Learning From Feeds:

Understanding Others’ Perspectives/Experiences By Following a Lot of People Who Are Different From You

In the acknowledgements of I am Charlotte Simmons, Tom Wolfe’s 2004 novel about undergraduate life at a fictional elite university, Wolfe thanks somebody who “enabled a fool to rush into undergraduate nightlife where wise men never went.” In essence, Wolfe was thanking students for letting him be a “fly on the wall” observer at parties to better understand their perspective, which was very different from his own, so he could write about it.

I was reminded of this during a few recent experiences of my own. One morning, for example, I was sitting in my office watching an Instagram livestream of what seemed to be an alcohol awareness class at a rural high school in the south, streamed from the perspective of a student who was snarkily and audibly commenting along and giggling with friends. Another day I was watching a video of another high-schooler in the south (who is openly gay and has over 30,000 Instagram followers) burn couches in his backyard with his family (with the self-referential-but-detached hashtag #redneckaf). I was also seeing mundane pics of everyday teen life scroll by in my Instagram feed: many teens post one or so selfies per day, some post periodic videos of simply listening to music in the car, others share photos with friends in clubs at school or promotions for events, and many are just hanging out.

Yes, I know what you’re thinking: “You’re over 40. You’re an academic. What are you doing with this stuff in your Instagram feed?”

Please, let me explain. And just to be clear, I’m following quasi-public behavior on Instagram and scouting out possible venues for future research, but am not currently doing research in this way.

As one who studies online self-presentation, particularly among gay men, I’m often trying to see where people are hanging out, meeting and interacting. Gay teens are an interesting case because they may not be welcome on (or appropriate users of) gay dating/hookup apps like Tinder and Grindr, but often want to meet other gay youth. Given the popularity of Instagram among teens and its network-building potential, I wondered what was happening on Instagram.

Via some hashtag searches on Instagram, I noticed a few gay teens who had startling numbers of followers (several over 30K). I also discovered an account (with about 45K followers) geared toward young gay men that many of these teens seemed to follow. This account shares face (and only face — this isn’t a hookup site in disguise as far as I can tell) pics and usernames submitted by young men, to serve as a sort of social networking tool within the social networking tool that is Instagram. (I’m not sharing the name of this account here to protect user privacy.)

Out of sheer curiosity, I followed that account and started following a few of the other people I was finding. I’ve generally only followed people whose accounts have privacy set to “public” so are visible to any Instagram user. I’m now following people with a wide range of follower counts, from the upper hundreds to tens of thousands. The general age range (for people who list their age) is from 14–25 or so. I’ve focused on people in the US, though there are quite a few from Europe and other countries in North America, with a smaller number from other parts of the world. It appears to be only gay men who are active, though I’m not sure if L_BTQ+ individuals are explicitly excluded or not.

They’ve taken over.

It turns out (perhaps not surprisingly) that these young men are way more active on Instagram than my actual friends and colleagues who use the app, so they’ve more or less taken over my feed and “stories bar” (NB: The “stories bar” shows ephemeral and currently live content available from one’s Instagram contacts.)

Just by opening my phone, as the examples above illustrate, I get a very close-up and sometimes personal view of a world that would otherwise be very hard for me to see. It reminds me a bit of people who have written observationally about their local campus Yik Yak feeds (e.g., Amy Bruckman’s blog post and Amy Butcher’s recent NYT op-ed). Yik Yak is anonymous and location based, so its feed gives a sense of what’s going on in one’s physical area, such as on a college campus. In contrast, what I experience is not anonymous in that I’ve started to recognize the people I follow, and it’s not local in that these people are geographically distributed around the US.

As I alluded earlier, what I see feels more analogous to being a fly-on-the-wall observer (a la Wolfe) in these people’s lives. I pretty quickly became aware that they use Instagram in a way very different from my peers’ and my own use. I noticed, for example, how quick many teens are to like each other’s posts and how quickly a selfie can rack up a few hundred likes, how a seemingly possessive long-distance boyfriend would comment on each of his boyfriend’s selfies with a ‘lock’ emoji and the word ‘mine,’ and how some seemed to follow each other’s romantic pursuits very closely (noting in one of their own posts that a breakup may have happened between two others, for example, both of whom were tagged). I also noticed how many people used photos accompanied by song lyrics to reflect a mood, ephemeral videos to show where they were from, and livestreams as a way to engage their followers. I also saw evidence of potential interlopers, such as an individual who repeatedly urged a livestreaming user to take of his shirt on camera.

All of this matters, I’d argue, as academic social media researchers like me struggle to understand what people are doing online these days. Even in this era of ubiquitous big data, this can be a major challenge. There are lots of reasons for this, but a few that stand out for me are:

1. Popular social platforms may have hundreds of millions (or billions) of users, and people’s individual experiences vary wildly based on who they’re connected to and what they see in their personalized, algorithm-driven feeds;
2. Platforms are proprietary, so it’s tricky to see what people are doing without relying on sample posts provided by users, surveys of their behavior that lack detail or sitting with participants and looking at their content; and
3. When we do get big swaths of data via scraping or collaborating with industry colleagues, they tend to be huge collections of content largely divorced from context, so the data reflect people’s activities but not the context of those activities. Where Facebook status updates are seen by others are part of a feed, for example, looking at thousands of status updates from disparate users doesn’t give us a good sense of what a given user’s feed looks like.

We, as a research community, need better ways to broaden our understanding of how people use and experience social platforms by adopting a more localized approach that focuses on multiple communities in parallel, rather than on trying to characterize social platforms as distinct communities in and of themselves. This is especially true for people who are very different from ourselves and our typical undergraduate and/or MTurk participants. Examples might include teens or older adults, people in different countries or cultures, people with different political beliefs, etc.

My Instagram experiences provide a potentially useful lesson here.

That is, watching a feed full of content from people very different from me sometimes felt like it could be a research method. For simplicity, let’s call this a proto-method and refer to it as ‘feed observation.’ Essentially this is an adaption of existing field study methods for the venue of a social feed on a platform like Facebook or Instagram. Doing research in this way, though raises some important new versions of old questions about benefits and drawbacks of this way working, ethical questions about how exactly to do the work, and the validity and reliability for addressing different questions. The purpose of this post is to raise and consider some of these questions about feed observation that we’d need to discuss if it were to be a full-fledged method.

What are the benefits?

First, let’s consider some reasons why doing research in this way might be valuable. We as a community already do lots of research on people’s social media experiences. We survey people, we use computational linguistic processing and statistical data science tools to analyze content, we interview users, we (ask them to share and then) look at their posts, we look at people’s feeds with them and ask us to tell us about the content, we use content coding to sort posts into categories. Why do we need another method?

One benefit is that feed observation would let researchers see people’s day-to-day, mundane posting behavior and — in doing so — get a sense of what they perceive to be normative behavior on a platform, what seems to be important to them on an ongoing basis and what it’s like to experience a daily feed of images from within a particular network.

In this way, it’s different from sampling from people’s posts or running a “logger” app to capture all of a particular user’s posts, or even using a screen recorder to log all of the posts that a particular person sees in a session. Logging and recording tools let us access content, but don’t provide a good sense of the overall setting for content within a feed, the temporal patterns of posting, or the dynamics of interactions around content, such as liking and commenting.

Another benefit is that observation can be done inconspicuously and unobtrusively, with the researcher being just one of many platform users who happens to be following a set of other users. Some of those users may notice the researcher as a follower and/or follow the researcher themselves, whereas others may not. A small but noticeable number of the people I’ve followed in my short experience have started following me. All of this, of course, raises ethical questions and I’ll talk about some of them below.

Next, feed observation could allow researchers to observe ephemeral behavior and see details that otherwise might get lost or forgotten in, say, an interview description or a selection of posts to share retroactively. These ephemeral behaviors are important, though, because it is through hanging out together and everyday interactions that relationships are built and sustained. This attribute becomes even more important as ephemeral posting via Stories and Livestreams become more common. As many of the Instagram Livestreams I’ve seen have involved fairly mundane “hanging out” and some engagement with remote viewers, this is a pretty fascinating view of others’ worlds, how they interact with their Instagram followers, and what sort of live interactions are taking place.

Feed observation also allows us to see how different groups or communities of people use particular features and/or affordances of social platforms. As researchers we sometimes assume that features are used the way we or our small set of participants use them, but that’s not always true. Some research on livestreaming (see, e.g., a recent CSCW panel), for example, suggests that there is usually an event or incident being streamed, but my informal experience suggests that livestreams are often used by some people just to “hang out” or even as a sort of “check-in” to show where one is at a particular moment. We might also assume that ‘stories’ are used to share ephemeral content, but miss that (again, based on my informal experience) these are also used to draw attention (and likes) to permanent posts. Indeed, I’ve learned that posting a selfie ‘story’ with the word ‘recent’ means the poster wants their followers to go ‘like’ their ‘recent’ regular Instagram post.

What About Ethics?

As with any method for studying human behavior, feed observation raises some interesting and important ethical concerns. It’s important to note here that my example of gay teens is an extreme case. That population raises questions specific to research with adolescents, but I’m not going to focus on those here because there’s a large body of work on this population and IRBs are generally well prepared to address these. Instead, I’ll focus on the specifics of feed observation.

Researchers have long discussed the ethics of observing online communities (see e.g., Amy Bruckman’s chapter about teaching students to do this; there’s also an active effort to update ACM’s Code of Ethics and Professional Conduct so it covers some of these issues). Feed observation shares some commonalities with online community observation in that a researcher would be observing others’ behavior in a quasi-public online context.

One key question is whether there is a “reasonable expectation of privacy” on Instagram and how strong this expectation is. Some might argue that no observation should take place at all without informed consent, though Hudson and Bruckman argued in a 2004 study of IRC chat rooms that a waiver of consent is appropriate in these cases because merely stating one is doing research in a chat room can affect behavior significantly. It also raises questions about what it means when users of a social platform set the privacy on their content to “public.” Do they do this because they want anybody to see it? To make themselves easier to find? Should we treat these behaviors as we do other behavior in quasi-public spaces? And if we only study public content, Casey Fiesler and colleagues recently pointed out that this itself might introduce some biases to factor in.

In any event, it’s important to protect participants’ identity and the privacy of their content in publications or presentations by not sharing identifiers or content directly and being careful to avoid using information that could be looked up to later identify them (note: I slightly altered the hashtag I mentioned above for this reason).

A big difference between feed observation and online communities lies in how and where the boundaries of a ‘community’ are drawn. Because feeds are personalized, there’s not always a clear and shared notion of community, and it is not always the case that a particular, discrete set of people are interacting only with each other. It is nonetheless true, though, that there are sets or clusters of people who are affiliated by common interests (e.g., young gay men) and find each other via hashtags, shared friends, etc. As such, it’s hard to know who is part of the community and who is not, and when the researcher him/herself becomes part of it. Is it enough just to join Instagram and follow some number of people, or must one be noticed? Is it different to follow users that have 200 followers vs. those with 20,000? There are some clear parallels here to ethical concerns that have been raised in the study of platforms like Twitter or Reddit, and specific sub-communities (e.g., Black Twitter).

Elastic and fluid boundaries make it tricky to determine when one is observing (or participating in) community behavior, and whether anybody is aware that the researcher is there watching. If one did need to announce that research was taking place or document consent, moreover, these fluid boundaries would make that difficult at best.

Another ethics question is how to treat very different types of content. Where we typically focus on a particular type (text posts, photos, chat utterances, etc.) and try to get lots of it, feed observation on platforms like Instagram, Facebook and Twitter would mean seeing many different types of content, and many different types of behavior. Some of the content becomes a permanent part of user profiles and timelines, whereas some of it is ephemeral and disappears. This raises questions about whether it’s ok for a researcher to capture or record any of this.

In the case of livestreaming it also foregrounds questions about expectations of privacy. I’ve seen, for example, streams of people just hanging out in a living room with their friends. A living room is typically a place where we’d say there’s an expectation of privacy. Does livestreaming change that? And does it change it in a way that’s different than it would for sharing a photo of people in the same living room?

These are just a few questions I’ve considered, but I know there are more and I’m eager to hear from you.

Validity and Reliability: What can we ask?

As with any possible method for research, it’s important to think about the kinds of questions that could be addressed, and the challenges to reliability and validity that are presented.

One key point is that it’s probably reasonable to make claims about the behavior of individuals that one is following, but harder to make broad experiential claims about the content of the feed itself as a unit of analysis. In the case of the gay teens on Instagram, for example, I know (from informally seeing comments and likes) that many of them follow each other. But I also strongly suspect that they follow and are followed by their straight, local friends and perhaps some family. So while my feed likely has a fair bit of content in common with theirs, it’s unlikely that I’m experiencing a feed that’s exactly like theirs. We need to be cautious about this.

The benefit described above of being able to observe ephemeral and in-the-moment content, of course, also — from a reliability standpoint — increases the imperative on the researcher to regularly check for ephemeral content and to do so at the times when this content is likely to be generated (e.g., often on evenings and weekends for some categories of users). This is especially true for livestreams which — on Instagram — do not remain visible to contacts after the stream ends. Signing in only at particular times of day or year could have significant consequences, just as it could for observations in physical places (i.e., imagine observing teens at a local mall on Wednesdays at 10:30am in October vs. July).

And this, of course raises, questions about external validity and generalizability. Seeing a particular subset of people behave in a particular way does not mean that all people (or even all people in that category) behave in that way or that we as observers are correct in interpreting our observations.

This means first that we must be careful in making claims about how all people in a demographic swath behave vs. seeing just a few of them. I have little doubt, for example, that there are differences between the gay teens I see on Instagram and their friends who have fewer followers or who post fewer mirror selfies. So it wouldn’t be fair to say that all teens post lots of mirror selfies, but it’s not uninteresting that these teens do.

It also means that this method, as with any observation, is probably best used in concert with other methods such as interviews and surveys with some members of the observed population to validate interpretations and see if any key behaviors are missed. For example, what interactions between users are we not seeing by focusing on a feed, such as private messages or interaction on other platforms like Snapchat or Messenger.

Moving Forward

So, dear reader, I’m posting this to get your feedback. Is this a possible way to do research, do you think? If it is, what else do we need to watch out for? If it’s not, why do you think it isn’t? Are there conditions under which you think it could be reasonable? What are your other thoughts? Please let me know by commenting here, on Facebook or by emailing me. Thanks!