Sunday, 1 May 2011

Agent-based modelling and Waves of Personal Regulation

I presented an agent-based model of communication the other week at the CAL11 conference in Manchester. The construction of agents in this model is one where each agent has three levels of personal regulation which relate to three types of communicative act that they can make: Exhortation, Disruption and Coercion. The communications that an individual makes are probabilistically related to their individual state at any point in time (whether their regulation is focused on Disruption, Coercion or Exhortation). This is a simple choice determined by the relative values of their communication.

To run the model, select the number of starting people (the first slider which says 'starting-...'), and click setup, then click go. You can change the transparency of communication by altering the comms-r slider. When people are red, they are unhappy! You can add individual people by clicking 'add person'. Clicking 'disrupt-every...' is interesting. You can click and drag people from one part of the screen to another (this is like building teams). And obviously you can slow it down.

Internal regulation is affected by the communications they receive from others, and there is a need (in order to remain happy and viable) to make communications which will result in communications which will continue the maintenance of wellbeing. Thus agents 'position' themselves with one another.

In my model at present, each regulating mechanism is basically a type of 'random walk' where, on perceiving new communications, the regulation of each individual at each level is adjusted by various 'deltas' (values which are added or taken away from the current value). Each adjustment to each regulating level has a consequent effect on other regulating levels, and furthermore on future communication.

In my presentation, I demonstrated how this setup could demonstrate some 'critical incidents' in the university e-learning strategy, like the use of champions to stimulate new practice, or the increase in transparency of communication to stimulated richer activity and greater cohesion amongst staff, or the effects of large-scale disruptions like changing the platform. But I think the reaction of individual agents is at the moment too 'quick' to adjust to new circumstances.

This is why I am asking whether my agents could instead have regulating systems which were based on waves rather than random walks. This might introduce new elements of unpredictability into their behaviour, and would also introduce some degree of latency in their responses, but I think all of these things are real: we have to wait for colleagues to 'come round' to new ways of thinking. In such cases, something is going on inside them... is it that they have to wait for cyclic (harmonic) functions to reach particular points before they feel able to change?

Another way of thinking this through is to think that each communication we make has a 'resonance' - past echoes in the environment have some bearing on what happens in the present. Echoes may decay according to harmonic rules... Sugata Mitra wrote a paper where a similar idea of playing with time with cellular automata produced some fascinating results... (see

No comments: