Decision Making "Biases" and Support for Assumption-Based Higher-Order Reasoning

Unaided human decision making appears to systematically violate consistency constraints imposed by normative theories; these biases in turn appear to justify the application of for­ mal decision-analytic models. It is argued that both claims are wrong. In particular, we will argue that the "confirmation bias" is premised on an overly narrow view of how conflicting evidence is and ought to be handled. Effective decision aiding should focus on supporting the contral processes by means of which knowledge is extended into novel situations and in which assumptions are adopted, utilized, and revised. The Non­ Monotonic Probabilist represents initial work toward such an aid. Conflicting Evidence Two broadly different approaches to conflict­ ing evidence have been adopted in systems that perform inference under uncertainty. In logic-based systems that are capable of "non-monotonic" reasoning (e.g., McDermott and Doyle, 1980}, if it is possible to derive a contradiction from a set of statements, then one or more of those statements must be false. Suppose, for example, we start from the following beliefs: Argument #I: p, p -+ q Argument #2: r, r-+ -q From these two arguments, we could infer an 1 This research was supported in part by Contract No. MDA903-86-C-0383 from the Army Research Institute. The COTR is Dr. George Lawton. 71 impossibility: the truth of both q and -.q. To remove the inconsistency, at least one of the beliefs responsible for it must be revised. We know we are wrong about p, p 4 Q, r, or r -+