Evidence-gathering without navel-gazing

Abstract What explains evidence-gathering? Some say desire for knowledge, or desire for true belief, or desire for belief that has some other feature. These answers are alike insofar as each ascribes to the agent a propositional attitude (desire) about her own propositional attitudes (knowledge or belief). Others say that what explains evidence-gathering is desire for a non-attitudinal outcome, such as obtaining food. That is the gist of an equation at the heart of I. J. Good’s Value of Information Theorem. But Good’s proof assumes that the agent believes that she will choose a future act that maximizes her future subjective expected utility; which suggests that it, too, ascribes to the agent a propositional attitude (belief) about her own propositional attitudes (future desires and beliefs). This paper seeks to explain evidence-gathering without ascribing to the agent any propositional attitudes about her own propositional attitudes. This is necessary, if we are ever to explain the evidence-gathering behaviour of non-self-aware animals. For some cases we can stay close to Good, although taking care to sidestep any ascriptions of second-order attitudes. For other cases, however—cases in which individual cognition does less work and evolved dispositions do more—we must diverge further from Good; and for some of them we need a new type of explanation, according to which acts of evidence-gathering are neither desired as means nor desired as ends, but desired nonetheless.