The key difficulty in making decisions is the uncertainty about the outcomes
that could result from any particular decision. In classical decision making under risk the
uncertainty is explicitly given as probabilities: a 50% chance of $100 or else nothing. In a
wide range of biologically important tasks, though, the organism must estimate the
uncertainties associated with outcomes. The sources of uncertainty can be perceptual,
motor, or environmental. Estimation can be based on simple frequency counts or explicit
models – as when we assume that a coin is fair despite never having tossed it.
Surprisingly, human performance in such tasks is often close to optimal
(Trommershäuser et al, 2008; Warren et al, 2012) despite the added burden. I will
describe a series of studies that let us examine how we estimate uncertainty in a variety
of perceptual, motor and cognitive tasks.
There is considerable evidence that humans systematically distort estimates of
uncertainty in making decisions and any complete description of how we estimate
uncertainty must explain such errors. I’ll present a model that captures how we distort
frequency in decision tasks and experimental evidence testing it.