27th March 2012

Post date: Mar 27, 2012 5:30:00 PM

A discussion on the 'computer-assisted reasoning' robust decision-making approach in papers from Lempert et al. (originally RAND Corporation). The 2002 PNAS paper gives an overview, with RAND Corporation's 2003 technical report going much deeper. There is also a 2003 paper by Bankes looking at using ensembles of models inside the scenario generator.

The ideas here are at the heart of the interface between social simulation and policy, and brought up questions such as on the actual process with stakeholders (e.g. when might subjective likelihoods of different plausible futures be used, or when would a strict 'all outcomes have to be taken as of unknown likelihood' be applied?) and--- crucially--- the nature of the scenario generator and the plurality of theory that it does or doesn't represent.

There was some later discussion on the nature and extent of prediction, and reference made to Troitzsch's 2009 discussion on the subject (a comment on previous papers, which echoes a previous group discussion).