Against Narrow Optimization and Short Horizons: An Argument-based, Path Planning, and Variable Multiattribute Model for Decision and Risk.

This paper proposes a mathematical approach to analysis of decision and risk that makes use of the constructive argument logics that have become commonplace recently in artificial intelligence. Instead of requiring an idealized, expected utility analysis of alternatives, in this paper, arguments appraise the desirability, comprehensiveness, and acceptability of incompletely described projections of the future. Instead of a qualitative risk management assessment process, threats and mitigations are represented numerically, but appraised with arguments, especially probability arguments and mitigation arguments, not averages. Arguments are given for or against the adequacy of commitments. Instead of using logic to derive the properties of acts that transform situations, e.g., to construct goal-satisfying plans, in this paper, dialectical burdens are placed on demonstrating to a standard that investments and response policies will attain each milestone on a proposed trajectory. Trajectories are extensible and valuations are multi-attribute with varying completeness as knowledge permits. Superior trajectory specificity will be related to superior argument specificity. The resulting picture of decision is a mixture of search, probability, valuation, and risk management; it should superficially bear a resemblance to satisficing mixed-integer discrete time control and many recent approaches to practical reasoning through argumentation. It is intended as an alternative to narrow optimization, which permits easy sacrifice of externalities on the grounds that they are hard to measure as real values. It is also intended as an alternative to fixed horizon decision-making, which produces unsustainable extremizations.