Sunday, 17 April 2016

Learning Analytics and the Futile Functionalist Alchemy of the Data analyst

I've been playing with R - trying to find something intelligent to say about a set of forum posts. They're very nice posts, and I'm sure the people making them got something out of it at the time when they submitted them. But one of the problems with analytics is that it's always trying to say something more about things which are ephemeral and about which there isn't really that much to say after the event. If I was a commercial organisation trying to sell people things, then perhaps I'd feel different. I'd take a punt on some of my speculative conclusions about the people posting messages and fire a few adverts at these people, and see if they bought something. If it works, I'll do it again and again and become rich! But then again, what I'd be doing is making the people fit my data by re-representing the data back to them as 'suggestions'.

I find with data analytics tools like R that a kind of fever comes over me. Manipulating matrices can be exciting - and one is tempted, having performed one transformation, to have a go at another - eventually some 'gold' will emerge. The problem is that we become so blinded by the technicalities of it all, and the hubris that some gold is there to be discovered, that it is easy to forget what it is we are looking for. It's all breathless and soulless. Where are the real people?

My deeper brush with analytics has concerned the Triple Helix, and the use of Shannon's information equations to model 'discourse dynamics'. I think there is some kind of discourse dynamic. It forms part of the environment within which we all live. But the important thing is that it only forms part of the environment. The internet itself, which is the domain which people tend to focus on when analysing discourse - is an even smaller part of the environment. Where's the rest of it? Well, we can't really see it, but each of us knows something is there. I prefer to call it 'constraint'. The Triple Helix is interesting because it acknowledges 'constraint' and uses Shannon's equations to attempt to get a handle on constraint.

The problem is that our understanding of constraints is constrained by constraints which we do not understand. Having said this, there are constraints which we do understand very well. The constraints of the market, for example. Or the constraints of tyrants. The effect of these constraints may have an impact on our discursive utterances, and this impact may perhaps be studied. But whatever we say about it is shrouded in constraints. Fundamentally, I'm re-stating Goedel's incompleteness theorem.

The curious thing is that learning is, in the final analysis, about overcoming constraints. Whether those constraints are fear, or physical ability, or intellectual challenge, when we overcome our constraints new things become possible. If we are to have a humane and sensible education system, then we may have to overcome the constraint that is 'learning analytics'.