Subjective Confidence Predicts Information Seeking in Decision Making – new paper out in Psychological Science

Awesome news: Our new paper with Annika Bolt and Nick Yeung is published in Psychological Science! You can find the paper here. The work done in this paper was performed back in 2015, when I spent three extremely fruitful months in the ACC lab at Oxford University.

Our main goal was to demonstrate that adaptive actions (such as seeking more information before making a decision) are driven by our confidence in an impending decision. To do so, we relied on a dissociation between accuracy and confidence that Annika unraveled during her PhD work. Doing so, we were able to demonstrate that participants solicited more additional evidence before committing to a decision in a condition associated with low versus high confidence, despite matched accuracies. This finding has important implications for current theories of decision making and confidence, most of which predict a tight link between evidence, accuracy and confidence. A particularly interesting aspect about this paper is that the journal editor requested that we perform a high-powered pre-registered replication as a pre-condition on the manuscript being sent out to review. Our replication showed that our main findings were highly robust, but we failed to replicate an exploratory correlation with meta-dā€™ that was highly significant in the original study. From this, we learnt that the need for caution and replication applies not only to testing primary hypotheses, but equally to analyses that are conducted in the spirit of fully interrogating the data after primary hypotheses are confirmed. A critical aspect of this story was the importance of journal policy ā€“ without the editorial request, the spurious correlation would have entered the literature qualifying as strikingly large and, more broadly, we would not have been forced to learn a lesson that we feel was very valuable to us as scientists.

Currently, we are writing up the results of an EEG version of this task, which is looking very good! Stay tuned!