Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 47 of the 49 references it cites were found in a registry or an index; 19 of the 43 sources its claims cite were checked for support.

Crowdsourcing punishment: Individuals reference group preferences to inform their own punitive decisions

Open access
Son, Jae-Young, Bhandari, Apoorva, FeldmanHall, Oriel · 2019 · Scientific Reports

Abstract

Justice systems delegate punishment decisions to groups in the belief that the aggregation of individuals’ preferences facilitates judiciousness. However, group dynamics may also lead individuals to relinquish moral responsibility by conforming to the majority’s preference for punishment. Across five experiments (N = 399), we find Victims and Jurors tasked with restoring justice become increasingly punitive (by as much as 40%) as groups express a desire to punish, with every additional punisher augmenting an individual’s punishment rates. This influence is so potent that knowing about a past group’s preference continues swaying decisions even when they cannot affect present outcomes. Using computational models of decision-making, we test long-standing theories of how groups influence choice. We find groups induce conformity by making individuals less cautious and more impulsive, and by amplifying the value of punishment. However, compared to Victims, Jurors are more sensitive to moral violation severity and less readily swayed by the group. Conformity to a group’s punitive preference also extends to weightier moral violations such as assault and theft. Our results demonstrate that groups can powerfully shift an individual’s punitive preference across a variety of contexts, while additionally revealing the cognitive mechanisms by which social influence alters moral values.

Overview

CoChat Summary

TL;DR

Across five experiments, individuals’ punishment decisions were strongly swayed by group punitive preferences, with predicted punishment shifting by as much as 40 percentage points. Computational modeling showed this conformity arose from lowered decision thresholds and increased valuation of punishment, with differences between victims and jurors.

Key findings

  • Group punitive preferences increased punishment endorsement in every experiment: predicted punishment shifted by as much as 30 percentage points in Experiment 1, 40 percentage points in Experiment 3, nearly 30 percentage points in Experiment 4, and also modulated crime-severity judgments in Experiment 5.
  • Conformity scaled with the proportion of punishers in the group, and even a past group’s preferences influenced punishment when subjects decided first (Experiment 2), though this effect was attenuated relative to deciding last.
  • In Experiments 3–4, group majority lowered decision thresholds (all posterior Ps < 0.001) and increased drift rate as the proportion of punishers rose, indicating greater valuation of punishment. Victims showed an overall bias toward compensation (average z = 0.42, 0.43, 0.44 for Mildly, Somewhat, Highly Unfair; all posterior Ps < 0.001), while Jurors were biased toward compensation only for Mildly Unfair offers (average z = 0.45, posterior P < 0.001).
  • Victim versus Juror differences were mixed: Experiment 4 found conformity significantly weakened for Jurors relative to Victims (β = 0.50, SE = 0.13, z = −2.58, p = 0.010), whereas Experiment 5 found the overall conformity effect significantly stronger for Jurors than Victims (β = 2.06, SE = 0.69, z = 3.00, df = 224.00, p = 0.003).

Methods

  • Experiments 1–4 recruited 176 subjects from Brown University and the surrounding community. Final analyzed samples were 41 (Experiment 1), 48 (Experiment 2), 41 with one partial dataset (Experiment 3), and 43 (Experiment 4). Subjects completed a Justice Game in which Player B could Accept, Compensate, or Reverse unfair splits from Player A; Experiments 1–2 varied whether subjects decided last or first in a five-person group, and Experiments 3–4 used Victim and Juror versions suitable for drift diffusion modeling.
  • Experiment 5 recruited 300 Amazon Mechanical Turk subjects, with final samples of 118 Victim-condition and 108 Juror-condition subjects after exclusions. Subjects read assault or theft vignettes crossed with high- versus low-intensity crime and rated punishment severity on a 100-point scale.
  • Main analyses used mixed-effects logistic or linear regressions. Experiments 3–4 used hierarchical Bayesian estimation of the Drift Diffusion Model via HDDM, with model selection by Deviance Information Criterion, posterior predictive checks, and parameter recovery simulations.

Limitations

  • The paper states that future work should examine repeated interactions with the same perpetrators, especially whether conformity ceases once a perpetrator is perceived as rehabilitated, and how social influence and attitude change operate as learning processes.
  • The paper states that it is difficult to obtain field data quantifying how much conformity effects shift real jurors’ punitive inclinations, and notes that Experiment 5’s divergence from Experiments 3–4 may be due to hypothetical choices or to the vignettes involving more severe moral transgressions.
  • Evident from the text: Experiments 1–4 relied on relatively small lab samples and computer-generated group responses under social deception, while Experiment 5 used hypothetical vignettes and excluded many MTurk responses for data-quality reasons; these features may limit generalizability to real-world punishment contexts.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41598-019-48050-2

Cites

47 of 47 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Task and trial structure for Experiment 1 (Last Decider) and 2 (First Decider). (a) In the modified Justice Game, Player A is endowed with $10 at the beginning of each trial. Player A offers some portion of that money with Player B, who can choose to Accept, Compensate, or Reverse the offer. In our task, Player As always made unfair offers. Because Player Bs could maximize their payout by choosing either Compensate or Reverse, decisions to Reverse can be interpreted as cost-free punishment for Player As. (b) Subjects in the Last Decider experiment observed their group’s punishment preferences prior to indicating their own punishment preference. (c) Subjects in the First Decider experiment indicated their punishment preference prior to observing their group’s preferences. Although group members are represented by icons in this figure, subjects were shown photographs in the actual experiment.
Figure 1 Task and trial structure for Experiment 1 (Last Decider) and 2 (First Decider). (a) In the modified Justice Game, Player A is endowed with $10 at the beginning of each trial. Player A offers some portion of that money with Player B, who can choose to Accept, Compensate, or Reverse the offer. In our task, Player As always made unfair offers. Because Player Bs could maximize their payout by choosing either Compensate or Reverse, decisions to Reverse can be interpreted as cost-free punishment for Player As. (b) Subjects in the Last Decider experiment observed their group’s punishment preferences prior to indicating their own punishment preference. (c) Subjects in the First Decider experiment indicated their punishment preference prior to observing their group’s preferences. Although group members are represented by icons in this figure, subjects were shown photographs in the actual experiment.CC BY · PubMed Central
Figure 2 Results for Experiments 1 (Last Decider) and 2 (First Decider). For both experiments, preference for punishment increases with the proportion of punishers. This effect remains significant, albeit is reduced, when subjects are observing the moral preferences of past groups who cannot affect the current choice (First Decider). Error bars reflect ± 1 SEM.
Figure 2 Results for Experiments 1 (Last Decider) and 2 (First Decider). For both experiments, preference for punishment increases with the proportion of punishers. This effect remains significant, albeit is reduced, when subjects are observing the moral preferences of past groups who cannot affect the current choice (First Decider). Error bars reflect ± 1 SEM.CC BY · PubMed Central
Figure 3 Schematics and behavioral results of Experiments 3 and 4 (Victim and Juror). (a) As in the Last Decider study, subjects in Experiments 3 and 4 observed their group’s punishment preferences prior to indicating their own punishment preference. In this example, the group is evenly split between punishers and compensators. (b) According to the Drift Diffusion Model (DDM), people make decisions by noisily accumulating evidence in favor of each of the options until a decision bound is reached. In this cartoon schematic, subjects are biased such that they initially prefer compensation over punishment, indicated by the starting point z; they require some threshold of evidence in order to commit to a choice, indicated by a; and they accumulate evidence in favor of compensation, indicated by v. Together, these parameters dictate the shapes of reaction time distributions for punishment and compensation. (c) In this example trial, the group unanimously prefers to punish the unfair offer. (d) We illustrate using a cartoon schematic how conformity to the group underlie shifts in the v and a parameters, thereby changing choice and reaction time distributions. (e) When deciding as either a Victim or Juror, preference for punishment increases with the proportion of punishers, replicating Experiment 1 (Last Decider). Error bars reflect ± 1 SEM.
Figure 3 Schematics and behavioral results of Experiments 3 and 4 (Victim and Juror). (a) As in the Last Decider study, subjects in Experiments 3 and 4 observed their group’s punishment preferences prior to indicating their own punishment preference. In this example, the group is evenly split between punishers and compensators. (b) According to the Drift Diffusion Model (DDM), people make decisions by noisily accumulating evidence in favor of each of the options until a decision bound is reached. In this cartoon schematic, subjects are biased such that they initially prefer compensation over punishment, indicated by the starting point z; they require some threshold of evidence in order to commit to a choice, indicated by a; and they accumulate evidence in favor of compensation, indicated by v. Together, these parameters dictate the shapes of reaction time distributions for punishment and compensation. (c) In this example trial, the group unanimously prefers to punish the unfair offer. (d) We illustrate using a cartoon schematic how conformity to the group underlie shifts in the v and a parameters, thereby changing choice and reaction time distributions. (e) When deciding as either a Victim or Juror, preference for punishment increases with the proportion of punishers, replicating Experiment 1 (Last Decider). Error bars reflect ± 1 SEM.CC BY · PubMed Central
Figure 4 Mean threshold parameter estimates for Victim and Juror DDM experiments. Significance asterisks above/below each bar represent differences from the Alone condition (which was used as the reference category and is represented by a grey dashed line in each panel), and lines indicate pairwise differences. Threshold is estimated to be higher for decision contexts in which subjects determine the final outcome (i.e., choosing alone and with evenly-split groups).
Figure 4 Mean threshold parameter estimates for Victim and Juror DDM experiments. Significance asterisks above/below each bar represent differences from the Alone condition (which was used as the reference category and is represented by a grey dashed line in each panel), and lines indicate pairwise differences. Threshold is estimated to be higher for decision contexts in which subjects determine the final outcome (i.e., choosing alone and with evenly-split groups).CC BY · PubMed Central
Figure 5 Mean drift rate parameter estimates for Victim and Juror DDM experiments. Significance asterisks above/below each bar represent differences from the Alone condition (which was used as the reference category and is represented by a grey dashed line in each panel), and significance bars indicate pairwise differences. Drift rates grow increasingly positive with the proportion of punishers, indicating that people accumulate evidence favoring punishment from groups.
Figure 5 Mean drift rate parameter estimates for Victim and Juror DDM experiments. Significance asterisks above/below each bar represent differences from the Alone condition (which was used as the reference category and is represented by a grey dashed line in each panel), and significance bars indicate pairwise differences. Drift rates grow increasingly positive with the proportion of punishers, indicating that people accumulate evidence favoring punishment from groups.CC BY · PubMed Central
Figure 6 Effects of fairness violation severity on Victims’ and Jurors’ decisions to punish. (a) As indicated by the drift rate intercept, Victims do not distinguish between differing levels of offer unfairness, whereas Jurors do. (b) As shown by the drift rate betas, each additional punisher in the group provides stronger evidence that punishment should be favored, more so for Victims than Jurors.
Figure 6 Effects of fairness violation severity on Victims’ and Jurors’ decisions to punish. (a) As indicated by the drift rate intercept, Victims do not distinguish between differing levels of offer unfairness, whereas Jurors do. (b) As shown by the drift rate betas, each additional punisher in the group provides stronger evidence that punishment should be favored, more so for Victims than Jurors.CC BY · PubMed Central
Figure 7 Change in punishment severity judgments for Victims and Jurors. (a) Schematic illustrating sample trials in which subjects are choosing alone or within a group. Vignettes were written in the second-party perspective in the Victim condition, and in the third-party perspective in the Juror condition. (b) Overall, the proportion of group members endorsing severe punishment increases subjects’ judgments for how severely a perpetrator should be punished for committing a crime. The midpoint of the scale is indicated using a grey dashed line. Error bars reflect ± 1SEM.
Figure 7 Change in punishment severity judgments for Victims and Jurors. (a) Schematic illustrating sample trials in which subjects are choosing alone or within a group. Vignettes were written in the second-party perspective in the Victim condition, and in the third-party perspective in the Juror condition. (b) Overall, the proportion of group members endorsing severe punishment increases subjects’ judgments for how severely a perpetrator should be punished for committing a crime. The midpoint of the scale is indicated using a grey dashed line. Error bars reflect ± 1SEM.CC BY · PubMed Central

Verification

Claims the paper makes

13 Valid · 1 Not valid · 10 Not checked · 76 Own results and interpretations

45 more claims

Create a free account to see all 100 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 38
  • It may state two separate claims as one. 20

Create a free account to see these claims.

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentz = -2.58reported p = .01 · recomputed p = .010

    “z = −2.58, p = .010”

1 of 1

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Citations

Cites

Show all 49 references

Cited by

See all 39

Metadata

Authors
Son, J.-Y., Bhandari, A., & FeldmanHall, O.
Affiliations
Brown University (US)
Corresponding author
FeldmanHall, Oriel
Venue
Scientific Reports
Published
2019-08-12
Volume
9
Issue
1
Pages
11625-11625
ISSN
2045-2322 (ISSN-L)
OpenAlex
W2968071192
Cited by
37
References
49
Citations by year
2025: 10 · 2024: 7 · 2023: 5 · 2022: 7 · 2021: 4 · 2020: 2 · 2019: 2
FWCI
2.18
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • PsyArXiv (OSF Preprints) (submittedVersion, cc-by)
  • PsyArXiv (OSF Preprints) (submittedVersion)
  • OSF Preprints (OSF Preprints) (submittedVersion)
Funders
National Institute of General Medical Sciences (P20GM103645)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Psychology of Moral and Emotional Judgment
  • Social Sciences › Law › Jury Decision Making Processes
  • Social Sciences › Economics, Econometrics and Finance › Economics and Econometrics › Experimental Behavioral Economics Studies
Keywords
punishment, reference groups, social influence, conformity, moral responsibility, restorative justice, group decision making, impulsivity, computational models of decision making, jury members
MeSH terms
Decision Making, Female, Humans, Male, Punishment, Crowdsourcing

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Crowdsourcing punishment: Individuals reference group preferences to inform their own punitive decisions. https://app.cochat.ai/papers/10.1038/s41598-019-48050-2
The paper
Son, J.-Y., Bhandari, A., & FeldmanHall, O. (2019). Crowdsourcing punishment: Individuals reference group preferences to inform their own punitive decisions. Scientific Reports. https://doi.org/10.1038/s41598-019-48050-2

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.