Confidence and Objective Probability Signaling in Perceptual Systems
- Author(s): Kowalsky, William
- Advisor(s): Burge, Tyler
- et al.
It's familiar to us from our thinking and deliberating that we can more or less confident about the truth of a proposition. For instance, you might be more confident that it will be overcast tomorrow than you are that it will rain. States of confidence are most familiar to us at the level of our thought. Are there states of confidence below the level of thought? Are states of confidence assigned by our psychological sub-systems? Recent discussions in the philosophy of perception and philosophy of psychology raise the question of whether states of confidence are assigned in perceptual systems. A perceptual system can represent an object as red, but can it also assign a level of confidence to the object’s being red?
This dissertation argues that perceptual systems do assign levels of confidence. To address the question, we must have some sense for what sort of state a level of confidence. We must explicate the notion of a level of confidence, in order to identify central psychological signatures and roles that such a state plays. Such signatures clarify what it would take for a perceptual state to count as a level of confidence. Having explicated the notion, we must then turn to empirical science. We must look carefully at perceptual psychology and allied fields to see whether perceptual systems in fact have states that satisfy the signatures. I follow this methodology.
In explicating levels of confidence, I draw on both commonsense understanding and normative disciplines that invoke levels of confidence, such as decision theory and formal epistemology. I identify two core signatures of levels of confidence. One signature is roughly that increasing levels of confidence in p tend to lead to an increased reliance, by subsequent psychological processes, on the truth of p. The other signature is that increasing levels of confidence in p tend to be formed on the basis of information that better supports the truth of p.
I also argue, on empirical grounds, that perceptual systems have sensory capacities for signaling objective probabilities. Most saliently, they have capacities for tracking the objective probability that the perceptual representations they produce are veridical. A perceptual representation as of red may be objectively more likely in one circumstance than in another to be veridical. Perceptual systems have capacities for signaling this probabilistic difference specifically. I argue for such states by considering experimental results in multimodal cue integration. I embed my account in broader accounts of objective probability and sensory signaling.
I then argue that the probability signaling states constitute levels of confidence in perceptual systems. The probability signaling states guide the use of perceptual representations in a way that satisfies the first signature. Because they track different probabilities of a representation's veridicality, they satisfy the second signature. I close by considering what other sorts of sensory phenomena might be explained in terms of confidence, by examining capacities for ``approximate number representation."