I quite like the word ‘reactions’ that Facebook is using to describe their new options to express feelings about a post. I wish I’d thought of it. This is a matter of much more than passing interest to me as it relates closely to something that occupied a lot of my time over some years of my life. In my own CoFIND social bookmarking system (that first saw the light of day about 18 years ago and underpinned my PhD work) I used to refer to something quite similar as ‘qualities’ – metadata (tags) to show not just that something is good or interesting but how it is good or interesting, that could then be used to rate and thus help to filter and rank a feed of bookmarked resources. CoFIND is an acronym – Collaborative Filter in N Dimensions – that refers to this n-dimensionality of ratings. Facebook’s Reactions feature is a simplified version of this: it’s about categories more than tags, but the thinking behind it is broadly similar. The differences, though, are interesting.
One of the things that is most notable about Facebook Reactions is that ratings are, like its Likes before them, binary: a simple ‘yes’ or ‘not-rated’. In most versions of CoFIND (it iterated a lot), users could choose to what extent something was good/loved/annoying/interesting/etc through a Likert scale. Giving the option to choose the strength of a feeling seems much more sensible when talking about fuzzy values like this. I want to be able to signify that I quite like something, or that is is mildly amusing, especially if my intent is to communicate my feelings to others. Facebook’s Reactions are a coarse as a means of expression: it is quite appropriate that its emoticons are literal caricatures. In all the methods I tried – radio buttons, clickable links, etc – introducing scalar ratings turned out to be way too complex to be usable, but web interfaces were not as rich in those days: I think things like popup draggable sliders (not dissimilar to Facebook’s interface) might make it more feasible nowadays.
Facebook Reactions are not just binary but fixed. CoFIND – I think, still uniquely – allowed individuals to create new qualities (reactions), which could then be used by anyone else. It was an n-dimensional rating system where ‘n’ could be any number at all. Qualities quite literally evolved for each community, with more used qualities surviving (being immediately available for use) and less used ones being relegated to backwaters of the system (effectively dying, albeit with the possibility of resurrection if added again). This allowed for such metadata to provide a mirror of the values that mattered most within a given community or network, rather than being imposed uniformly on everyone, and for those values to evolve as the community itself evolved. While I appreciate the simplicity of Facebook’s interface (CoFIND’s most fatal flaw was always that its interface was far too complex to be usable) I still think that user-created ways of emoting – what I have since called ‘fuzzy tags‘ – lead to much more useful reactions that matter within a given community, especially when users can choose the degree to which a fuzzy tag applies. When CoFIND was used in an educational setting, qualities like ‘good for beginners’, ‘authoritative’, or ‘comprehensive’ tended to emerge – they were pedagogical metadata. When used in other contexts, such as to discover what HCI students considered important in a website, site-ranking qualities like ‘slow’, ‘boring’, ‘artistic’ and ‘informative’ appeared.
One of the things I hate most vehemently about Facebook is that it same-ifies everything: a person in Facebook has a single unchanging (and permanently reified) identity, with a single network, a single facade, a single caricatured way of being in the world, notwithstanding the odd nod to diversity like pages and lists. Facebook’s business model relies on this, because any clustering or parcellation reduces the potential to connect, and connections are everything to Facebook. This makes me highly sceptical of its claimed ‘discovery’ that people are actually separated by only 3.57 degrees rather than six. Given that the system very deliberately drives them to friend as many others as much as possible, on most tenuous grounds of connection, this is hardly surprising. It shows not that previous studies are mistaken but the extent to which Facebook has manipulated human networks for profit. Apart from evolving to fit a single community, another of the things CoFIND did was to deliberately parcellate the environment, allowing different sets of values to evolve in different contexts. What is ‘good’ in the context of learning to read is not likely to be ‘good’ in the context of learning geometry, so different topics each evolved a (largely) separate set of qualities. This might not have been the best way to drive the growth of large networks, but it was a much better way to enable the self-organized emergence of meaningful communities. It also allowed individuals to express and embrace different facets of themselves, which in turn made it easy to accommodate changing needs and interests: essential in the context of learning, which is (if nothing else) about change.
You can read about the tortuous process of CoFIND’s development and the thinking behind it in my PhD thesis. I continued to develop CoFIND into the mid 2000s but, though the final version was a bit more usable and scalable (I rewrote it in PHP and changed a lot of the mechanisms, simplifying a fair number of things, including losing the fuzzy ratings) I’m still most fond of the final version that is described in the thesis.