The notion of what constitutes my ethnographic ‘field’ continues to reappear in various situations. Sometimes this is from people who know better than me that I need to articulate precisely what I mean by it, and sometimes it’s from people less familiar with ethnography who can’t conceive what an online field might be. Traditionally, ethnographic fieldwork, and more specifically participant observation, is marked by a number of factors. It assumes the ethnographer will be resident in a limited geographical locale in which they experience face-to-face relationships (Wittel, 2000) with an ‘object’ of study – an ‘Other.’ There will be clearly identified boundaries where it is straightforward to establish what is included and also what is excluded. In an online, digital, virtual or cyberethnography, residence and geography have less meaning, interaction is mediated and boundaries blur. The ethnography becomes one of movement and flexibility, responding to the ebb and flow of the people and practice under study. My field then becomes one of the people I follow and those they bring into view; the learning practices (and others?) in which they’re engaged; and the areas into which they take that practice. Twitter usually, though not always, provides the point of entry to that field; there I might remain, or be whisked off elsewhere as I follow the actors.
An obvious question might be, ‘but isn’t your ‘field’ determined by those who you follow?’ Of course, that’s true and if I followed different people, my field might be very different. The platform might be the same, but the Twitter I experience as a result of those different people might be very different indeed. When I reach the point of having to inscribe my positionality to allow others to review my study, I feel I need to render more closely what the point of entry is, in terms of those who I follow – my followees.
Although I’m conducting a qualitative study, I elected to undertake one quantitative module within the Research Masters programme I’m also following – ‘Survey Methods.’ For the assignment, I chose to conduct a ‘survey’ of my followees, so that I might have a better sense of their demographics – gender, location, educational phase etc. This would help me to better appreciate what this part of my field is; at the bit determined by those who I follow. So for example, given my professional background, I might be mainly following secondary school physics teachers, or educational technology co-ordinators. I think that my followees are men and women, more experienced and new to teaching, from across the world and teaching a wide range of subjects and young people from within a range of educational sectors, but the truth is, I simply have no idea. Now I could have sent out a traditional questionnaire; a Google form perhaps. A lot of the data however, is already visible on people’s profiles, so why ask them to give up time when I could simply gather the data for myself … or perhaps find an automated process which might gather some of them for me. Well I did try, but in the end, this wasn’t a high enough priority to spend the time learning enough about data mining to gather all the data. During the study module however, I did learn far more about the process of sampling and how statistically significant results can be produced if a simple random sampling process is indertaken.
My first step was to produce a list of everyone I was following – the sampling frame, which on this occasion was the entire population, and from which the random sample would be drawn. Although I didn’t have the skill to draw down all the data I wanted, I had just enough experience with DataMiner to pull down a list of all my followees, together with some basic supplementary information from their profiles e.g. number of tweets, followers, followees. Importantly, I could then assign a random number to each of them, and from that construct a true random sample. To provide a reasonable representation of the population of my followees, I needed a sample of 344 individuals. However, when it came to the practicalities of collecting the data, the manual part of it proved incredibly time consuming – around six hours for only half the sample. Reality now kicked in and I decided I couldn’t spare the time to collect the full sample; I only needed a handful of results to illustrate my methodology for the assignment I needed to produce. The truth is I don’t really need the findings to be statistically robust since I’m only seeking an impression of the population I’m following. Here then is a summary composed from a random sample of 152 of those people.
There are few surprises in this opening set of results, other than the proportion of people that don’t follow me back. I (falsely) assumed that since I’m following approximately the same number of people who follow me, that if I was following a particular person/account, they would be following me back. Clearly not so. Since I wasn’t aware of it, I’d never really thought about the consequences. The audience I thought I had for my tweets was based on the people I had chosen to follow and now I know that’s clearly not the case. I wonder to what extent that’s the same for other Twitter users? All but one of my followees have public accounts, and the majority are in fact people, as opposed to parody or organisational accounts. Remarkably few chose to hide their whole identity; some might have an unusual handle, but still provide their real name, whilst some chose a graphic rather than photo for their avatar. Half identify as teachers and of those, only 60 in primary and secondary, the two sectors where my research focus is aimed. Does that matter? [spotted a discrepancy in the above graphic that I need to follow up] I appear to be following more women than men, but nowhere near the proportion of women which makes up the teaching profession in general. In choosing those who I follow, do I unwittingly attempt to follow roughly equal numbers of men and women … and if the proportion of men and women followees don’t correspond to that within the profession in general, what implications does that have for the observations I make?
The next set of data is related more closely to the behaviour and activity, rather than the attributes, of those I follow and who follow me. (In each chart, the red bar indicates where the data from my own profile would locate me)
It appears to be most likely that someone I follow will have tweeted between one and five thousand tweets; I’m slightly above that. Though what this means, I’m not entirely sure. For example, someone who only joined Twitter a couple of years ago, but has tweeted 15k tweets is likely to have a bigger impact in my timeline than someone who’s also tweeted 15k times, but opened their account in 2008. Perhaps tweet-rate might have been a better indicator of … well, of what? Engagement? This is when we begin to see how important the qualitative side to a study of this nature is likely to be. What effects do people with a high or low tweet rate have on those who follow them? Having generated just under 10k tweets, I’m around the middle of that set of data; does that mean I’m fairly typical? Similar to those who I follow?
Without unpicking each of the subsequent graphs in detail, it’s nevertheless possible to see I’m close(ish) to the middle, if not slightly above for each metric. However, it’s clear to see I’m far less prolific than the average user when it comes to ‘Liking,’ and have created far more ‘Lists’ than most of my followees do. I can offer some possible explanations for these differences, but more importantly, what I take away is that in some ways I’m far from a typical user. When I processed the figures for ‘Likes’ I couldn’t believe (based on my own behaviour) how many ‘Likes’ some people post; but then I only know my rationale for ‘Liking’ and not that of other folks. On the other hand, I seem to have created far more lists than most people do; around a third of my followees create no lists at all. That too seems strange from my perspective where I view lists as a powerful way of filtering the flow of data into more meaningful streams; why wouldn’t other people want to do that?
What have I learned?
Firstly I think I’m typical enough to be considered an insider; I’ve clearly not just parachuted in to conduct research with no intention of attempting to become part of the ‘community.’ This is also part of one’s ethical sensibility, which I discussed at greater length here. Whilst some lurking is inevitable as a participant observer, and is also an accepted and understood behaviour on social media, if one was to stretch that too far and make no positive contributions to the community, that would be far from ethical in this contributory space where sharing is the norm. So the figures seem to suggest I might be able to legitimately claim an emic or insider view, at least if I acknowledge that there are aspects of what I do that are somewhat different from the norm. But I guess that too is the same for many users; there is no ‘average’ user?
Secondly, I have to be incredibly careful in extrapolating what I do and what I see as being typical of other teacher tweeters. Here I think that part of the hidden materiality (the devices, the applications used, the environment where Twitter is accessed, the time of day) all become significant. One of the reasons I post fewer photos than many people is that I don’t access Twitter using a smartphone, so am less likely to fire off a quick photo/tweet of my current activity.
What began as a statistical exercise resulted in me exploring issues I probably wouldn’t even have become aware of, let alone have targeted for reflective consideration. I’m pretty sure there’s a lesson to be learned here.
Wittel, A. (2000, January). Ethnography on the move: From field to net to Internet. In Forum Qualitative Sozialforschung/Forum: Qualitative Social Research (Vol. 1, No. 1).