Adapting Information Search through Subjective Confidence and Accumulated Evidence
Mon-B16-Talk I-03
Presented by: Linda McCaughey
Decisions are almost always preceded by information search aimed at finding the option that is most preferable. How much information one acquires before making a decision should depend on how sure one wants to be. This aspired confidence level presumably plays an important role in regulating pre-decisional search.
To investigate this role we conducted two experiments (combined N=168), in which participants made binary decisions based on self-truncated sampled information and reported their confidence. The first experiment manipulated general confidence levels while the second manipulated decision importance.
Participants in the first experiment underwent a trial phase in which one condition received negatively distorted decision feedback while the control received accurate feedback before completing the remainder of the decisions without any additional feedback.
As expected, participants in the distorted-feedback condition collected significantly more information, on average. Nevertheless, they also reported significantly lower confidence levels, but tended to truncate their decision after a very similar amount of evidence. A regression revealed that higher evidence was positively related with confidence for both conditions, but the distorted-feedback condition needed more evidence to reach the same confidence level as the control condition.
The second experiment manipulated decision importance within participants. While participants collected significantly more information for more important decisions, this did not result in higher evidence or higher reported confidence levels. Hence, situational influences may influence how much evidence we need to reach an aspired confidence level, but sample size seems to have an additional effect, which might be maladaptive.
To investigate this role we conducted two experiments (combined N=168), in which participants made binary decisions based on self-truncated sampled information and reported their confidence. The first experiment manipulated general confidence levels while the second manipulated decision importance.
Participants in the first experiment underwent a trial phase in which one condition received negatively distorted decision feedback while the control received accurate feedback before completing the remainder of the decisions without any additional feedback.
As expected, participants in the distorted-feedback condition collected significantly more information, on average. Nevertheless, they also reported significantly lower confidence levels, but tended to truncate their decision after a very similar amount of evidence. A regression revealed that higher evidence was positively related with confidence for both conditions, but the distorted-feedback condition needed more evidence to reach the same confidence level as the control condition.
The second experiment manipulated decision importance within participants. While participants collected significantly more information for more important decisions, this did not result in higher evidence or higher reported confidence levels. Hence, situational influences may influence how much evidence we need to reach an aspired confidence level, but sample size seems to have an additional effect, which might be maladaptive.
Keywords: sampling, information search, confidence, meta-cognition, self-regulation