Purchasing Information to Reduce Uncertainty in Trajectory Prediction

Abstract

Previous studies show people underestimate the amount of uncertainty in predictive states. This experiment tested an implication of such findings: that people will be less likely (than optimal) to acknowledge their uncertainty, and hence less likely to purchase information that reduces future uncertainty. In a trajectory prediction task, participants had the opportunity to reduce uncertainty by a constant amount by purchasing either an expensive (bad deal), neutral-cost, or cheap (bargain) sensor. Participants were biased, relative to expected value theory, in integrating cost with uncertainty reduction in their decisions to purchase. In contrast to our primary hypothesis that they would place more decision weighting on reducing the cost of the purchase, than on gaining information, our results revealed the opposite. The findings are interpreted in the context of models of attribute salience in decision-making.

Paying a price for information: Non-optimal decision making and preference for information over expected value outcomes

“…The neutral condition purchase rate of 65% approximately replicates the 60% value obtained by Wickens et al, (2016) but in the previous case without the current direct financial implications. This 65% value again reveals the information seeking bias, supporting our first hypothesis.…”

“…First, as noted, there were financial rewards for engaged participation, which would induce greater purchasing with better deals; and such behavior was clearly observed in the data. Second, such behavior, and the approximate 15% information purchase bias, replicated that observed within the experimental laboratory where on line observation of participants was available (Wickens et al, 2016). As a consequence, we have considerable faith in the reliability of our data, and resulting conclusions.…”

“…We note that Wickens et al (2016) did examine information purchase tendencies in a paradigm similar to that employed here and found such a bias to overpurchase. However, these results were less than conclusive because (a) exhibiting this bias did not penalize participants (the more-information purchase option had the same notional NEV as less information option) and (b) neither actual financial rewards, nor penalties were imposed.…”

“…The paradigm is a variant of one that we have employed frequently in past research (Herdener et al, 2016(Herdener et al, , 2019Pugh et al, 2018) which involves predicting the location of a trajectory endpoint, such as a hurricane landfall or location of a downed aircraft, when there is expected variability or uncertainty in that trajectory. In these past studies there is a strong trend that people underestimate the variability of a set of such trajectories, as if overconfident in their ability to predict the location of the endpoint.…”

“…The current experiment tests whether a preference for greater information reflects a bias that will persist even when the choice to purchase reflects a loss in NEV. That is, the cost of the additional information is clearly more than the financial gain expected from its purchase, as quantified by the increased likelihood of finding the valued target. We also examined purchase preference when that purchase is neutral (as in Wickens et al, 2016), and when purchase is associated with increased expected value.…”