Thesis defense Andrea Bertana (Donders series 489)
22 February 2021
Promotor: Prof. dr. A.J.van Opstal
Co-promotor: Dr. J.F.M. Jehee
Computational basis of human confidence in vision
In the current thesis, we investigated the computational basis of human confidence estimates during perceptual decisions. A central point of our investigation was the Bayesian theory of confidence, in which confidence is linked to the quality of the perceptual evidence underlying the decision. However, we considered also an alternative hypothesis, the heuristic-based strategy of confidence: observers could use certain image properties, such as image contrast or variability, as external cues to confidence. By creating new qualitative signatures of those models and focusing on random fluctuations in behavior due to internal noise, we showed that confidence estimates can be linked to the uncertainty in perceptual decisions. However, in some cases confidence can be based on another quantity: the perceived variability in the stimulus. Moreover, we showed that random fluctuations in behavioral reports, linked to confidence, are not caused by temporal variability in sampling strategies.