Slides from talk on: "Possibilistic prediction and risk analyses"

Arguing for an approach for complexity scientists/modellers to interact with those making decisions (policy, business etc) in situations, in a way that does not deprive those decision makers of responsibility and which leaves them in control, whilst informing them.

A talk given at the EA Conference, Bonn, May 2015.

Abstract:

It is in the nature of complex systems that predictions that give a probability are not possible.

Indeed I argue that giving "the most likely" or "rough" prediction is more harmful than useful.

Rather an approach which maps out some of the possible outcomes is outlined. 

Agent-based modelling is ideal for producing these - including, crucially, possibilities that could not have been conceived just by thinking about it (due to the fact that events can combine in ways that are more complex than the human brain can cope with directly).

A characterisation of the real future possibilities and their nature allows some positive responses to events:
* putting in place 'early warning indicators' for the emergence of identified possibilities
* contingency planning for when they are indicated. 

Such an approach would allow policy makers to better 'drive' their decision making, without abnegating responsibility to experts

 Slides availabe at: http://www.slideshare.net/BruceEdmonds/be-eatalkfinal