Outcome uncertainty influences probability perception and risk attitudes

Elston, Thomas W; Mackenzie, Ian Grant; Mittelstädt, Victor

doi:10.1098/rsos.210307

Cited by 3 publications

(14 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, while information format did not itself drive the perceptual inflation of low probabilities, our results suggest that it does modulate probability perception. This is consistent with our prior study of the DE gap (Elston et al, 2021) that compared choice patterns across risky experience-based and description-based options and did not include a safe option. Our prior study indicated that subjective outcome likelihoods associated with experience-based options were perceptually inflated across all probabilities tested (20%, 50%, 80%).…”

Section: Information Format the Description-experience (De) Gap And T...supporting

confidence: 89%

“…All candidate emojis were simple in color and form and did not contain faces or face-adjacent information (c.f. Elston et al, 2021). The probabilities of 20%, 50%, and 80% were randomly assigned to the emojis for each subject.…”

Section: Apparatus and Stimulimentioning

confidence: 99%

“…Furthermore, as per our pre-registrations, we excluded individual trials with outlier choice reaction times (RTs). Based on our previous study (Elston et al, 2021) and manual inspection of the data, we defined outlier RTs as those less than 200 ms and greater than 1500 ms. These excluded individual trials were rare and never comprised more than 3% of an individual's responses.…”

Section: Exclusion Criterion and Data Preparationmentioning

confidence: 99%

“…We used RL models to provide trial-wise estimates of subjective value associated with each stimulus. As pre-registered, we fit and compared 2 RL models: (1) the standard temporaldifference Q-learning algorithm (TDLR; Sutton & Barto, 2018) and ( 2) the RS-RL model (Elston et al, 2021;Niv et al, 2012). RS-RL differs from traditional temporal-difference learning in that it fits two learning rates, one for positive prediction errors (α + ) and one for negative prediction errors (α -), rather than a single learning rate used for all prediction errors.…”

Section: Reinforcement Learning Model Of Risk Sensitivitymentioning

confidence: 99%

“…Thus, elongated decision times reflect elongated cognitive processing and, in turn, cognitive resources. One interpretation related to outcome uncertainty is that outcomes associated with risky options are internally represented as a distribution and that reaction times partially reflect the process of integrating over that distribution (see Elston et al 2021). Thus, because risky options are inherently more uncertain than safe ones, it may be that the process of computing values associated with risky options takes longer because the distribution to integrate over is wider.…”

Section: Risky Decisions and Loss-mitigating Decisions May Require Mo...mentioning

confidence: 99%

See 4 more Smart Citations

Extreme outcomes accentuate overweighting of low probabilities and underweighting of high probabilities in experience-based choice

Elston¹,

Leuthold²,

Mackenzie³

et al. 2023

Preprint

View full text Add to dashboard Cite

Risk permeates decision making because choice outcomes are often uncertain. When basing decisions on prior experience, previous research suggests that the prospects of gain and loss differentially produce subjective inflation and deflation of risk-associated values, respectively. These effects have been explained in terms of an “extreme-outcome rule”, which posits that valuations of risky options are biased towards the best/worst possible outcome. However, prior studies typically examined risky choices only at the 50% level, and it was unclear whether the risk-preferences predicted by the “extreme outcome rule” would generalize to other probabilities. Furthermore, the cognitive processes underlying risk preferences remained unclear. We addressed these questions through three pre-registered experiments where the critical condition involved decisions between equivaluable safe and risky options across three probabilities (20%, 50%, 80%) with the goals of either maximizing gains or minimizing losses. We found that probability and risk context mutually influenced choice-behavior: low probabilities were particularly subjectively inflated, leading to a strong preference for rare, risky options when seeking gains and a strong aversion when mitigating losses. We also found that decisions involving risk require more cognitive resources, as indicated by longer decision times. Additionally, a risk-sensitive reinforcement model showed that risk-seeking/averse individuals tended to overweight positive/negative outcomes. This study highlights that outcome uncertainty, the range of possible outcomes a given choice option could yield, is a critical component in probability distortion and that subjective probability distortions may drive risk attitudes.

show abstract

Section: Information Format the Description-experience (De) Gap And T...supporting

confidence: 89%

Section: Apparatus and Stimulimentioning

confidence: 99%

Section: Exclusion Criterion and Data Preparationmentioning

confidence: 99%

Section: Reinforcement Learning Model Of Risk Sensitivitymentioning

confidence: 99%

Section: Risky Decisions and Loss-mitigating Decisions May Require Mo...mentioning

confidence: 99%

See 3 more Smart Citations

Extreme outcomes accentuate overweighting of low probabilities and underweighting of high probabilities in experience-based choice

Elston¹,

Leuthold²,

Mackenzie³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

Extreme Outcomes Accentuate Overweighting of Low Probabilities and Underweighting of High Probabilities in Experience‐Based Choice

Elston,

Leuthold,

Mackenzie

et al. 2024

Behavioral Decision Making

Self Cite

View full text Add to dashboard Cite

Subjective inferences of probability are critical for decisions involving uncertainty. Recent studies have suggested that extreme outcomes bias beliefs about the value of uncertain options toward the best/worst outcome possible when learning the odds through experience, leading to increased preferences for uncertain options over equivaluable sure bets when there is the prospect of gain and, conversely, aversion to uncertain options when there is the prospect of loss. However, prior studies regarding the influence of extreme outcomes on decisions involving uncertainty have only done so using 50/50 gambles, and it was unclear whether extreme outcomes biased probability perception more broadly. Across three pre‐registered experiments, we found that when people made decisions between equivaluable certain and uncertain options, they particularly preferred uncertain options at low probabilities (20%) when there was the prospect of gain and avoided them when there was the prospect of loss, with these preferences being reduced or even reversed at medium (50%) and high (80%) probabilities. We also found that uncertainty preferences were influenced by outcome extremity and the relative certainty associated with safe reference options. We conclude that extreme outcomes accentuate the overweighting of low probabilities and the underweighting of high probabilities in experience‐based choice.

show abstract

The influence of reward and loss outcomes after free- and forced-tasks on voluntary task choice

Mittelstädt,

Mackenzie,

Leuthold

2024

Psychological Research

View full text Add to dashboard Cite

In four experiments, we investigated the impact of outcomes and processing mode (free versus forced) on subsequent voluntary task-switching behavior. Participants freely chose between two tasks or were forced to perform one, and the feedback they received randomly varied after correct performance (reward or no-reward; loss or no-loss). In general, we reasoned that the most recently applied task goal is usually the most valued one, leading people to prefer task repetitions over switches. However, the task values might be additionally biased by previous outcomes and the previous processing mode. Indeed, negatively reinforcing tasks with no-reward or losses generally resulted in more subsequent switches. Additionally, participants demonstrated a stronger attachment to free- compared to forced-tasks, as indicated by more switches when the previous task was forced, suggesting that people generally value free over forced-choice task goals. Moreover, the reward manipulation had a greater influence on switching behavior following free- compared to forced-tasks in Exp. 1 and Exp. 3, suggesting a stronger emphasis on evaluating rewarding outcomes associated with free-task choices. However, this inflationary effect on task choice seemed to be limited to reward and situations where task choice and performance more strongly overlap. Specifically, there was no evidence that switching behavior was differentially influenced after free-and forced-task as a function of losses (Exp. 2) or reward when task choice and task performance were separated (Exp. 4). Overall, the results provide new insights into how the valuation of task goals based on choice freedom and outcome feedback can influence voluntary task choices.

show abstract

Outcome uncertainty influences probability perception and risk attitudes

Cited by 3 publications

References 25 publications

Extreme outcomes accentuate overweighting of low probabilities and underweighting of high probabilities in experience-based choice

Extreme outcomes accentuate overweighting of low probabilities and underweighting of high probabilities in experience-based choice

Extreme Outcomes Accentuate Overweighting of Low Probabilities and Underweighting of High Probabilities in Experience‐Based Choice

The influence of reward and loss outcomes after free- and forced-tasks on voluntary task choice

Contact Info

Product

Resources

About