Sets and Nets

If you are not involved in social computing, social network analysis, sociology or similarly network-focussed disciplines and interest groups, this is going to seem like a pretty odd thing to make a fuss about (surely this is all common sense) so you might want to look away now…

Of late I have been increasingly concerned that the field of social computing is becoming dominated by a single world-view. To an increasing extent, research in the area has become dominated by various forms of network analysis, almost completely excluding the ‘social’ part of the term and anything else that avoids talking about abstract connections. Sometimes it is taken to ludicrous extremes. The other day I reviewed a paper that purported to show a small world network structure in what were described as ‘communities’ of people whose only connection was that they had used the same tags to describe content they had uploaded. They had not tagged anyone else’s uploads, only their own. There was absolutely no interaction between individuals, even in a mediated or artefactual way.  They were as much a network as all the people in the world that like cake.

I think nets are great. Network analysis is incredibly useful: it provides a powerful and flexible tool for understanding interactions between people and objects, lets us gain rich insights into complex interacting and dynamic systems. But nets are only a small part of what makes the field of social computing interesting.

The main part of social computing is, of course, made up of people and all their multifaceted wonderfulness (gotta love em), which should make up the chief object of study in any rational universe but that are generally simply treated as nodes in social computing conferences and journals (which are not rational universes).

However a significant other class of object in a social computing system is made up of sets of things – people, resources, dialogues, groups and so on. Things in sets (interchangeably, collections, aggregations, classes or bags) don’t have a particular order or internal connections between them like networks. They are defined simply by membership (or, in fuzzy sets, degrees of membership). They are just things we lump together for some reason. We could even have sets of nets. Or nets of sets if it makes sense or is useful to do so. Or sets of sets. The point is, sets and nets are useful in different ways and for different purposes.

Sets are powerful tools, especially in an environment (like this one) where people naturally fall into them – classes, research groups, centres, schools and so on. If I am designing a system for learning in an environment like The Landing, it is at least as important to me to know who is in a class as it is to know about the connections between them. It matters that I see myself in a group (set) of people who like to think about online learning. It matters that I can find them, not because they are a network but because they have defined themselves (through tags and profiles) as part of that set. I guess you could, it you wished, see the person-tag-person triad as a second-order net, but why bother when the most natural thing in the world is to call us part of the same set? It’s also computationally way easier and less expensive to do. Of course, once I have found them then first order networks come into play and that’s important too.

Sets are also wonderful for for lumping, averaging, summing, counting, weighting and rating, comparing and sorting.

Sets are perfect when we want to find something and we know what kind of thing it is – in structured data, especially.

Sets are great when we want to model the entities in the world, to find out what kinds of things are out there, how many there are, what they are like, what most interests them as a whole. The vast majority of databases in the world owe their forms to set theory and are composed of sets.

Sets are fabulous tools to filter not just things that are in a set but also the things that are not. 

Sets are just made for harnessing collectives: for instance, tag clouds are based on sets, not nets – typically, we count how many times a tag has been used and weight it in the list by popularity. Similarly, sets are far better than nets for voting: it is mighty interesting that a knows b and b knows c but, in some contexts, it is way more important that there were two votes for x and one vote for y. 

While we could (if we were particularly obsessive) see nets in everything from atoms to galaxies and model almost all sets as nets, we would lose something important in doing so. The fact that we can even use words like ‘atom’ and ‘galaxy’ means that we have already lumped things into categories – i.e. we have put them in sets.  Sets tell us about what matters to us, how we categorise, how we lump things, what the world is like to other people. Sets describe identity. We need sets in order to begin to have nets, as almost every net we are interested in is composed of sets: i.e. things we have no need to analyse as nets and that would have a different meaning and identity if we did.

I am part of the set of people who teach at Athabasca University. I may be part of a similarly named network too, but the fact that I perceive myself as a member, not just joined as a node in a network, matters. Those overlapping sets are as much what defines me as the connections I explicitly or implicitly form.

Nets are cool and most things can be seen in terms of them. But sets are cool, and most things can be seen in terms of them too. Clumping matters as much as connecting. 


I am a professional learner, employed as a Full Professor and Associate Dean, Learning & Assessment, at Athabasca University, where I research lots of things broadly in the area of learning and technology, and I teach mainly in the School of Computing & Information Systems. I am a proud Canadian, though I was born in the UK. I am married, with two grown-up children, and three growing-up grandchildren. We all live in beautiful Vancouver.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.