The Psychology of Conditionals and Non-Constructive Inference

SPEAKER

David Over

ABSTRACT

We will review psychological results showing that people judge the
probability of the indicative conditional in natural language, P(if p
then q), to be the conditional probability, P(q|p). Some psychologists
have consequently developed psychological accounts of conditional
reasoning in which P(if p then q) = P(q|p), making use of the normative
proposals of Ramsey (1931), de Finetti (1937), and Adams (1998). It is
invalid in these accounts to infer "if not-p then q" from "p or q", and
yet intuitively this inference often appears to be a very strong one. It
is also endorsed by people in some psychological experiments. We will
argue that the inference is strong when the disjunction is supported by
a non-constructive justification, but not when it is justified
constructively. In the former type of justification, our analysis
implies that P(if not-p then q) will be close to P(p or q). We will also
present experimental evidence that confirms our analysis.