Melbourne School of Psychological Sciences - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemHow we evaluate and overrule our perceptual decisions( 2020)To navigate the world safely, it is critical that we are able to rapidly evaluate and overrule about our perceptual judgements. This thesis investigated the cognitive processes which underlie two forms of decision evaluation: discrete ‘changes of mind’ (i.e. decision reversals) and continuous confidence judgements. Three specific research questions were posed and addressed. Firstly, what sources of sensory information influence the likelihood and speed with which we change our minds about a perceptual decision? Secondly, what are the moment-to-moment information processing dynamics that underlie decision reversals? And finally, can information which is seemingly extraneous to a perceptual decision, specifically the amount of physical effort invested into reporting a decision outcome, affect retrospective judgements of decision confidence? These research questions formed the basis of three studies, which make up the core empirical chapters of this thesis. Study 1 investigated whether ‘absolute’ sensory information affects change-of-mind behaviour, across two experiments. In both experiments, participants indicated which of two flickering grey squares was the brightest with a button press. Following each initial decision, the stimuli remained on screen for a brief period and participants were free to change their response. To manipulate absolute sensory evidence the overall brightness of the two squares was varied, while either their luminance difference (Experiment 1) or luminance ratio (Experiment 2) was held constant. In both experiments increases in absolute evidence led to faster, less accurate initial responses and slower changes of mind. Change-of-mind accuracy decreased when the luminance difference was held constant, but remained unchanged when the luminance ratio was fixed. To account for these findings, we examined the predictions of six models: three existing change-of-mind models and three alternative models which have previously been used to account for the effects of absolute evidence on one-off decisions. Overall, a leaky competing accumulator model best accounted for participants’ behaviour. This suggests that the biologically relevant features of leak and partial inhibition within a decision process may be important in accounting for change-of-mind behaviour. Study 2 investigated the information processing dynamics underlying initial decisions and changes of mind. In particular, this study addressed the outstanding question of whether information processed prior to a decision being made (‘pre-decisional information’) has any influence on the likelihood and speed with which that decision is later reversed. As in Study 1, participants indicated which of two flickering grey squares was the brightest. Following each decision, the stimulus briefly remained on screen and participants were free to change their response. Critically, with each screen refresh a random luminance value was added to the mean luminance value of each square. Using psychophysical reverse correlation, we then retrospectively examined the impact that this luminance noise had on participants’ decisions on a frame-by-frame basis. Strikingly, we found that even the very first frame of sensory evidence participants saw influenced the likelihood and speed of later decision reversals. This indicates that pre-decisional information can influence later change-of-mind behaviour, and challenges the most prominent model of perceptual changes of mind, the extended Diffusion Decision Model (extended DDM), which predicts a complete insensitivity to pre-decisional information. To account for our findings within the DDM framework, we developed a novel variant of the extended DDM in which initial sensory information exerts a long-lasting bias over ongoing evidence accumulation. When fit to just the behavioural response data alone, this model was able to recreate the information usage patterns we observed. This suggests that an initial 'snapshot' of sensory information may exert a long-lasting bias over later sensory evidence accumulation, thus influencing later self-corrective behavior. Finally, Study 3 investigated the effect of foregone physical effort expenditure on decision confidence judgements. In this study, a dynamic luminance discrimination task was again employed. However, participants reported their decisions by squeezing one of two hand-held dynamometers until a pre-specified force threshold was reached. To manipulate the amount of effort required to report a choice, we varied how hard participants needed to squeeze on each trial across three individually calibrated levels (low, medium, high). After each decision, participants gave a confidence rating on a continuous scale ranging from 0 (‘certainly incorrect’) to 100 (‘certainly correct’). It was found that when more effort had been invested into reporting a decision, participants were more confident that the decision was correct. Broadly put, this suggests that people are sensitive to a ‘motoric sunk cost effect’ whereby greater foregone effort expenditure leads to an inflated sense of decision confidence. Overall, these findings suggest that: a) change-of-mind decisions are sensitive to absolute as well as relative sources of sensory information, b) that initial, pre-decisional, sensory information can influence the speed and likelihood with which a decision is later reversed, and c) that additional sources of information beyond sensory evidence, specifically action dynamics, can feed into and/or modulate the processes which underlie self-evaluative behaviour. These findings are consistent with post-decisional evaluative behaviours arising out of a continued unfolding of the initial decision process, with time-varying dynamics, which receives top-down modulation from additional self-monitoring process(es).