Contributions of insula and superior temporal sulcus to interpersonal guilt and responsibility in social decisions
Hypotheses The paper tests whether anterior insula encodes interpersonal guilt as responsibility-contingent affect following self-caused partner harm, whether momentary happiness incorporates partner reward prediction errors with a responsibility-weighted rule, and whether superior temporal sulcus represents partner experiential state via partner RPEs conditional on participant responsibility.
Claims Across an fMRI study (N=40) and a behavioural replication (N=44), happiness decreases more following low partner outcomes when the participant rather than the partner made the choice; a computational model with a socially-weighted partner-RPE term outperforms alternatives, with the social_pRPE weight reliably positive in both studies. Left anterior insula BOLD is elevated in Social relative to Partner trials after negative partner outcomes (peak MNI [-28, 24, -4]), left STS tracks partner RPEs specifically under participant responsibility, and gPPI reveals condition-dependent insula–right IFG coupling.
Inferences The convergence of behavioral, computational, and neural evidence onto a single anatomical claim localizes interpersonal guilt to anterior insula and identifies STS as a mentalizing locus for partner-state tracking gated by agency, reframing social decision-making under risk as a responsibility-weighted affective computation rather than a pure reward comparison. The partner-algorithm deception is a structural caveat all social claims inherit.
▸ Summary
▸Hypotheses tested
Anterior insula encodes interpersonal guilt — the responsibility-contingent affect arising specifically when one's own choice has caused harm to another.
The guilt hypotheses predict a behavioral interaction: happiness should drop more after negative partner outcomes when the participant — rather than the partner — made the choice.
Tested by
Happiness ratings decrease more after negative partner outcomes when the participant made the choice (Social condition) than when the partner made the choice (Partner condition), operationalizing interpersonal guilt as responsibility-contingent unhappiness about partner harm.
The insula-as-guilt-substrate hypothesis predicts elevated anterior insula BOLD in Social vs Partner trials specifically after negative partner outcomes — the (condition × outcome) interaction in an a priori insula ROI.
Tested by
Anterior insula BOLD activity is significantly elevated in the Social condition compared to the Partner condition specifically after negative partner outcomes, tracking the guilt effect (responsibility-contingent partner unhappiness).
Happiness incorporates partner reward prediction errors with a responsibility-weighted rule — partner RPEs caused by the participant's own choices receive an independent, non-zero weight in the happiness computation.
The responsibility-weighting hypothesis predicts that a happiness model with a social_pRPE × responsibility interaction term outperforms nested alternatives, with a positive social_pRPE weight in both studies.
Tested by
A computational model incorporating responsibility-modulated happiness better accounts for the behavioral guilt effect than models without responsibility weighting, providing formal parameterization of the social decision-making component.
The guilt hypotheses predict a behavioral interaction: happiness should drop more after negative partner outcomes when the participant — rather than the partner — made the choice.
Tested by
Happiness ratings decrease more after negative partner outcomes when the participant made the choice (Social condition) than when the partner made the choice (Partner condition), operationalizing interpersonal guilt as responsibility-contingent unhappiness about partner harm.
STS represents the partner's affective state via partner reward prediction errors, but only when the participant bears responsibility for the choice that produced those errors.
The mentalizing-under-responsibility hypothesis predicts that left STS tracks partner reward prediction errors specifically when the participant is responsible for the choice — a parametric modulator that is significant in Social but not in Partner trials.
Tested by
Superior temporal sulcus (STS) BOLD activity tracks partner reward prediction errors specifically when the participant was responsible for the choice, consistent with mentalizing about the partner's experiential state.
▸Dissociations
Being the decision-maker (Social + Solo vs Partner condition) reduces participant happiness independently of outcome (Study 1: t(3600)=−3.92, p<0.0001, β=−0.14, 95%CI=[−0.20, −0.07]; Study 2: t(2870)=−6.07, p<0.0001, β=−0.24, 95%CI=[−0.31, −0.16]), consistent with a psychological cost of assuming responsibility for others.
Happiness ratings decrease more after negative partner outcomes when the participant made the choice (Social condition) than when the partner made the choice (Partner condition), operationalizing interpersonal guilt as responsibility-contingent unhappiness about partner harm.
The guilt effect (happiness decrease after low partner outcomes under participant vs partner choice) is present regardless of whether the participant received the high or low lottery outcome: high own outcome: Study 1 t(39)=−3.58, p<0.001, d=0.56, BF10=32; Study 2 t(43)=−2.68, p=0.01, d=0.40, BF10=3.8; low own outcome: Study 1 t(39)=−3.39, p=0.002, d=0.54, BF10=19; Study 2 t(43)=−3.58, p<0.001, d=0.54, BF10=33.5.
Participant momentary happiness correlates with both participant reward (Study 1: F(1,3598)=691.5, R²=0.16; Study 2: F(1,2630)=650.8, R²=0.20) and partner reward (Study 1: F(1,2426)=128.6, R²=0.05; Study 2: F(1,1735)=111.8, R²=0.06), confirming that participants were emotionally responsive to partner outcomes throughout the task.
Individual dot-product values between participants' guilt-related BOLD responses and the Yu & Koban neural guilt signature do not correlate with individual behavioral guilt effect scores (Spearman's Rho=−0.058, p=0.725), indicating the neural signature does not track individual differences in guilt sensitivity as measured behaviorally.
The anterior insula guilt effect is consistent with a previously published neural guilt signature map (Yu & Koban), providing convergent validity for the insula as a guilt-tracking region.
Participants chose lotteries more often in the Solo condition than in the Social condition in Study 1 (t(4796)=2.54, p=0.011, β=0.164, 95%CI=[0.038, 0.291]) but not in Study 2 (t(3829)=0.23, p=0.82, β=0.015), suggesting a weak and inconsistent effect of social context on risk-taking that does not replicate across studies.
▸Eliminations & validating controls
The guilt effect (happiness decrease after low partner outcomes under participant vs partner choice) is present regardless of whether the participant received the high or low lottery outcome: high own outcome: Study 1 t(39)=−3.58, p<0.001, d=0.56, BF10=32; Study 2 t(43)=−2.68, p=0.01, d=0.40, BF10=3.8; low own outcome: Study 1 t(39)=−3.39, p=0.002, d=0.54, BF10=19; Study 2 t(43)=−3.58, p<0.001, d=0.54, BF10=33.5.
Probability of lottery selection increases with expected value advantage in both Study 1 (t(4796)=9.26, p<3.1e-20, β=0.074, 95%CI [0.059-0.090]) and Study 2 (t(3829)=10.62, p<5.3e-26, β=0.093, 95%CI [0.075-0.110]), confirming participants were sensitive to expected value in their choices.
Bilateral ventral striatum activation increases with the computational model's expected certain rewards and expected values of chosen lotteries (left: pFWE=0.002, T=5.63, d=0.75, Z=5.41, 110 voxels, MNI [−14 8 −8]; right: pFWE=0.005, T=5.46, d=0.70, Z=5.26, 80 voxels, MNI [10 10 −4]), validating the model-based fMRI approach as a manipulation check before the STS analysis.
Bilateral ventral striatum shows greater BOLD response when participants choose the risky (lottery) rather than safe option, irrespective of Social or Solo condition (left cluster: Cohen's d=0.72; right cluster: d=0.85), replicating established striatal responses to risky decision-making.
▸Interpretations
The anterior insula guilt effect is consistent with a previously published neural guilt signature map (Yu & Koban), providing convergent validity for the insula as a guilt-tracking region.
▸Standalone empirical findings
Functional connectivity (gPPI) between anterior insula and inferior frontal gyrus varies by experimental condition, with increased coupling in the Social condition associated with the guilt effect.
Three clusters — precuneus (d=0.79), left temporoparietal junction (TPJ; d=0.59), and medial prefrontal cortex (mPFC; d=0.54) — show greater BOLD response when participants decide for themselves and their partner (Social condition) than for themselves alone (Solo condition), replicating the association of these regions with social decision-making.
▸Scope qualifiers
The "partner" was an EV-maximizing algorithm, not a real person. All social-comparison claims inherit this structural validity assumption.
Two-study design (fMRI N=40, behavioural N=44) using a within-subject lottery paradigm with an EV-maximising algorithmic partner. All neural claims are conditional on Study 1; all behavioural claims are tested in both studies.
▸All claims (alphabetical)
- agency-reduces-happiness fig3 (text, no dedicated panel)
- guilt-effect-independent-of-own-outcome fig3d, fig3h
- guilt-reduces-happiness-after-partner-loss fig3d, fig3h
- guilt-signature-no-individual-difference fig4 (supplement, text)
- happiness-correlates-partner-reward fig3a, fig3b, fig3e, fig3f
- hypothesis-insula-tracks-interpersonal-guilt hypothesis
- hypothesis-responsibility-weights-partner-rpes hypothesis
- hypothesis-sts-mentalizes-partner-state-under-responsibility hypothesis
- insula-guilt-replicates-yu-koban-signature fig4 (supplement)
- insula-ifg-connectivity-guilt fig5
- insula-tracks-guilt-effect fig4e, fig4f
- interprets-yu-koban-guilt-signature fig4 supplement (convergent-validity analysis)
- lottery-choice-increases-with-ev fig2a, fig2d
- partner-algorithm-deception-assumption
- precuneus-tpj-mpfc-social-decisions fig4b
- prediction-behavioral-guilt-effect prediction
- prediction-insula-tracks-guilt prediction
- prediction-responsibility-model-wins-comparison prediction
- prediction-sts-tracks-partner-rpe-under-responsibility prediction
- responsibility-modulates-guilt-computational fig3, table1
- risk-premiums-null-social-solo fig2b, fig2e
- scope-two-study-design scope
- social-prpe-weight-positive fig3c, fig3g, table1
- solo-vs-social-choice-difference fig2a (Study 1 only; null in fig2d Study 2)
- sts-tracks-partner-reward-prediction-errors fig4h
- ventral-striatum-tracks-computational-reward fig4g
- ventral-striatum-tracks-risky-choices fig4a
Abstract mapped to claims
The paper's abstract is shown with each sentence linked to the claim(s) it represents in the dependency graph. Hover or click a sentence to highlight the corresponding claim cards. Below: what the graph contains that the abstract leaves out, and vice versa.
1This study investigated the neural mechanisms involved in feelings of interpersonal guilt and responsibility evoked by social decisions in humans. 2In two studies (one during fMRI), participants repeatedly chose between safe and risky monetary outcomes in social contexts. 3Across conditions, each participant chose for both themselves and a partner (Social condition), or the partner chose for both themselves and the participant (Partner condition), or the participant chose just for themselves (Solo condition, control). 4If the risky option was chosen in the Social or Partner condition, participant and partner could each receive either the high or the low outcome of a lottery with 50% probability, independently of each other. 5Participants were shown the outcomes for themselves and for their partner on each trial and reported their momentary happiness every few trials. 6As expected, participant happiness decreased following both low lottery outcomes for themselves and for the partner. 7Crucially, happiness decreases following low outcomes for the partner were larger when the participant rather than their partner had made the choice, which fits an operational definition of guilt. 8This guilt effect was associated with BOLD signal increase in the left anterior insula. 9Connectivity between this region and the right inferior frontal gyrus varied depending on choice and experimental condition, suggesting that this part of prefrontal cortex is sensitive to guilt-related information during social choices. 10Variations in happiness were well explained by computational models based on participants’ and partners’ rewards and reward prediction errors. 11A model-based analysis revealed a left superior temporal sulcus cluster that tracked partner reward prediction errors that followed participant choices. 12Our findings identify neural mechanisms of guilt and social responsibility during social decisions under risk.
- C2 lottery-choice-increases-with-ev fig2a, fig2d Probability of lottery selection increases with expected value advantage in both Study 1 (t(4796)=9.26, p<3.1e-20, β=0.074, 95%CI [0.059-0.090]) and Study 2 (t(3829)=10.62, p<5.3e-26, β=0.093, 95%CI [0.075-0.110]), confirming participants were sensitive to expected value in their choices.
- D4 solo-vs-social-choice-difference fig2a (Study 1 only; null in fig2d Study 2) Participants chose lotteries more often in the Solo condition than in the Social condition in Study 1 (t(4796)=2.54, p=0.011, β=0.164, 95%CI=[0.038, 0.291]) but not in Study 2 (t(3829)=0.23, p=0.82, β=0.015), suggesting a weak and inconsistent effect of social context on risk-taking that does not replicate across studies.
- C3 risk-premiums-null-social-solo fig2b, fig2e Risk premiums do not differ between Solo and Social conditions in either study (Study 1: t(39)=1.53, p=0.134, Cohen's d=0.24, BF10=0.49; Study 2: t(43)=−0.21, p=0.84, d=−0.03, BF10=0.17), providing evidence against social-context-driven changes in risk aversion in this paradigm.
- C1 guilt-effect-independent-of-own-outcome fig3d, fig3h The guilt effect (happiness decrease after low partner outcomes under participant vs partner choice) is present regardless of whether the participant received the high or low lottery outcome: high own outcome: Study 1 t(39)=−3.58, p<0.001, d=0.56, BF10=32; Study 2 t(43)=−2.68, p=0.01, d=0.40, BF10=3.8; low own outcome: Study 1 t(39)=−3.39, p=0.002, d=0.54, BF10=19; Study 2 t(43)=−3.58, p<0.001, d=0.54, BF10=33.5.
- D1 agency-reduces-happiness fig3 (text, no dedicated panel) Being the decision-maker (Social + Solo vs Partner condition) reduces participant happiness independently of outcome (Study 1: t(3600)=−3.92, p<0.0001, β=−0.14, 95%CI=[−0.20, −0.07]; Study 2: t(2870)=−6.07, p<0.0001, β=−0.24, 95%CI=[−0.31, −0.16]), consistent with a psychological cost of assuming responsibility for others.
- C5 ventral-striatum-tracks-risky-choices fig4a Bilateral ventral striatum shows greater BOLD response when participants choose the risky (lottery) rather than safe option, irrespective of Social or Solo condition (left cluster: Cohen's d=0.72; right cluster: d=0.85), replicating established striatal responses to risky decision-making.
- E2 precuneus-tpj-mpfc-social-decisions fig4b Three clusters — precuneus (d=0.79), left temporoparietal junction (TPJ; d=0.59), and medial prefrontal cortex (mPFC; d=0.54) — show greater BOLD response when participants decide for themselves and their partner (Social condition) than for themselves alone (Solo condition), replicating the association of these regions with social decision-making.
- C4 ventral-striatum-tracks-computational-reward fig4g Bilateral ventral striatum activation increases with the computational model's expected certain rewards and expected values of chosen lotteries (left: pFWE=0.002, T=5.63, d=0.75, Z=5.41, 110 voxels, MNI [−14 8 −8]; right: pFWE=0.005, T=5.46, d=0.70, Z=5.26, 80 voxels, MNI [10 10 −4]), validating the model-based fMRI approach as a manipulation check before the STS analysis.
- I1 insula-guilt-replicates-yu-koban-signature fig4 (supplement) The anterior insula guilt effect is consistent with a previously published neural guilt signature map (Yu & Koban), providing convergent validity for the insula as a guilt-tracking region.
- D3 guilt-signature-no-individual-difference fig4 (supplement, text) Individual dot-product values between participants' guilt-related BOLD responses and the Yu & Koban neural guilt signature do not correlate with individual behavioral guilt effect scores (Spearman's Rho=−0.058, p=0.725), indicating the neural signature does not track individual differences in guilt sensitivity as measured behaviorally.