Kopsacheilis, Orestis
(2019)
The description-experience gap in individual and social decisions under risk and uncertainty.
PhD thesis, University of Nottingham.
Abstract
This thesis contributes to the understanding of the ‘Description - Experience (DE) gap’, which posits that risky decisions depend on the way information about uncertainty is communicated: from description or from experience. The canonical interpretation of findings in this literature suggests that when people make decisions from description, they behave as if overweighting rare
events relative to their probability, whereas, when they make decisions from experiential formats, they behave as if underweighting them.
Chapter 1 provides an overview of the main topics and research methodologies presented in Chapters 2, 3 and 4.
Chapter 2 reports the results of an experimental study that provides a cohesive account of the forces behind the DE gap. Our experimental protocol allows us to quantify the effect of each factor in isolation. Moreover, to address methodological concerns in this literature, we employ an elicitation method which allows us to measure these effects both with and without the mediation of a behavioural model that accounts for probability weighting. We find an overall significant DE gap which is equal in size with the literature’s average.
Despite being mostly driven by an informational asymmetry (sampling bias), other factors pertaining to preferential (ambiguity) and cognitive (likelihood representation and memory) aspects of decision making proved important.
Examining the shape and relative position of probability weighting curves we discuss intriguing behavioural implications of our findings.
Chapter 3 focuses on how people search for information in decisions from experience and examines how different search patterns influence ensuing risky choices. In a lab-experiment we find that people search more from options with rarer events. We also find that sampling amount decreases over time periods. Both of these findings become less salient however after the introduction of a history table which records and displays previously sampled outcomes during the lottery evaluation. The cue that dominates the treatment
where such a table was present, is the table’s maximum capacity. With respect to choices, we elicit and compare probability weighting functions from treatments
where decisions were made from experience with a treatment where decisions were made from description. Our treatment comparison reveals evidence for a variant of the canonical interpretation of the gap. We refer to it as the ‘relative underweighting hypothesis’, which states that rare events in individual risky decisions are overweighted in Experience too, but less so than
in Description.
Chapter 4 explores the DE gap in a social context, where we investigate whether the format in which social information is obtained -descriptive or
experiential- influences cooperation in social dilemmas and if so, how. We develop and implement in an online experiment, a variation of the prisoner’s dilemma game that allows us to observe cooperative responses over a range of likelihoods of cooperation. The likelihoods are communicated either in descriptive or experiential formats. We find that conditional cooperation - the willingness to cooperate if others do the same - is prevalent in our study as cooperation rates increase with the probability of cooperation across treatments.
Nonetheless, there are significant differences in the cooperation patterns between Description and Experience. Interestingly, we find evidence that this gap in social decision making, is in the opposite direction from what
the canonical interpretation in the individual context would have predicted.
Rare events (of cooperation or defection) appear to be more overweighted in Experience rather than in Description. Another asymmetry with the individual
domain is that sampling bias, the predominant driver of the DE gap in risky choices, does not affect the gap in the social domain. We conclude that this reversal is due to people being less sensitive towards social information, when they receive this information experientially compared to receiving it descriptively.
Lastly, we discuss why such reversals are less likely to occur in individual decision settings.
Chapter 5 provides a summary of the previous chapters’ results, identifies their limitations and discusses potentially interesting directions for future research.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
