CoChat
Sign Up
Cited by 102Experimental Behavioral Economics StudiesNeural and Behavioral Psychology StudiesPsychology of Moral and Emotional Judgment

Neurobiological mechanisms of responding to injustice

Open access
Stallen, Mirre, Rossi, Filippo, Heijne, Amber, Smidts, Ale, De Dreu, Carsten K.W., Sanfey, Alan G.·2018·The Journal of Neuroscience
DOI 10.1523/jneurosci.1242-17.2018PMC PMC6596077
Open this paper in CoChat — free accountOpen PDF

Abstract

People are particularly sensitive to injustice. Accordingly, deeper knowledge regarding the processes that underlie the perception of injustice, and the subsequent decisions to either punish transgressors or compensate victims, is of important social value. By combining a novel decision-making paradigm with functional neuroimaging, we identified specific brain networks that are involved with both the perception of, and response to, social injustice, with reward-related regions preferentially involved in punishment compared with compensation. Developing a computational model of punishment allowed for disentangling the neural mechanisms and psychological motives underlying decisions of whether to punish and, subsequently, of how severely to punish. Results show that the neural mechanisms underlying punishment differ depending on whether one is directly affected by the injustice, or whether one is a third-party observer of a violation occurring to another. Specifically, the anterior insula was involved in decisions to punish following harm, whereas, in third-party scenarios, we found amygdala activity associated with punishment severity. Additionally, we used a pharmacological intervention using oxytocin, and found that oxytocin influenced participants' fairness expectations, and in particular enhanced the frequency of low punishments. Together, these results not only provide more insight into the fundamental brain mechanisms underlying punishment and compensation, but also illustrate the importance of taking an explorative, multimethod approach when unraveling the complex components of everyday decision-making. SIGNIFICANCE STATEMENT The perception of injustice is a fundamental precursor to many disagreements, from small struggles at the dinner table to wasteful conflict between cultures and countries. Despite its clear importance, relatively little is known about how the brain processes these violations. Taking an interdisciplinary approach, we combine methods from neuroscience, psychology, and economics to explore the neurobiological mechanisms involved in both the perception of injustice as well as the punishment and compensation decisions that follow. Using a novel behavioral paradigm, we identified specific brain networks, developed a computational model of punishment, and found that administrating the neuropeptide oxytocin increases the administration of low punishments of norm violations in particular. Results provide valuable insights into the fundamental neurobiological mechanisms underlying social injustice.

AI snapshot

Written from the paper’s full text.

TL;DR

In a novel Justice Game with fMRI and intranasal oxytocin, participants preferred punishment over compensation, and punishment decisions dissociated into willingness to punish and severity. Anterior insula tracked willingness to punish especially in second-party contexts, amygdala tracked severity in third-party contexts, and oxytocin increased low punishments and fairness expectations.

Key findings

  • Participants spent more chips in second-party punishment (M = 38.07, SD = 23.55) than third-party punishment (M = 23.52, SD = 18.47) and compensation (M = 13.93, SD = 11.63); punishment exceeded compensation, t(53) = 6.06, p < 0.001.
  • Ventral striatum activity was higher for punishment versus compensation decisions, while temporal parietal junction activity was associated with deciding not to punish.
  • The computational model best fit was Model 4 (LL = 162, AIC = 333, R² = 0.68, accuracy = 46%), supporting separable willingness-to-punish and punishment-severity parameters; anterior insula correlated with willingness to punish in second-party games, and amygdala correlated with punishment severity in third-party games.
  • Oxytocin did not affect mean punishment or compensation (F(1,52) = 0.12, p = 0.742) but increased low punishments in second-party games (interaction p = 0.050) and, in third-party games, more strongly as the amount taken increased (p = 0.017); oxytocin-group participants more often changed expectations to expect no chips (12/27 vs 3/27, χ²(1) = 7.45, p = 0.006).

Methods

  • Randomized, double-blind, placebo-controlled between-subjects design with 55 males (28 oxytocin); behavioral analyses used 54 participants, and neuroimaging/model analyses used 53 participants.
  • The Justice Game comprised 144 trials across second-party punishment, third-party punishment, and third-party compensation conditions, with preprogrammed fair and unfair Taker choices; fMRI data were analyzed with GLMs and ROI analyses in anterior insula, amygdala, and DLPFC.
  • A computational utility model with parameters for willingness to punish and punishment severity was fit by minimizing residual sum of squares and compared with Nash equilibrium and inequity aversion models.

Limitations

  • Stated: the lack of oxytocin effects on compensation may be due to restriction of range; neural interactions may require larger sample sizes; DLPFC involvement in third-party punishment needs larger-sample follow-up.
  • Stated: inferring psychological processes from brain activity requires caution.
  • Evident: all participants were male, limiting generalizability; the exploratory correlational ROI analyses cannot establish causal mechanisms.

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

Integrity signals

Registry

Crossref · 10.1523/jneurosci.1242-17.2018

Cites

47 of 47 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 45 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The study identified specific brain networks involved with both the perception of, and response to, social injustice.

    “By combining a novel decisionmaking paradigm with functional neuroimaging, we identified specific brain networks that are involved with both the perception of, and response to, social injustice, with reward-related regions preferentially involved in punishment compared with compensation.”
    Quote in the paperSubstantial claim
  • Findingown data

    The neural mechanisms underlying punishment differ depending on whether one is directly affected by the injustice, or whether one is a third-party observer of a violation occurring to another.

    “Results show that the neural mechanisms underlying punishment differ depending on whether one is directly affected by the injustice, or whether one is a third-party observer of a violation occurring to another.”
    Quote in the paperSubstantial claim
  • Findingown data

    In third-party scenarios, amygdala activity is associated with punishment severity.

    “Specifically, the anterior insula was involved in decisions to punish following harm, whereas, in third-party scenarios, we found amygdala activity associated with punishment severity.”
    Quote in the paperSubstantial claim
  • Findingown data

    Reward-related regions are preferentially involved in punishment compared with compensation.

    “By combining a novel decisionmaking paradigm with functional neuroimaging, we identified specific brain networks that are involved with both the perception of, and response to, social injustice, with reward-related regions preferentially involved in punishment compared with compensation.”
    Quote in the paperSubstantial claim
  • Findingown data

    The anterior insula is involved in decisions to punish following harm.

    “Specifically, the anterior insula was involved in decisions to punish following harm, whereas, in third-party scenarios, we found amygdala activity associated with punishment severity.”
    Quote in the paperSubstantial claim
  • May combine two claims. 14
  • Classified as probably not a substantial claim. 14
  • The quote only partly states this claim. 1

  • May combine two claims.

    Findingown data

    Participants were willing to incur costs to both punish and compensate in response to unfairness.

    “Repeated-measures multivariate ANOVAs with participants' mean levels of punishment and compensation behavior as dependent variables, confirmed that participants were willing to incur costs to both punish and compensate in response to unfairness, and also that behavior differed across game types (F (2,51) ϭ 24.05, p Ͻ 0.001, partial ␩ 2 ϭ 0.49).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Oxytocin influenced participants' fairness expectations, and in particular enhanced the frequency of low punishments.

    “Additionally, we used a pharmacological intervention using oxytocin, and found that oxytocin influenced participants' fairness expectations, and in particular enhanced the frequency of low punishments.”
    • Also stated in: § Significance Statement · plain
      “Using a novel behavioral paradigm, we identified specific brain networks, developed a computational model of punishment, and found that administrating the neuropeptide oxytocin increases the administration of low punishments of norm violations in particular.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In the oxytocin group, 12 of 27 participants changed their beliefs after treatment to expect other players to take no chips at all, whereas only 3 of 27 in the placebo group changed their expectations.

    “12 of 27 participants in the oxytocin group changed their beliefs after treatment, now expecting other players to take no chips at all, whereas only 3 of 27 participants in the placebo group changed their expectations”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    No neural activity correlated with individual estimates of parameter α, nor with interactions between treatment and estimates of willingness to punish or α.

    “No neural activity correlated with individual estimates of parameter ␣, nor with interactions between treatment and estimates of or ␣.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Punishment is not a single, unitary process, but comprises distinct subcomponents.

    “Recent neuroscientific work suggests that punishment is not a single, unitary, process, but rather comprises distinct subcomponents (Krueger and Hoffman, 2016).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Krueger and Hoffman, 2016

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Effects of oxytocin are present between 30 and 60 min after intranasal administration.

    “This timing conforms to other studies showing that effects of oxytocin are present between 30 and 60 min after intranasal administration (Baumgartner et al., 2008;De Dreu et al., 2010, 2011;Stallen et al., 2012).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Baumgartner et al., 2008

      Not checked in the free run.

    • cites De Dreu et al., 2010, 2011

      Not checked in the free run.

    • cites Stallen et al., 2012

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    One account of punishment in response to unfairness posits that punishment is driven by social preferences for equitable outcomes.

    “One account of punishment in response to unfairness posits that punishment is driven by social preferences for equitable outcomes (Fehr and Schmidt, 1999).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Fehr and Schmidt, 1999

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    The final determination of appropriate punishment actions has been suggested to critically depend on the DLPFC.

    “whereas the final determination of appropriate punishment actions has been suggested to critically depend on the DLPFC (Knoch et al., 2006;Buckholtz et al., 2008Buckholtz et al., , 2015;;Treadway et al., 2014;Zhong et al., 2016).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Knoch et al., 2006

      Not checked in the free run.

    • cites Buckholtz et al., 2008

      Not checked in the free run.

    • cites Treadway et al., 2014

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Many studies first demonstrated oxytocin as promoting specifically prosocial behavior, but recent work has suggested a more complex role, potentially also relevant to negative, punishment, behavior.

    “Many studies first demonstrated oxytocin as promoting specifically prosocial behavior, but recent work has suggested a more complex role (Ma et al., 2016), potentially also relevant to negative, punishment, behavior (Krueger et al., 2013;Daughters et al., 2017).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Ma et al., 2016

      Not checked in the free run.

    • cites Krueger et al., 2013

      Not checked in the free run.

    • cites Daughters et al., 2017

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The deliberate act of taking resources from another elicits very negative emotions in the person from whom the resources are taken, and this is typically experienced as a stronger norm violation than receiving an unfair share of resources.

    “Experimental work in economics has shown that this deliberate act elicits very negative emotions in the person from whom the resources are taken, and that this is typically experienced as a stronger norm violation than receiving an unfair share of resources (Bosman and Van Winden, 2002;Ben-Shakhar et al., 2007;Korenok et al., 2014), hence our use of this manipulation.”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Bosman and Van Winden, 2002

      Not checked in the free run.

    • cites Ben-Shakhar et al., 2007

      Not checked in the free run.

    • cites Korenok et al., 2014

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Alternative accounts suggest that affective experiences, such as anger, envy, or retaliation, are key drivers in sanctioning norm violators.

    “However, alternative accounts suggest that affective experiences, such as anger, envy, or retaliation, are key drivers in sanctioning norm violators (Rabin, 1993;Dufwenberg and Kirchsteiger, 2004;Pedersen et al., 2013).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Rabin, 1993

      Not checked in the free run.

    • cites Dufwenberg and Kirchsteiger, 2004

      Not checked in the free run.

    • cites Pedersen et al., 2013

      Not checked in the free run.

  • Classified as probably not a substantial claim. The quote only partly states this claim.

    Study design and conductcitation

    The study avoided directly using the terms 'punishment' and 'compensation' to prevent influencing individuals' actual fairness norms.

    “We avoided directly using the terms”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Pedersen et al., 2013

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    The findings on fairness and unfairness align with previous neuroimaging work, providing support for the validity of the novel task.

    “These findings align well with previous neuroimaging work on fairness and unfairness, providing support for the validity of our novel task”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Sanfey et al., 2003

      Not checked in the free run.

    • cites Tabibnia et al., 2008

      Not checked in the free run.

    • cites Harle ´et al., 2012

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    There may not be a one-to-one mapping between the willingness-to-punish parameter and sensitivity to unfairness; the final decision to punish may also depend on other factors, such as whether one is hurt directly.

    “suggesting there may not be a one-to-one mapping between parameter and sensitivity to unfairness, but that the final decision to punish may also depend on other factors, such as whether one is hurt directly”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductnone given

    We developed a novel behavioral task and a new computational model of punishment that allowed us to study both the psychological and neural mechanisms underlying punishment and compensation in response to norm violations.

    “we developed a novel behavioral task and a new computational model of punishment that allowed us to study both the psychological and neural mechanisms underlying punishment and compensation in response to norm violations”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The Inequity Aversion model argues that people value equality and demonstrate a preference that everyone receives the same amount, even if this is zero.

    “Second, we tested performance of Inequity Aversion model (AI), a prominent model of social preferences that argues that people value equality and demonstrate a preference that everyone receives the same amount, even if this is zero (Fehr and Schmidt, 1999).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Fehr and Schmidt, 1999

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Pure fairness considerations (represented by model-derived AIns activation) may play a decisive role in second-party punishment situations, whereas additional contextual information is integrated in punishment decisions in third-party contexts, possibly involving the DLPFC.

    “Possibly therefore, pure fairness considerations (as represented here by model-derived AIns activation) play a decisive role in second-party punishment situations, whereas additional contextual information is integrated in punishment decisions in third-party contexts, when others are harmed (possibly involving the DLPFC).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The TPJ is associated with perspective-taking.

    “the TPJ, an area associated with perspective-taking”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Carter and Huettel, 2013

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Larger sample sizes may be required to find significant interactions between treatment and parameter correlates at the neural level.

    “To find significant interactions between treatment and parameter correlates at the neural level larger sample sizes may be required.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The study prespecified three brain regions of interest: amygdala, anterior insula, and dorsolateral prefrontal cortex.

    “These prespecified brain regions included the amygdala, the anterior insula (AIns) and the dorsolateral prefrontal cortex (DLPFC).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    The computational model allows testing whether the willingness to punish and the decision of how severely to punish are separable processes.

    “Another advantage of the development of the computational model is that this allows us to test whether the willingness to punish and the decision of how severely to punish are separable processes.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Individuals' willingness to punish depends on their fairness considerations, with fairness preferences being stronger when one was directly involved than when another person was the victim.

    “a reasonable interpretation is that individuals' willingness to punish depends on their fairness considerations, with individuals' fairness preferences being stronger when one was directly involved than when another person was the victim”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites Gu ¨rog ˘lu et al., 2010

      Not checked in the free run.

    • cites Corradi-Dell'Acqua et al., 2013

      Not checked in the free run.

    • cites Zhong et al., 2016

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The DLPFC is known for its role in integrating information and response selection.

    “The DLPFC is known for its role in integrating information and response selection”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Ridderinkhof et al., 2004

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    The present findings provide more insight into the fundamental brain mechanisms underlying punishment and compensation and demonstrate how combining a computational approach with neuroimaging methods can help disentangle the psychological motives underlying complex decision-making processes.

    “The present findings not only provide more insight into the fundamental brain mechanisms underlying punishment and compensation, but also demonstrate how combining a computational approach with neuroimaging methods can help disentangling the psychological motives underlying complex decision-making processes.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    Oxytocin had a systematic effect on punishment decisions at both the behavioral and neural levels.

    “with regard to punishment decisions, oxytocin had a systematic effect at both the behavioral and neural levels”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Previous work on third-party punishment in hypothetical criminal contexts has found the DLPFC involved in integrating information about the responsibility of the wrongdoer and the amount of harm done.

    “previous work on third-party punishment in hypothetical criminal contexts has found this area involved in integrating information about the responsibility of the wrongdoer and the amount of harm done”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Buckholtz et al., 2015

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The study conducted analyses only on these three brain regions because of strong a priori hypotheses.

    “We conducted our analyses only on these three brain regions, as we had strong a priori hypothesis about these areas in particular.”
    Quote in the paperSubstantial claim
27 of 27

  • The paper relies on this without support.

    People are particularly sensitive to injustice.

    “People are particularly sensitive to injustice.”

  • The paper relies on this without support.

    Violations of social norms trigger an emotional response in the victims of such violations and can also lead to strong responses from third parties.

    “In fact, violations of social norms, standards of behavior that are based on widely shared beliefs about how one should behave, not only trigger an emotional response in the”

  • The paper relies on this without support.

    Relatively little is known about how the brain processes violations of social norms.

    “Despite its clear importance, relatively little is known about how the brain processes these violations.”

  • The paper relies on this without support.

    The amount of chips taken by the Taker indexes the magnitude of the social norm violation.

    “the more that was taken from the victim”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Recent research has shown that willingness to punish severely was associated with increased amygdala activity in third-party punishment.

    “recent research has shown that, in a scenario in which an observer punishes transgressors due to social norm violation (i.e., third-party punishment), the willingness to punish severely was associated with increased amygdala activity (Stallen et al., 2018)”

    From A neurobiological association of revenge propensity during intergroup conflict (2020) · cites this paper as Stallen et al., 2018

Self-check

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

Each contradiction between two passages listed here was confirmed by an independent check.

  • Two different third-party punishment constraints are stated

    Minor

    The first third-party constraint depends on the chips taken and subtracts punishment spending twice. Equation 5 instead gives a constraint of less than 200 chips for the same game type; the two define different feasible choices.

  • Reported AIC does not follow the stated formula

    Minor

    For Model 4, the stated formula uses k = 4 and LL = 162. Its 2k − 2LL terms alone equal −316, whereas the table reports an AIC of 333; that value is consistent with treating 162 as a negative log-likelihood instead.

  • Response-scale options disagree with the modeled choices

    Minor

    A scale starting at 1 and advancing in steps of 10 would offer 1, 11, and so on, whereas the model specifies choices starting at 10 and advancing in steps of 10.

  • Oxytocin-group count does not reflect the imaging-data loss

    Minor

    The behavioral sample contained 27 oxytocin participants. If the one participant whose imaging data was lost was in that group, the 53-person neuroimaging sample would contain 26, not 27, oxytocin participants.

4 of 4

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. 10.2307/2117561 · DOI

Cited by

Sorted by citations

Related

Authors
Stallen, M., Rossi, F., Heijne, A., Smidts, A., De Dreu, C. K. W., & Sanfey, A. G.
Affiliations
  • Leiden University (NL)
  • University of California San Diego (US)
  • Radboud University Nijmegen (NL)
  • Erasmus University Rotterdam (NL)
ORCID
Venue
The Journal of Neuroscience
Publisher
Society for Neuroscience
Published
2018-02-19
Volume
38
Issue
12
Pages
2944-2954
ISSN
  • 0270-6474 (ISSN-L)
  • 1529-2401
DOI
10.1523/jneurosci.1242-17.2018
PMC
PMC6596077
OpenAlex
W2788810732
Semantic Scholar
aea67e7dd595fa5973a5eaac11a11d1ffcb8fca4
Cited by
102
References
47
Citations by year
2026: 6 · 2025: 15 · 2024: 11 · 2023: 18 · 2022: 14 · 2021: 12 · 2020: 15 · 2019: 6 · 2018: 5
FWCI
18.94
Type
journal article
Language
en
Open access
Licence
Locations
  • Journal of Neuroscience (publishedVersion, cc-by-nc-sa)
  • PubMed Central (publishedVersion)
  • PubMed (publishedVersion)
  • Leiden Repository (Leiden University) (submittedVersion)
  • EUR Research Repository (Erasmus University Rotterdam) (submittedVersion, other-oa)
  • RePub (Erasmus University Rotterdam) (submittedVersion, other-oa)
  • University of Groningen research database (University of Groningen / Centre for Information Technology) (submittedVersion, other-oa)
  • Radboud Repository (Radboud University) (submittedVersion)
  • Radboud Repository (Radboud University) (submittedVersion, other-oa)
  • Data Archiving and Networked Services (DANS) (submittedVersion, other-oa)
  • Data Archiving and Networked Services (DANS) (submittedVersion)
  • ERIM Top-Core Articles
Funders
  • Erasmus Research Institute of Management
  • FP7 Ideas: European Research Council (313454)
Topics
  • Social Sciences › Safety Research › Experimental Behavioral Economics Studies
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Psychology of Moral and Emotional Judgment
Keywords
social injustice, punishment decisions, compensation decisions, fairness, third-party punishment, anterior insula, amygdala, oxytocin, reward processing, norm violation
MeSH terms
Brain, Computer Simulation, Decision Making, Double-Blind Method, Humans, Male, Oxytocin, Punishment, Reward, Social Justice, Compensation and Redress, Young Adult

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

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