Some Thoughts on Social Media for Science

What follows are some more or less connected thoughts on what social media for science could and should be. There are excellent articulations of what a social media future for science might look like, such as the multiple articles and blogs by Bjoern Brembs. This is not that! Instead, I’m trying to articulate for myself some constraints, tensions, and road blocks to such a future. My hope is that deeper discussion of those can help move us forward.

So here my basic constraints:

We live in a climate emergency. Our remaining carbon budget for 1.5 degrees is apparently virtually exhausted. It seems obvious to me that I can’t just go back to pre-pandemic levels of travel (which were already unconscionable). My research, in the grand scheme of things, simply isn’t that important. Scientists should be leading the desperately required transformation of how we do things, and that means fundamental change.

In person meetings are helpful, and chance encounters at conferences have repeatedly turned out to be unexpectedly valuable for me, but the climate emergency would dominate that even if conferences had been a hugely efficient way to advance actual research for me, which they were definitely not (maybe only my talks have audiences that include people preparing their own slides and catching up on mail…). Most importantly, the pandemic completely transformed my sense of what can be achieved through online contact only. I developed a whole range of deep professional collaborations and personal connections with researchers that I haven’t ever met face to face. This includes phd students and early career researchers, which matters because this is the group who genuinely most needs conferences to forge contacts.

Many of those contacts arose from the initiative scibeh.org, an attempt to bring the wider expertise of the behavioural science community into the pandemic response, specifically the many and diverse researchers of all career levels who were not previously engaged in policy advice. As part of that, we both used Twitter and we set up dedicated discussion forums on Reddit. Blog posts detailed both the project and its problems (see here and here), so I’ll limit myself to my main takeaways from trying to set up, moderate, and maintain a high quality forum for discussing science:

This indicates (severe) limits on participation and volume of quality content that, I strongly suspect, will not only exert themselves in *any* online discourse format, but be the dominant factor in how things develop. So we should, ideally, be building based on those known constraints and their implications for design, rather than trying to wedge science discourse into extant tools.

So we need much more than just online conferences plus a relocated science twitter. What should that more look like? That’s the big question (and one we should be throwing academic knowledge and research activity at..). It’s easier to say what it shouldn’t look like and I’ll stick to that here. It probably shouldn’t be

Perceived personal value and value to science are very much not the same thing, and we shouldn’t just think about what people personally want out of social media, but also how that aligns with science’s wider goals and constraints. The best online medium for science discourse will be the one that people most prefer to use from among the options and designs compatible with those constraints. Arguably, the relative weight of ‘what best serves science’ will become greater the more ‘successful’ social media are and the more we encourage social media use as part of the academic job. And the more an academic is using social media in a way that foregrounds their professional role (title and affiliation in bio, posts about their science..) the more normal rules of science should, arguably, apply. “My feed, my rules” becomes more problematic the more substantive scientific debate and dissemination of scientific opportunities shift online.

In the time I’ve been an academic, there have been huge changes in what counts as appropriate in work related discussion and where it takes place. When I first started, the focal event was the Friday post-seminar visit to the pub. That faded away in good part because of people thinking more about participation, the exclusionary effects of environments, and issues around informal networks of influence. It might just be me, but I see a lot about online social media that currently seems poorly aligned with changes that have occurred in those other contexts.

The social aspect of science. It might just be me because my own biggest disappointment with science (relative to what I imagined science would be like as a phd student) has long been how sociologically determined it is: ‘who knows who’ matters so much more than I ever imagined it would and so much more than I think it should given that the ideal type of science involves the co-creation of a body of knowledge on the basis of evidence itself. The ideal of science discourse involves the exchange and evaluation of arguments, not evaluation of positions based on the people putting forward the arguments. One thing I’ve noticed about social media is how much of the description of its utility to scientists often sounds like fuelling those sociological factors as if they were simply a good thing: use social media to “broaden your reach”, “raise your profile”, “make contacts”. It is not at all obvious to me that the wide spread use of platforms that are organised around the core concept of “followers” doesn’t inherently amplify (what strike me as) distorting features of science, as opposed to fostering those aspects we want. There is some role in science for giving certain people a privileged platform: we have long invited researchers with a measure of track record in the field to give key notes, for example. But those are carefully selected and, importantly, circumscribed and delimited opportunities: once that person has spoken, and the conference is over, so is their moment in the sun. It is a totally different proposition to have scientists with online platforms with tens or even hundreds of thousands of followers. Such online accounts have been handed a megaphone which is in no way restricted to the topic of their expertise. Crucially, those scientists are doing nothing wrong when they use that platform to muse publicly about some interesting new thing on which they have little to no prior expertise. They, like everyone else, are legitimately expressing an opinion in the discourse. They may even have an interesting, unique, outsider perspective. So the problem (if indeed there is one) lies not with that expression, but with the outsized attention it is then paid. One could argue that all of this is harmless, but in a world of finite resources information invariably competes. Time spent reading and discussing this post or issue, is time spent not discussing something else. I can’t really see such allocation of attention as anything other than likely suboptimal, just as I can’t see collective epistemic outcomes being improved by paying more attention to those whose communications happen to be witty. And it seems to me the more one believes in the power of online social media for ‘reach’, ‘profile’ and ‘contacts’, the more one inherently believes in the possibility of distortions.

Online social media don’t have to be structured this way. While Twitter (and replacements) are structured around people, other platforms like Reddit are not. I've been phantasising about the possibility of building a platform that is both: an inside-out-internet, as it were, like sweaters that have one pattern worn straight up and another one when worn inside out. Imagine being able to take a collection of posts on a microblogging platform like Twitter (an utterance graph) and re-rendering them as an argument graph that assigns different tokens of the same argument or opinion to a single, once displayed, type. And all relevant answers are connected to that type, regardless of when and where they occurred in the discourse. Imagine then being able to toggle between the two views.

There are many problems such a possibility would address: in particular, the reduction down to unique arguments would make it much easier for expertise to assert itself. It’s inherent in the nature of expertise that there are fewer experts than non-experts. One unfortunate manifestation of this is the futility of countering misinformation on a platform such as Twitter where the same mis-or disinformation can be re-started endlessly, and the small group of experts will find it impossible to keep up. While that constitutes an extreme end of the spectrum, the same underlying structural problem also plagues scientific discourse. The number of readers (and reviewers!) that can fully parse and critically evaluate all details of a paper are an (often very small) minority among potential readers. This structural feature of expertise suggests that we need to help people find expertise rather than naively assume it will come to assert itself in the discourse. By levelling the playing field, at least in numerical terms, an argument focussed display can help that.

An argument map also downplays the temporal dynamics: in a source based discourse like Twitter, you might miss that one vital post and now it’s gone. In an argument map, it’s given its rightful place, and can be viewed asynchronously. The overall discourse itself is also freed from the need to be hugely stimulating in real time (see 4. above). And, if nothing else, reduction of discourse to an argument map would allow one to gauge easily how productive an exchange actually is: does it contain nuance and counter-arguments or is it just endless repeat.

How close to technically possible is this? No idea, but NLP argument extraction and summary tools are getting better and better. In the absence of such a tool, we should think carefully about the extent to which we want to structure exchange around arguments or around people, and what kind of scale and time frame a particular exchange merits. In an ideal world, we would have a range of options at our disposal that we could flexibly choose from, moving our exchange from one format to another as constraints demand.

The long and short of all of this, however, is that I think we should be using this moment in time to really start building something better and that means something different to anything we've had in the past.