Decision-making, for humans and machines, under fundamental uncertainty

Abstract:

I discuss my strategic research work, which informs the basic direction of DST's autonomous systems SRI under Project Tyche, in which we are developing or sponsoring the development of new approaches to Artificial Intelligence (AI), and the new Modelling Complex Warfighting SRI in Operations Analysis (OA). Decision-makers - whether human or machine - face the inescapable challenge of having to exercise choice under conditions of irreducible uncertainty, whereby the consequences of their choices will play out in a future that they can neither really predict nor totally control. By and large, this distinguishes the automations of the past fairly sharply from the autonomy of the future, and the OA methods and practices of today from the analysis needed to meet the Force Design challenge in the future. These distinctions are brought into sharp focus by the particularly extreme environments of military operations.

Decision-making is resource allocation under uncertainty, and uncertainty entails awareness (construction of rational beliefs in economics terms), and hence I view Force Design, Command and Control, and Autonomy fundamentally as economic problems (though the details vary). Economics has long held a kind of love-hate relation with uncertainty; the economic phenomena collectively known as the 2007-2008 Global Financial Crisis and subsequent extended slump has prompted a renewed interest in uncertainty. I will talk about the different kinds of uncertainty identified in the economic literature, in contrast to the unreasonably strong invariant condition assumptions that lie just under the surface of much existing AI and OA as well as in some areas of economic theory, and then discuss my work on connecting these concepts through ergodic theory and nonlinear dynamics to incompleteness, unsolvability and incompressibility phenomena. Rather than the somewhat amorphous view of uncertainty of the economic theory, the mathematics yields a view into a richly structured and beautiful universe of worlds of different kinds of uncertainty, and the basis for knowing how to handle them effectively by essentially shifting problem choices more than by refining solution methods.

In the latter part of the talk, I will again turn back towards economics for some somewhat concrete examples of how uncertainty plays out in real and simulated systems, and for inspiration from a long history of dealing with different kinds of uncertainty - albeit with varying levels of success - for developing effective approaches for decision-making. If time permits, I will also talk about some of the compounding economic factors of organisational decision-making that greatly magnify the complexity of Force Design, Command and Control, Autonomy and even inter-organisational collaboration, and how these factors play might be incorporated better into algorithms, analytical methods and human management practices.

Biography

Dr Darryn J Reid is a Principal Scientist in Defence Science and Technology Group.

He has been with DST since 1995 (then known as DSTO), and has worked in distributed systems, machine learning and artificial intelligence, interoperability, formal reasoning and logic, operations research, simulation, optimisation and optimal control, electronic warfare, intelligence analysis, missile targeting and control, command support systems, complexity, nonlinear dynamics and ergodic theory, web-based technologies, software development, hardware design, formal languages and model theory, theory of computation and algorithmic information theory, crowd modelling, theoretical and applied economics, philosophy of science and military theory.

He holds the degrees of Bachelor of Science in Mathematics and Computer Science, Bachelor of Science with First Class Honours in Mathematics and Computer Science, and Doctor of Philosophy in Theoretical Computer Science from the University of Queensland.

He has strong research interests in pure and applied mathematics, theoretical and applied computer science, philosophy, military theory and economics. He is also works as an artist in oils, pastel, soft-colour and various drawing media and specialises in portraiture. In other words, he knows just enough to understand how profoundly ignorant he is. He is currently trying to age as disgracefully as possible, with the support of his beautiful wife Julie and their wonderful son Tyler.

Date & time

1–2pm 14 Jun 2017

Location

Speakers

Dr Darryn J Reid

Updated:  10 August 2021/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing