first phase of our empirical work and to answer our first research question about the trust experiences of lay social media users, we generated data via both interviews with and auto-ethnographies by student participants. <back>

4) Trust criteria was one of our central parent-codes. Combining inductive and deductive reasoning it became apparent that people's experiences and articulations of trust are complex, and trust can be thought of as consisting of various assessments and interpretations that may vary depending on the situation. Thus, we coded with a number of child-codes for so-called aspects, components or conditions of trust, i.e., "attention-seeking," "authentic/inauthentic," "coherent/incoherent," "competent," "credentials," "complexity vs oversimplification," "relatable," "repetitive," "validated," and others. <back>

Adams, Tony & Holman Jones, Stacy (2008). Autoethnography is queer. In Norman Denzin, Yvonna Lincoln & Linda Tuhiwai Smith (Eds.), Handbook of critical and indigenous methodologies (pp.373-390). London: Sage.

Anderson, Leon (2006). Analytic autoethnography. Journal of Contemporary Ethnography, 35(4), 373-395.

Are, Carolina (2021). The shadowban cycle: An autoethnography of pole dancing, nudity, and censorship on Instagram. Feminist Media Studies, 22(8), 2002-2019, https://doi.org/10.1080/14680777.2021.1928259 [Accessed: February 23, 2025].

Barroso, Julie & Powell-Cope, Gail M. (2000). Metasynthesis of qualitative research on living with HIV infection. Qualitative Health Research, 10(3), 340-53, https://doi.org/10.1177/104973200129118480 [Accessed: February 23, 2025].

Becker, Howard S. (1970). Sociological work: Method and substance. New Brunswick, NJ: Transaction Books.

Berger, John (1972). Ways of seeing. London: British Broadcasting Association and Penguin.

Bergdahl, Elisabeth (2019). Is meta-synthesis turning rich descriptions into thin reductions? A criticism of meta-aggregation as a form of qualitative synthesis. Nursing Inquiry, 26(1), 1-8.

Bochner, Arthur P. & Ellis, Carolyn (1992). Personal narrative as a social approach to interpersonal communication. Communication Theory, 2(2), 165-172.

Boll, Tobias (2024). Meet "me" in the field(-notes): The selves and self-relations of autoethnography. Qualitative Inquiry, 30(8-9), 674-683, https://doi.org/10.1177/10778004231196921 [Accessed: February 23, 2025].

Breckner, Roswitha & Mayer, Elisabeth (2023). Social media as a means of visual biographical performance and biographical work. Current Sociology, 71(4), 661-682, https://doi.org/10.1177/00113921221132518 [Accessed: February 23, 2025].

Bruner, Jerome S (1986). Actual minds, possible worlds. Cambridge, MA: Harvard University Press.

Bruns, Axel (2008). Blogs, Wikipedia, Second Life, and beyond: From production to produsage. New York, NY: Peter Lang.

Bucher, Taina & Helmond, Anne (2018). The affordances of social media platforms. In Jean Burgess, Thomas Poell & Alice E. Marwick (Eds.), The Sage handbook of social media (pp.233-253). London: Sage.

Charmaz, Kathy (2014 [2006]). Constructing grounded theory (2nd ed.). London: Sage.

Clarke, Adele E. (2005). Situational analysis: Grounded theory after the postmodern turn. London: Sage.

Davidjants, Jaana & Tiidenberg, Katrin (2022). Activist memory narration on social media: Armenian genocide on Instagram. New Media & Society, 24(10), 2191-2206.

Dubois, Elizabeth & Ford, Heather (2015). Trace interviews: An actor-centered approach. International Journal of Communication, 9(25), 2067-2091, https://ijoc.org/index.php/ijoc/article/view/3378/141 [Accessed: January 8, 2024].

Ellis, Carolyn; Adams, Tony E. & Bochner, Arthur P. (2011). Autoethnography: An overview. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 12(1), Art. 10, https://doi.org/10.17169/fqs-12.1.1589 [Accessed: March 14, 2025].

Emerson, Robert M.; Fretz, Rachel I. & Shaw, Linda L. (2011). Writing ethnographic fieldnotes. Chicago, IL: University of Chicago Press.

Eschler, Jordan (2016). A critical reflection on social media research using an autoethnographic approach. In Tung X. Bui & Ralph H. Sprague, Jr. (Eds.), 2016 49th Hawaii international conference on system sciences (HICSS), Koloa, HI, USA (pp.1871-1880). Los Alamitos, CA: Conference Publishing Services.

Estabrooks, Carole A.; Field, Peggy Anne & Morse, Janice M. (1994). Aggregating qualitative findings: an approach to theory development. Qualitative Health Research, 4(4), 503-511.

Geertz, Clifford (1973). The interpretation of cultures: Selected essays. New York, NY: Basic Books.

Gibbs, Martin; Meese, James; Arnold, Michael; Nansen, Bjorn & Carter, Marcus (2014). #funeral and Instagram: death, social media and platform vernacular. Information, Communication & Society, 18(3), 255-268.

Gomez Cruz, Edgar & Lehmuskallio, Asko (Eds.) (2016). Digital photography and everyday life: Empirical studies on material visual practices. London: Routledge.

Goodall, Harold L. (2001). Writing the new ethnography. Walnut Creek, CA: AltaMira Press.

Greschke, Heike (2023). "Positioning" analysis with autoethnography—epistemic explorations of self-reflexivity: Introduction to the special issue. Qualitative Inquiry, 30(8-9), 659-662, https://doi.org/10.1177/10778004231193762 [Accessed: February 23, 2025].

Hand, Martin (2013). Ubiquitous photography: Digital media and society series. Cambridge: Polity Press.

Holman Jones, Stacy (2016). Living bodies of thought: The "critical" in critical autoethnography. Qualitative Inquiry, 22(4), 228-237.

Joffe, Helene (2008). The power of visual material: Persuasion, emotion and identification. Diogenes, 55(1), 84-93.

Jorgensen, Kristian M. (2016). The media go-along: Researching mobilities with media at hand. MedieKultur, 32(60), 32-49, https://doi.org/10.7146/mediekultur.v32i60.22429 [Accessed: February 23, 2025].

Latzko-Toth, Guillaume; Bonneau, Claudine & Millette, Mélanie (2017). Small data, thick data: Thickening strategies for trace-based social media research. In Anabel Quan-Haase, & Luke Sloan. (Eds.), The Sage handbook of social media research methods (pp.199-214). London: Sage.

Leaver, Tama; Highfield, Tim & Abidin, Crystal (2020). Instagram: Visual social media cultures. Cambridge: Polity Press.

Li, Yiyi & Xie, Ying (2020). Is a picture worth a thousand words? An empirical study of image content and social media engagement. Journal of Marketing Research, 57(1),1-19.

Light, Ben; Burgess, Jean & Duguay, Stefanie (2018). The walkthrough method: An approach to the study of apps. New Media & Society, 20(3), 881-900.

Mann, Peter (2016). The research interview: Reflective practice and reflexivity in research processes. New York, NY: Palgrave Macmillan.

Marcus, George (1998). Ethnography through thick and thin. Princeton, NJ: Princeton University Press.

Markham, Annette N. (1998). Life online: Researching real experience in virtual space. Walnut Creek, CA: AltaMira Press.

Markham, Annette N. (2013a). Remix cultures, remix methods: reframing qualitative inquiry for social media contexts. In Norman K. Denzin & Michael D. Giardina (Eds), Global dimensions of qualitative inquiry (pp.63-81). New York. NY: Routledge.

Markham, Annette N. (2013b). Fieldwork in social media: What would Malinowski do?. Journal of Qualitative Communication Research, 2(4), 434-446.

Markham, Annette N. (2018). Critical pedagogy as a response to datafication. Qualitative Inquiry, 25(8), 754-760.

Markham, Annette N. & Pronzato, Riccardo (2023). A critical (theory) data literacy. Information and Learning Sciences, 125(5/6), 293-320.

Marwick, Alice E. (2015). Instafame: Luxury selfies in the attention economy. Public Culture, 27(1), 137-160.

Marwick, Alice E. & Boyd, Danah (2014). Networked privacy: How teenagers negotiate context in social media. New Media & Society, 16(7), 1051-1067.

Mavroudis, Jonathan (2018). Fame labor. In Crystal Abidin & Megan Lindsay Brown (Eds.), Microcelebrity around the globe (pp.83-93). Bingley: Emerald.

Miguel, Cristina (2016). Visual intimacy on social media: From selfies to the co-construction of intimacies through shared pictures. Social Media + Society, 2(2), https://doi.org/10.1177/2056305116641705 [Accessed: February 23, 2025].

Miller, Daniel; Costa, Elisabetta; Haapio-Kirk, Laura; Haynes, Nell; Sinanan, Jolynna; McDonald, Tom; Nicolescu, Razvan; Spyer, Juliano; Venkatraman, Shriram & Wang, Xinyuan (2019 [2017]). Contemporary comparative anthropology—The why we post project. Ethnos, 84(2), 283-300.

Mitchell, William J. (2005). What do pictures want?. Chicago, IL: Chicago University Press.

Müller, Michael R. (2011). Das Körperbild als Selbstbild. In Michael R. Müller, Hans-Georg Soeffner & Anne Sonnenmoser (Eds.), Körper haben: die symbolische Formung der Person (pp.87-106). Weilerswist: Velbrück Wissenschaft.

Polkinghorne, Donald E. (1988). Narrative knowing and the human sciences. Albany, NY: Suny Press.

Porritt, Kylie & Pearson, Alan (2013). The historical emergence of qualitative synthesis. Philadelphia, PA: Lippincott, Williams and Wilkins.

Reckwitz, Andreas (2003). Grundelemente einer Theorie der sozialen Praktiken. Zeitschrift für Soziologie, 32(4), 282-301.

Reis, Shmuel; Hermoni, Doron; Van-Raalte, Riki; Dahan, Rachel & Borkan, Jeffrey M. (2007). Aggregation of qualitative studies—From theory to practice: patient priorities and family medicine/general practice evaluations. Patient Education and Counseling, 65(2), 214-222.

Robards, Brady & Lincoln, Sian (2017). Uncovering longitudinal life narratives: Scrolling back on Facebook. Qualitative Research, 17(6), 715-730.

Rose, Gillian (2022 [2001]). Visual methodologies: An introduction to researching with visual materials (5th ed.). London: Sage.

Roth, Wolff-Michael; Breuer, Franz & Mruck, Katja (Eds.) (2003). Subjectivity and reflexivity in qualitative research II. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 4(2), https://www.qualitative-research.net/index.php/fqs/issue/view/18 [Accessed: March 14, 2025].

Schatzki, Theodor (1996). Social practices: A Wittgensteinian approach to human activity and the social. Cambridge: Cambridge University Press.

Schreiber, Maria (2017). Audiences, aesthetics and affordances. Analysing practices of visual communication on social media. Digital Culture & Society, 3(2), 143-164.

Schreiber, Maria (2020). Digitale Bildpraktiken: Handlungsdimensionen visueller vernetzter Kommunikation. Wiesbaden: Springer VS.

Schreiber, Maria (2023). #Strokesurvivor on Instagram: Conjunctive experiences of adapting to disability. MedieKultur: Journal of Media and Communication Research, 39(74), 50-72, https://doi.org/10.7146/mk.v39i74.132468 [Accessed: February 23, 2025].

Sondheim, Alan (Ed.) (1996). Being online: Net subjectivity. New York, NY: Lusitania Press.

Spradley, James (1980). Participant observation. New York, NY: Holt, Rinehart and Winston

Spry, Tammy (2001). Performing autoethnography: An embodied methodological praxis. Qualitative Inquiry, 7(6), 706-732.

Stone, Rosanne Allequere (1995). The war of desire and technology at the close of the mechanical age. Cambridge, MA: MIT Press.

Tiidenberg, Katrin (2015). Great faith in surfaces—A visual narrative analysis of seflies. In Airi Alina Allaste & Katrin Tiidenberg (Eds.), "In search of …": New methodological approaches to youth research (pp.233-256). Cambridge: Cambridge Scholars Publishing.

Tiidenberg, Katrin (2018). Selfies, why we love (and hate) them. Bingley: Emerald.

Tiidenberg, Katrin (2025/Forthcoming). Visual social media. In Martin Hand, Barry Sandywell & Victoria Szabo (Eds.), The Bloomsbury encyclopedia of visual culture: Digitisations, transformations, and futures. London: Bloomsbury.

Tiidenberg, Katrin; Hendry, Natalie Ann & Abidin, Crystal (2021). Tumblr. Cambridge: Polity press.

Tiidenberg, Katrin; Markham, Annette N.; Pereira, Gabriel; Rehder, Mads; Dremljuga, Ramona; Sommer, Jannek K. & Dougherty, Meghan (2017). "I'm an addict" and other sensemaking devices: A discourse analysis of self-reflections on lived experience of social media. Proceedings of the 8th International Conference on Social Media & Society (#SMSociety17), Toronto, Canada, July 28-30, https://doi.org/10.1145/3097286.3097307 [Accessed: February 23, 2025].

Wolfinger, Nicholas H. (2002). On writing fieldnotes: Collection strategies and background expectancies. Qualitative Research, 2(1), 85-93.

Zhao, Sumin & Zappavigna, Michele (2018). Beyond the self: Intersubjectivity and the social semiotic interpretation of the selfie. New Media & Society, 20(5), 1735-1754.

Katrin TIIDENBERG is professor of participatory culture at the Baltic Film, Media and Arts School of Tallinn University, Estonia. She is the author and editor of multiple books on social media, digital visual cultures and digital research methods. She is currently leading an international research project on visual digital trust (TRAVIS) as well as the Participatory Wellbeing Research Group of the newly funded Estonian Centre of Excellence of Wellbeing Sciences. Her research interests span social media, digital cultures, networked visuality, internet governance and self-care.

Tallinn University
Baltic Film, Media and Arts School
Narva rd 25, 10120, Tallinn, Estonia

Annette N. MARKHAM is a long-time digital culture researcher who specializes in ethical research practice and creative methods for studying multi-modal and digitally-saturated social contexts. She is currently professor of future media literacies and public engagement at Utrecht University.

Utrecht University
Department of Media and Culture Studies
Muntstraat 2A (room 0.02), 3512 EV Utrecht, Netherlands

Maria SCHREIBER is a postdoctoral researcher in the Communication Studies department at the University of Salzburg. She focuses on visual practices and representations and their entanglement with platforms, digital cultures, biographical and social transformations. She is one of the chairs of the European Communication Research and Education Association (ECREA) section Visual Cultures, and a principal investigator in the Chanse project TRAVIS.

University of Salzburg
Department of Communication Studies
Rudolfskai 42, 5020 Salzburg, Austria

Andrea SCHAFFAR works as a post-doc researcher. She is part of the Austrian TRAVIS team at the Department of Communications at the University of Salzburg. She teaches at several universities in Austria. In her work, she focuses on the use of qualitative methods and methodologies at the intersections of different disciplines. Her research and teaching areas include media studies and media pedagogy, urban studies and housing, and social dynamics. Additionally, she works as a trainer and consultant in the field of organizational development and group dynamics.

Tiidenberg, Katrin; Markham, Annette N., Schreiber, Maria & Schaffar, Andrea (2025). Patterns of surprise and ambivalence: Studying social media visuality by way of aggregated autoethnography [52 paragraphs]. Forum Qualitative Sozialforschung / Forum: Qualitative Social Research, 26(2), Art. 10, https://doi.org/10.17169/fqs-26.2.4182.

Forum Qualitative Sozialforschung / Forum: Qualitative Social Research (FQS)

ISSN 1438-5627

Funded by the KOALA project

Creative Commons Attribution 4.0 International License

Source	Snippet	Comment
Austrian field notes	"So-called 'reality checks' are a well-known example of body positivity on Instagram. In the Explore section of Instagram, I came across another example from the @username (screenshots inserted, link inserted). This is an album posting consisting of two images. One photo that was taken at a good angle and one that represents reality—without posing. This reveals things like a slightly protruding belly or a bit of love handles. The last three pictures in the album are quotes. The user is smiling in each of these photos."	The auto-ethnographer positioned all of these as their reflections and notes on trustworthy and untrustworthy images, and in particular, on trustworthy and untrustworthy practices by social media users, and the norms and expectations among audiences that surround those (e.g., showcasing the curated nature of content creation, the difference that posing makes). By noticing, over and over again, the claims to trustworthiness and authenticity
Austrian field notes	"I find the criticisms in the comments section interesting, as they suggest that the user is deliberately making herself fatter—exactly the opposite of what other users who want to portray themselves as slimmer are trying to do. Both are a form of distortion of reality—one to get reach for diverse body images, the other to avoid diverse body images on IG."	made by content creators via visual and genre tactics, the autoethnographer complicates straightforward interpretations of causal links between social media images and wellbeing, and explores the different discourses and actors (creator, the viewer, other audience members who comment) involved in social media meaning making.
Austrian field notes	"I am also amazed that you can change a lot with posing. These pictures show once again how important it is to surf social media in a reflective way (screenshot included). People who are constantly comparing themselves with others should be aware that some things are staged and edited. Posts like these create transparency."
Estonian field notes	"Most of her posts created by herself have this heavy amateurish vibe all over it. Same goes with the current post. But since the point of the account seems to be selling stuff and a certain lifestyle, then this amateurish aspect might be good, but in general it feels very unprofessional."	Here, the autoethnographer had a number of field notes like the one included, where they noticed themselves noticing the quality and professionalism of the visual content and sometimes having an emotional reaction to it (e.g., annoyance, or a feeling of
Estonian Brain dump	"Another thing that stands out is about the professional aspects of how the content was created. For some reason, if content is over-produced, with all kinds of lights and whistles added to, for example, I find it annoying, which impacts the emotional side of experiencing the post. I mean, less is more. The rule of editing a video (or compiling a post) is simple—if the element you want to add does not make the video (or post) better or worse, then it is not needed. Add some flashy details only if it makes the content better. If not, then it is going to direct the attention away from the idea, message and meaning of the post towards the flashy/cool/colorful aspect of the content. I have also noticed that if the content is simple and has a certain amateurish vibe, then it creates this kind of warmth about it that has a more positive effect on me. I feel like I am not a customer, but the person behind the post is really interested to get the message to me and that he/she finds this message also very important."	affinity). In the brain dump, the autoethnographer moves to what BOLL (2023) might call the retrospective self-narrative, where the self-positioning is less of immediate experience and more of a preliminary, tentative search for patterns in terms of one's own experiences and reactions. The research team noticed such internal negotiations in assigning meaning to visual quality of content on social media across participants and countries.
UK interview	"And then I find it quite quickly becomes quite disingenuous to me, where I'm just like, a lot of these women are like—they are really like—they've got really nice bodies. None of them are overweight or in bad shape. [...] So I find it quite disingenuous sometimes, when I feel like they're like forcing themselves into like an awkward shape, or taking an unflattering angle, in quite a—it's almost like in a very staged way. But the opposite of staged, where everything's perfect. It's very staged in a way to make themselves look as bad as possible."	The interviewee is reflecting on their perceptions of body-image photography and how it is framed by creators. They also reflect on themselves noticing how a particular framing of "unflattering angles" (also mentioned by the Austrian autoethnographer), which makes appeals to trustworthiness, might misfire.

1.		What kind of health and wellness-related visual social media content do you trust and distrust? (photos, selfies, memes, infographics, short or long form videos, etc.)
	a)	What does your (dis)trust depend on (e.g., the platform/app; the creator; the exact topic; what is depicted; the genre; something else entirely)?
2.		How and why do you trust health and wellbeing-related visual content? What are indicators of trustworthiness for you?
3,		How do specific features and functions of platforms and apps shape your experiences of trust? Do different apps and platforms feel like different spaces in terms of trust?
4.		What are your "trust rituals" on social media? What do you routinely do?
	a)	Do you do anything to make sure that you can trust what you consume?
	b)	Does trust seem to have any part in how enjoyable your social media use is?
5.		When you create content yourself, do you think people trust it?
	a)	Why, why not? Does it matter to you?
	b)	How do you enact trustworthiness?

1.		Make a plan. Your plan should specify:
	a)	How often and for how long will you "do observations?"
	b)	How will you take notes? Pen and paper? Word document? Excel spreadsheet? A private social media account to which you can post and repost with explanatory captions? Do you need to combine notes, screenshots and brief spoken memos you dictate into your phone? Do you need to sometimes film yourself (and your screen)?
	c)	What are the important observations to include (i.e., location, how you felt, what you were doing, how other people acted, what interactions were had, what you noticed, emerging patterns of use, intentions of participation, unintended consequences of participation etc.)?
2.		Observe your own experiences of trust in the context of visual social media for 3-7 days. Be honest, be thorough. Try different strategies for observing, note taking, and gathering data.
3.		Reflect. Create a short (2 min., or 250 words) reflection, where you discuss first thoughts, reactions and feelings regarding your observation. You may want to focus on a particular instance of trust or lack thereof, or you may want to focus on what feels like a budding thought regarding your participation on social media in the context of health and wellness-related visuals and trust.
4,		Analyze what works and what doesn't in your observation plan and revise your plan. Adjust the plan so it serves you better. The aim is to have an observation plan that allows nuanced exploration of how you experience visual trust on social media. It is very likely you will decide to significantly narrow your focus here (the point is NOT in quantitative self-tracking). Write a brief (~200 words) commentary on how and why you changed your plan.