Search Results

This chapter examines a recent model of the basis and accuracy of confidence judgements (Koriat 2012), focusing on the metatheoretical assumptions underlying the model. According to the self-consistency model (SCM), when participants must choose between two alternative answers to a question, they retrieve a sample of representations of the question and base their confidence on the consistency with which the chosen answer is supported across representations. Self-consistency represents a crude mnemonic cue that reflects the amount of deliberation and conflict experienced in making a choice. Although information is retrieved from within, the process is assumed to have much in common with the sampling of observations from the outside world to test a hypothesis about a population and to assess the likelihood that the conclusion reached is correct. Thus, subjective confidence is modelled by the logic underlying the calculation of statistical level of confidence, and represents the assessed likelihood that a new sample of representations will yield the same choice. Several predictions regarding the basis of confidence judgements were confirmed for general-information tasks and perceptual judgements, and for social beliefs and social attitudes. The confidence — accuracy correlation was shown to be a by-product of the consistency — correctness relationship: it is positive only when the consistently selected answer is correct but is negative when the consistently selected answer is wrong. The results support the idea that metaknowledge is intimately tied to knowledge.

Avians have been shown to be highly cognitive. This chapter asks whether they are also metacognitive in two separate experimental situations. In Study 1, recognition of confidence of their immediate decision in perceptual tasks was tested. Pigeons and bantams performed a visual search to peck at a target colour among distracters. Completing the search led them to a choice between a ‘risk’ icon and a ‘safe’ icon. The ‘risk’ icon always rewarded the bird’s correct search responses by food but punished incorrect responses by a timeout. The ‘safe’ icon rewarded the response at a lower frequency irrespective of their responses. The birds chose the ‘safe’ icon more frequently after an incorrect search than after a correct search. This differential use of icons generalized to the same tasks with new stimuli in most birds, and in one pigeon to a novel line-length discrimination task. In Study 2, it was asked whether pigeons would seek a hint when they were unsure of what to do next. They performed a list learning task, in which they had to peck at three icons in an arbitrary sequence. When novel icons appeared, the birds had to find out a correct sequence by trial and error. However, on half of the trials, there was an additional ‘hint’ icon, and pecking at it marked the next icon in the sequence. Some of the birds used this ‘hint’ icon more frequently upon starting to learn novel lists and less frequently later. These results suggest that some birds could be metacognitive in some situations, although other non-metacognitive accounts were also considered.

The chapter is an extension of Hampton's approach of asking a monkey to respond metacognitively. It focuses on confidence judgments, which in humans are typically made verbally. It describes an experiment on rhesus macaque monkeys. These monkeys were given the opportunity to express their confidence by placing bets on the accuracy of their tasks in a cognitive task. The subjects were required to respond on all trials, easy and hard. After each trial, subjects were required to select a high- or a low-confidence icon. Having shown that the subjects chose the high- and low-confidence icons appropriately, this chapter argues that the metacognitive ability of monkeys is similar to that observed in human subjects in experiments that use the confidence judgment paradigm. Judgment of learning asks how certain they are that they will be able to remember a recently learned item in the future.