Metacognition is our own subjective evaluation of memory quality. For example, we can judge how confident we are in a memory (i.e., there are some that we are more confident in than others), or how well we learned something (i.e., the likelihood of retrieving it later on). The interesting thing about metacognition is that our subjective evaluation doesn’t always match the objective memory quality (whether the memory is accurate or not)!
Conscious Experience During VWM Encoding
We are used to rich experiences that disappear immediately upon the loss of sensory information. We are then left to our own devices to piece together the once vibrant scene from whatever information we encoded. Generally, the set of things we encoded are accessible, while the remainder is not. But what drives this accessibility? Is it related to our subjective experience at the time? Counterintuitively, this is not always the case.
In an experiment, we attempt to replicate this strange link between experience and accessibility. Further, we also explore the roles of attention in both the spatial (where we are attending) and temporal (at what time scales) domains. From here we can build models that attempt to predict what is accessible to us from moment-to-moment.
As we go about our lives, we like to think that our memories have fidelity – that is, the representation we have in our minds is accurate to how it was first encoded. In fact, we are often quite confident in the fidelity of certain memories, sometimes to our own detriment. Think of an event that both you and your friend witnessed yet you are both confident in differing accounts, or witnesses of crime who are asked to select a criminal from a photograph and who are very sure that they have selected the right person. Just because we are confident about a memory doesn’t always mean that memory is accurate. Your friend may have remembered something incorrectly, or the wrong person may be convicted of a crime. In fact, previous research has shown that our confidence is often over-inflated as compared to our actual memory accuracy.
We attempt to ‘fix’ the discrepancy between memory confidence and accuracy by examining the short-term metacognitive accuracy of individuals. Using a whole-report change detection task with feedback, we train participants to make more highly-confident-and-correct responses and reduce their highly-confident-and-incorrect responses. We reward participants with more points when they make a high-confidence response that is also correct (i.e., select the correct colour or line orientation) but subtract points when they make high-confidence responses that are incorrect. With their goal as gaining the most points as possible, participants are able to ‘calibrate’ their responses to be highly-confident-and-correct, leading to memories that are not only accurate but supported by a high confidence judgement (high metacognitive accuracy). Older adults may not adapt to the same training in the way that younger adults do for a variety of reasons, which is why we extend this study from younger adults to an older population.