On the nature of certainty in value-based decisions: Disentangling variability- and extremity of experiences

Tan, Tiemen Sebastiaan 1 ; Raghoebar, Sanne1; Zhang, Chao2; Quandt, Julian3; Veling, Harm1

Published Mar 10, 2026 on Dryad. https://doi.org/10.5061/dryad.c59zw3rnh

Data files

Mar 10, 2026 version files 26.98 MB

certainty_in_value-based_decisions.csv

26.97 MB
README.md

5.84 KB

Abstract

Consistent value-based decision-making is crucial for outcomes such as maintaining a healthy lifestyle, or being perceived as trustworthy. Confidence in evaluations is related to consistency of value-based decisions, but its nature is poorly understood. Here we examined how evaluation confidence is formed through the possibly unique contributions of “variability”- and “value-extremity” of prior experiences with objects. We conducted a preregistered 2×2 within-subjects experiment (N = 188) where variability- and value-extremity of experiences with objects were manipulated independently. Participants learned value-distributions for objects, estimated their value, and rated confidence in their evaluation. As predicted, results show that variability of experiences is negatively related to confidence, independent of value-extremity. Surprisingly, we find no evidence for a positive effect of value-extremity on evaluation confidence. These results point to the possibility that the previously observed relation between value-extremity and evaluation confidence is the result of an ecological confound between variability and “value-extremity” of prior experiences.

Dataset DOI: 10.5061/dryad.c59zw3rnh

Description of the data and file structure

This data was collected to investigate the independent effects of variability of previous experiences and the extremity of previous experience of participant's certainty in their own value-estimations of stimuli.

Files and variables

File: certainty_in_value-based_decisions.csv

Description: The datafile contains information on the learning process of (novel) stimuli, value-estimations of stimuli, certainty ratings of value-estimations, and binary choices between stimuli.

participantNumber: Individual identifier for each participant.

date: Moment the experiment was started, in YYYY-MM-DD format with time of day added after an underscore ("_")

psychopyVersion: Software (PsychoPy) version used for the experiment

OS: Operating system the participant used.

valuePractice: (Numerical) values presented in the practice learning phase of the experiment. Values correspond to (euro)cents.

xMouseSunsetScaleEvaluation: The x position of the cursor at the moment participants estimate the value of an stimulus. Ranges from -1 (left) to +1 (right)

yMouseSunsetScaleEvaluation: The y position of the cursor at the moment participants estimate the value of an stimulus. Ranges from -1 (bottom) to +1 (top)

evaluationValue: Participant's estimated average value associated with an stimulus. On a scale of 0-200 (euro)cents.

reactionTimeEvaluation: Time difference between start and end of evaluation routine, in which participants estimate stimulus' average value. In seconds.

extremity: The average value of a stimulus, in cents.

fractal: Stimulus identifier per stimulus.

xMouseConfidence: Same as "xMouseEvaluationValue" but for confidence routine.

yMouseConfidence: Same as "yMouseEvaluationValue" but for confidence routine.

confidenceValue: Reported certainty in a value estimation ("evaluationValue") on the same row. On a scale ranging from "not at all confident" to "very confident", with a precision of 200 points.

reactionTimeConfidence: Time difference between start and end of certainty reporting routine. In seconds.

builtDistribution: list of 101 sublists. Each sublist represents a column. The farther right the sublist, the higher the values associated with it (0-200, steps of 2). So the first column represents value 0, second column values 1 & 2, third - 3 &4, ..., 101th - 199 & 200. The number in the sublist shows how many blocks (out of 100 (or 10 for the practice stimuli)) were placed in that column. This can be used, among others, to calculate & visualize spread of blocks and average values.

condition: List that tracks, per row, which of the six conditions (extremity x outcomeStability) was associated with that row. First index is the extremity, second index is the outcome stability.

outcomeStability: Width (standard deviation) of the underlying probability distribution values were drawn from in the learning phase. This is either 4 or 12.

shownValue: List of values shown in the learning phase to a participant. Values are numerical and represent monetary value (cents).

randomNumber: Random number shown to be reported (0-200) in between evaluationValue trials.

randomNumberEvaluation: Particpant's reported value on randomNumber trials. On a scale of 0-200.

differenceRNEvaluation: Difference between shown random number and reported evaluation on randomNumber Trials.

chosenFractal: Chosen stimulus in the binary choice task, per trial.

notChosenFractal: Not chosen stimulus in the binary choice task, per trial.

chosenSide: Side of the screen the chosen stimulus was on. (Left / Right)

reactionTimeBinaryChoice: Difference between start of binary choice routine and the choice being made. In seconds.

meanChosenFractal: average value of the selected stimulus (same as value in "extremity" for a stimulus).

SDChosenFractal: Standard deviation associated with a stimulus (same as value in "outcomeStability" for a stimulus).

meanNotChosenFractal: same as meanChosenFractal but for the not chosen stimulus.

SDNotChosenFractal: same as SDChosenFractal but for the not chosen stimulus.

correctChoice: Whether the stimulus with the higher average value was chosen (yes / no). For ties (equal average value), says "equal".

reactionTimeChoiceConfidence: Time difference between start and end of binary choice confidence routine. In seconds.

fractalLeft: Stimulus shown on left hand side of the screen. Per binary choice trial.

fractalRight: Stimulus shown on right hand side of the screen. Per binary choice trial.

xMouseBinaryChoice: Same as xMouseConfidence but for confidence in the previous binary choice instead for the previous value estimation.

yMouseBinaryChoice: Same as yMouseConfidence but for confidence in the previous binary choice instead for the previous value estimation.

confidenceBinaryChoice: same as confidenceValue, but for previous binary choice made.

bonus: Earned bonus, in euro's.

Code/software

The files are .csv.

They can be read by, for example, RStudio, Python or LibreOffice.

Access information

Other publicly accessible locations of the data:

https://osf.io/khc7u/files/osfstorage

Human subjects data

Consent was obtained to make the de-identified data available for future research. We de-identified data by assigning a random participant number to each participant, using participant numbers to distinguish individual participants in our data analyses.