Reasoning on conflicting information: An empirical study of Formal Argumentation.

Abstract

According to the Argumentative Theory, human reasoning has an argumentative function, which consists of devising and evaluating arguments for and against various claims. It is however unclear how humans handle conflicting claims they face in everyday life (i.e., "Bob is telling me that Alice is at the library" vs. "Charles is telling me that Alice is at home"). We here investigate human argumentative reasoning in the light of Formal Argumentation, a research field that develops formal methods to give a normative account of argumentation and reasoning about conflicting information. In Formal Argumentation, multiple argumentation semantics that allow selecting sets of jointly acceptable arguments have been proposed. Nonetheless, it is unclear which of these semantics predicts best how humans evaluate the acceptability of conflicting arguments. We conducted an empirical study in which 130 young adults judged natural language arguments. We instructed them to draw the attack relation between the given arguments and to evaluate the acceptability of each of these arguments. Our results show that human judgments on the existence and directionality of attacks between the arguments conform to theoretical predictions from Formal Argumentation. We further found out that some less well-known argumentation semantics predicted human evaluation better than the most well-known semantics. These findings support the cognitive plausibility of variants of Formal Argumentation and bring new insights into reasoning about conflicting information.