Why Model Dynamic Decision Making?

Abstract

This chapter outlined the motivation for modelling dynamic decision making and introduced related background material on human decision making, situation awareness, the BDI model, team modelling, and the need for transparent agent models in mixed human/machine teams.

The modelling of dynamic decision making is important in a number of application areas including (i) capability analysis (simulation of tactical scenarios in order to identify gaps in capability); (ii) training (realistic simulation of non-player characters in a virtual environment); (iii) tactics development and evaluation (using decision-making models in a constructive simulation context to explore tactical options); (iv) autonomous systems (UxVs whose domain-specific decision-making competence needs to be on a par with that of humans); and (vi) human/machine teamwork (transparent agents that can explain their behaviour at the level of goals and intentions). In developing TDF, we were interested in those particular applications areas and so have focused on how to engineer effective and understandable dynamic decision-making systems, rather than emphasising other aspects such as the fine-grained modelling of human performance at the millisecond level, which is more of a concern for psychology research.

Following on from Dennett [19], TDF is grounded in the intentional stance and the BDI paradigm, as this has been shown to be a good folk psychology model of human reasoning (i.e. how we think we think), and so facilitates the development of decision-making systems that can be understood by domain experts and interacted with by human team members. Although TDF is rooted in the BDI model, it is agnostic with regard to the content of the decision-making models; one is free to overlay higher level frameworks such as RPD and the OODA loop.