Cognitive illusions and the lying machine: A blueprint for sophistic mendacity.

摘要

This dissertation concerns the development of a prototype lying machine---a mechanical sophist whose disingenuous rhetoric seduces humans into believing falsehoods. The development approach is to empower a machine with a predictive psychological theory of human reasoning (specifically, Philip N. Johnson-Laird's mental models theory) and to use this theory as part of the machine's computational theory of mind, i.e., its ascriptive theory about the mental contents and operations of its audience. In operation the machine uses mental models theory to exploit cognitive illusions, highly fallible heuristics and biases in human reasoning the machine incorporates cognitive illusions into its arguments, which are thus called illusory arguments. Since the machine articulates illusory arguments with the intent to deceive its audience, the machine is a liar and a teller of sophistic lies.The dissertation lays out the philosophical, psychological, and computational foundations for this lying machine, and provides empirical evidence that the machine reliably deceives ordinary humans. It presents the philosophic rationale for the claim that the machine lies, or at least simulates lying, and includes description of sophistic lies and sufficient conditions for lying sophistically (the form of lying enacted by the machine). Additionally, the work contains a formal specification of a mental models theory for propositional reasoning, and theoretically and computationally extends this theory to the generation of credible arguments. Also described are two psychological experiments that test the illusory and persuasive effect of the machine's sophisms. The results of these experiments confirm that the machine is a potent liar, and validate the approach to mechanizing sophistic deception.The described research builds on and is contrasted with contemporary philosophic accounts of lying, cognitive and psychological theories of human reasoning, and the thematic and theoretic pursuit of persuasive technologies and computational models of natural argument. This work is inspired by and linked to philosophy of mind and language, artificial intelligence, computational cognitive modeling, argumentation theory, fallacy theory, and classical Greek rhetoric.