Monday 23 April 2012

Visualising Visualising

Visual analytics is the fast emerging field where data is presented graphically and interactively for individuals to explore. The problem that it addresses is simple and fundamental: too much data; too little meaning. The visual analytics approach aims to provide users with interactive tools for allowing them to explore the data in a variety of ways identifying what appears to be meaningful, and what appears to be superfluous. There has been a lot of work going on in this area, including a large EU project called Vismaster (see

I've complained in this blog about the learning analytics/visualisation movement's obsession with pretty pictures (see and, and I have argued that the principal objective of any kind of analytics is decision and control. Too frequently this objective is lost as we become hypnotised by the pictures. The fundamental question is always "what does it mean?"; not only "what does it mean?" to me, but "what does it mean?" to us collectively (as we try to steer our institutions, learners, etc).

Recently I have also tried to get a grip on 'meaning'... or (I guess) the meaning of meaning. I am quite convinced by the arguments presented by Leydesdorff, following von Foerster's ideas of eigenform and Krippendorff's ideas around 'content analysis' that meaning has to do with expectations or anticipations. For Leydesdorff, meaning is "the structuring of expectations" (see I like that. The meaningful moments when I look at a beautiful sunset seem to me precisely moments where my anticipation and expectation is transformed; somehow the gears crunch together in a new way. The same is the case with the meaningful moments in music: at the climax, the expectations shift, the elements come together. This is probably similar (although a bit more precise) to Koestler's bisociation model (see

But there is a problem here and it has to do with abstraction. I love to think abstractly (this blog is a testament to that!!), but the dream of abstract thinkers is somehow to put the world in a formula or a model - or a 'visualisation'. To have the answer. Yet whilst we are (I am) driven to think like this (intellectual progress depends on it), it's a false hope. Experience happens over real time whilst models and formulae abstract time. The time of experience is not the same as the time of formulae or models. Their time is the time of clocks. The time of experience, however, is more fluid. Time ebbs and flows, sometimes (at deeply meaningful moments) standing still.

The meaningfulness of me writing this now is meaningful now. In the future, all I have is a vestige of what might have been meaningful at some point. But yet, there is a vestige of meaning in the document I produce, and as I browse my archives, my experience might recapture some essence of the meaningfulness that was experienced at the time of writing. This is a kind of hermeneutic analysis, and it has  been practiced for centuries.

But to come back to the problem which visualisation tries to address: too much information, too little meaning; No decision and control. If abstractions can't deliver the goods, then what might?

This is where I find the 'visual analytics' work most interesting. Because in visual analytics, what is created are experiences for interactively engaging with the data. So we have an original experience in time; we have a visualisation of the data from that experience which is also experienced in time. This means that users exploring the data explore their meanings of the data through interaction with it, and relate those meanings to the meanings of original experiences.

But this visualisation process could go deeper and more recursive. What about visualising the visualising of the data? or visualising the visualising of the visualising of the data? That's fascinating me because it starts to look like a von Foerster 'eigenform'. It also conforms with the fractal mathematical models of Dubois that desribe anticipatory systems.

But most interestingly this 'recursive visualisation' deals with the problem of abstraction by always having time and process as the fundamental underpinning of a meaning-making process of exploration. So time isn't stuffed into an equation or a pretty picture; it remains part of the process. In this way, analysing data and looking for meaning starts to look more like analysing music.

No comments: