Human cooperative behavior is essential for the rational exploitation of limited resources and the successful completion of team tasks. Individuals in groups frequently encounter mixed-motivation situations where personal interests conflict with collective interests, or short-term gains oppose long-term benefits. It is precisely these dilemmas of mixed motivations that expose people to various social dilemmas (Dawes and Messick, 2000; Hauser et al., 2019). Researchers define cooperative behavior as an individual's behavioral choice in social dilemmas that suppresses short-term self-interest to maximize collective benefits (Dou et al., 2018; Henrich and Muthukrishna, 2021). When confronting conflicting interests, the ability to eschew hostile competition in favor of cooperative coexistence holds profound significance for enhancing personal wellbeing and advancing human society (Declerck et al., 2013; Rand, 2016; Yin et al., 2025).

To empirically investigate cooperative behavior, researchers commonly employ the ecologically valid public goods game (PGG) paradigm, which simulates real-world dilemmas through a “give-game” mechanism. In this paradigm, N participants receive experimental currency units (ECUs) and independently decide how much to contribute to a public account. The total contributions are multiplied (simulating resource growth) and equally redistributed to all members. Crucially, while higher aggregate contributions lead to greater public account growth and increased individual payoffs, free-riders (those contributing minimally or nothing) can exploit the system by retaining more ECUs than high contributors while still receiving equal shares. If no participant contributes, the public good collapses entirely. Therefore, an individual's contribution level directly indexes their cooperation intensity in public goods dilemmas (Dou et al., 2018; Sun et al., 2024).

Extensive experimental research demonstrates that individuals exhibit fairness preferences even while pursuing self-interest (Choshen-Hillel and Yaniv, 2011; Tsoi and McAuliffe, 2020). Fehr and Schmidt (1999) integrated this social preference for fairness into behavioral economics by proposing the concept of inequity aversion—the tendency to avoid unequal resource distributions (whether advantageous or disadvantageous) and even incur personal costs to achieve equality (Han et al., 2018; Fehr and Schmidt, 1999). Researchers typically operationalize IA through the Ultimatum Game (UG), where participants acting as Responders reject unequal allocation proposals from Proposers. The rejection rate serves as a validated proxy for IA intensity, a methodology robustly supported by empirical studies (Corradi-Dell'Acqua et al., 2013; Hu and Mai, 2021; Wang et al., 2022). While existing evidence confirms that individuals act as both homo economicus (maximizing monetary gains) and homo socialis (sacrificing self-interest to punish norm violators) in resource allocation contexts (Fehr et al., 2008; Nishi and Christakis, 2015; Tanaka et al., 2024), the mechanisms through which inequity aversion influences cooperative behavior in interpersonal interactions remain underexplored.

Real-world social dilemmas are inherently embedded in contexts of objective inequality, where individuals' decision-making processes are inevitably shaped by prior experiences of subjective unfairness (Hauser et al., 2019). At the macro level, wealth disparity and resource inequality amplify social dilemmas—a pattern resonating with Confucius's admonition in Analects 16.1: “The greatest concern lies not in scarcity but in unfair distribution”—indicating that distributive injustice jeopardizes social stability more severely than absolute poverty. Empirical evidence corroborates this ancient wisdom: Oishi and Kesebir (2015) analyzed socioeconomic data from 34 nations and demonstrated that national economic 3growth often fails to enhance collective wellbeing when accompanied by escalating economic inequality, as the psychological costs of resource polarization frequently negate the benefits of material prosperity. At the micro level, exposure to unfair conditions—such as socioeconomic status asymmetry or income inequality—systematically erodes the sustainability of altruistic, prosocial, and cooperative behaviors while heightening unethical conduct (Bechtel et al., 2018; Cheng et al., 2021; Du et al., 2021; Jachimowicz et al., 2020). These findings collectively position unfairness experiences as critical inhibitors of cooperative behavior in interactive games.

Humans' social preference for inequity aversion is rooted in distinct neural substrates. Pioneering fMRI research by Sanfey et al. (2003) revealed that unfair proposals activate emotion-related anterior insula, cognition-control-associated dorsolateral prefrontal cortex (DLPFC), and conflict-resolution-linked anterior cingulate cortex (ACC). Crucially, the activation intensity of the anterior insula exhibits a significant positive correlation with rejection rates of unfair offers—a finding that underscores the role of emotional processing in social decision-making (Sanfey et al., 2003). Complementing these results, Tabibnia and Lieberman (2007) demonstrated that fair proposals engage reward-processing regions, including the ventromedial prefrontal cortex (vmPFC), amygdala, ventral striatum, and orbitofrontal cortex (OFC), indicating that equitable outcomes function as intrinsic “reward stimuli.” While neuroimaging studies have mapped fairness-related brain networks (Fliessbach et al., 2012; Kish-Gephart, 2017; Wang et al., 2022), ERPs provide unparalleled temporal resolution advantages for dissecting inequity aversion dynamics in iterative games. Recent ERP evidence highlights two key components: MFN and P300.

The MFN component, peaking at 250–350 ms post-stimulus over frontocentral regions (with generators localized to the ACC; Gehring and Willoughby, 2002), reflects rapid semi-automatic evaluation of fairness violations during mid-stage processing (Leng and Zhou, 2010). Empirical studies consistently demonstrate that MFN amplitudes are enhanced for negative outcomes Boksem et al., (2011) and stimuli violating social norms (Li et al., 2025; Wang et al., 2017; Wu et al., 2011). In UG, unfair proposals elicit stronger MFN than fair offers (Spape et al., 2019), with (Boksem et al. 2011) further demonstrating that MFN sensitivity to unfairness correlates with individual fairness concerns. These findings collectively establish MFN as a neural marker of inequity sensitivity, where norm-deviant unfair proposals amplify negative deflections.

In contrast, the later-emerging P300 component (300–600 ms post-stimulus, maximal over parieto-occipital regions) indexes advanced cognitive processes during outcome evaluation, including emotional significance appraisal and attentional resource allocation (Li et al., 2025; Mussel et al., 2022). P300 primarily indexes advanced cognitive processes during the late-stage evaluation of fairness outcomes (Li et al., 2025; Mussel et al., 2022). Empirical evidence indicates that P300 amplitude reflects the integration of emotional significance appraisal and attentional resource allocation during decision-making (Hu and Mai, 2021; Leng and Zhou, 2010). Crucially, its amplitude is modulated by both outcome valence and reward magnitude, with positive feedback and larger monetary rewards eliciting enhanced P300 positivity (Bellebaum et al., 2010; Kobza et al., 2011). For instance, Xu et al. (2020) employed a third-party punishment paradigm to investigate neural responses to donation proposals, demonstrating that fair allocations evoke larger significantly lager P300 amplitudes than unfair ones. Converging evidence from UG studies further shows that fair proposals—which align with social norms and yield higher personal gains—consistently elicit more pronounced P300 responses than inequitable offer (Mussel et al., 2022; Wu et al., 2011). Collectively, these findings position P300 as a neural signature of fairness valuation, integrating both normative adherence and utilitarian benefit computations.

Despite extensive evidence that humans balance self-interest with fairness enforcement in social dilemmas, critical gaps persist in understanding how neural dynamics of unfairness processing (e.g., MFN/P300) calibrate cooperative decisions following unfair experiences. The current study integrates the UG and PGG to address these questions through a neurobehavioral lens. We hypothesize that: (1) Individuals will demonstrate stronger aversion to unfair proposals than fair ones, evidenced by higher rejection rates in UG; (2) Participants will exhibit greater cooperation (higher PGG contribution rates) with fair proposers compared to unfair counterparts; (3) Unfair proposals will elicit more negative MFN amplitudes, whereas fair proposals will evoke more positive P300 responses; (4) Neural activity (MFN/P300 amplitudes) triggered by unfairness experiences will significantly correlate with cooperative behavior in repeated games. By combining ERPs with sequential game paradigms, we bridge neurocognitive mechanisms and behavioral outcomes to decode how unfairness reprograms social decision-making.

A priori power analysis was conducted using G^*Power 3.1.9.7 (Faul et al., 2009). Following methodological recommendations for ERP studies (Hu and Mai, 2021; Wang et al., 2022), we set parameters with α = 0.05, power (1-β) = 0.80, and a medium effect size (f = 0.25). The analysis assumed default values for within-subject measurement correlation (ρ = 0.5) and non-sphericity correction (ε = 1.0), yielding a minimum required sample size of 28 participants. This power threshold (80%) aligns with established standards in ERP research where medium effect sizes are typically expected (Larson and Carbine, 2017). We successfully recruited 26 healthy university students, approaching our target sample size of 28. Two participants were excluded due to excessive artifacts in the EEG recordings. The final sample (N = 24) provided 75% power to detect medium effects (f = 0.25), exceeding sensitivity thresholds of observed large effects (f > 0.40). Ultimately, 24 participants (13 males; age range 18–21 years, M_age= 19.50 ± 1.06) met all inclusion criteria: right-handedness, normal or corrected-to-normal vision, no neurological or psychiatric history, and no prior experience with economic games. The study protocol was approved by the Academic Ethics Committee of Jinan University. Participants received ¥50–60 compensation after providing written informed consent.

The study employed a single-factor within-subjects experimental design. The independent variable was the type of proposal scheme, comprising three levels: fair, moderately unfair, and extremely unfair. Dependent variables included: the frequency of proposal rejection in the UG, evoked EEG components, and the amount of money contributed in the PGG. The complete experimental procedure is illustrated in Figure 1.

This study established a dyadic interactive simulation paradigm. We first employed the UG to manipulate participants' perceived unfairness (fair, moderately unfair, extremely unfair), followed by the PGG to measure cooperation intentions during social interactions, thereby testing how unfairness experiences predict cooperative behavior. Critically, each trial involved a new anonymous partner pair. Participants were explicitly instructed: “You will interact with different strangers every round—each UG proposer and PGG co-player is unique and will never reappear.” This one-shot stranger paradigm eliminated reputation effects while allowing clean measurement of unfairness spillover.

To control for potential confounding variables, participants completed self-report measures assessing key extraneous factors, including gender (Irwin et al., 2015), mood state (Dou et al., 2018), dispositional trust (Bianchi and Brockner, 2012), and empathic concern (Kowalski et al., 2015). Sex was recorded using dummy coding (0 = female, 1 = male) to account for potential sex differences in unfairness experience and cooperative behavior. Dispositional trust was evaluated using a five-item questionnaire (Schuessler, 1982) on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree), with higher scores reflecting greater trust propensity. Empathic concern was assessed using a six-item subscale from the Interpersonal Reactivity Index (Davis, 1980), also on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree), with higher scores indicating greater empathic concern. Mood state was measured using the Positive and Negative Affect Scale (PANAS; Watson et al., 1988), which includes 10 items assessing positive and negative affect on a seven-point Likert scale (1 = not at all, 7 = extremely), with higher scores indicating more pronounced mood states.

Specifically, participants first reported their mood states before completing the initial task. After completing the UG and PGG tasks, they required to complete PANAS and a retrospective fairness perception questionnaire under different proposal condition on a seven-point Likert scale (1 = very unfair, 7 = very fair), with higher scores indicating a stronger perception of fairness. Participants were instructed to reflect on their real-time experiences and the opponent's behavior during the task to provide accurate ratings. This approach ensured that both their immediate emotional state and their retrospective fairness judgments were accounted for in the analysis.

Electroencephalogram (EEG) data were continuously recorded using a 64-channel BrainAmp Standard system (Brain Products, Munich, Germany) at a sampling rate of 500 Hz. Electrodes were positioned according to the 10–10 International system using Ag/AgCl ring electrodes. Vertical electrooculogram (VEOG) was monitored via electrodes placed 1 cm above and below the left eye, and horizontal EOG (HEOG) via electrodes at the outer canthi of both eyes. During recording, FCz (standard in 10–10 system) served as reference and AFz as ground, with all impedances maintained below 10 kΩ.

Offline preprocessing in BrainVision Analyzer 2 (Brain Products) followed ARTEMIS standards and (Keil et al. 2014) guidelines. Re-referencing to averaged mastoids (TP9/TP10), followed by ICA-based ocular artifact correction (Jung et al., 2000). The signal was filtered by a 0.05–30 Hz digital band-pass filter (24 dB/octave). The EEGs were segmented from 200 ms before to 800 ms after the presentation of the offer, and baseline correction was performed based on the 200 ms interval before the offer was presented. Segments containing maximum amplitudes that exceeded an absolute value of 100 μV or a voltage step from one sample to the next of 50 μV were excluded by means of automated artifact rejection. Finally, the EEG epochs were averaged separately for the fair, moderately unfair, and extremely unfair. MFN and P300 components were quantified based on grand-averaged waveforms and topographic maps across 24 participants. Temporal windows were defined as: MFN: 260–400 ms (negative peak amplitude within window; Miraghaie et al., 2022); P300: 450–650 ms (positive peak amplitude; Xu et al., 2020). Statistical analyses employed repeated-measures ANOVA. Greenhouse-Geisser correction addressed sphericity violations, with Bonferroni adjustment for multiple comparisons. All conditions retained >40 artifact-free trials after averaging.

We conducted a two-way 3 (proposal scheme) and 2 (gender) repeated measure ANOVA for perceived fairness. There was a significant main effect of proposal scheme, F_{(2, 44)} = 60.35, p < 0.001, ηp2 = 0.73. Post-hoc comparisons confirmed that the perceived fairness of fair proposal (self/other = 5:5) were significantly than moderately unfair (4:6 or 3:7) and extremely unfair proposals (2:8 or 1:9) (ps < 0.001). However, the main effect of gender [F_{(1, 22)} = 0.57, p = 0.46, ηp2 = 0.03] and the interaction effect of these two factors were not significant [F_{(2, 44)} = 1.68, p = 0.21, ηp2 = 0.07]. These results confirm the validity of the UG manipulation in eliciting graded unfairness perceptions and indicate that gender differences did not confound the experimental effects.

The correlation analysis revealed significant negative correlations between dispositional trust and cooperative behavior under both moderately unfairness (r = −0.50, p < 0.05) and extremely unfairness conditions (r = −0.53, p < 0.05). Additionally, empathic concern was negatively correlated with cooperative behavior in the moderately unfairness condition (r = −0.41, p < 0.05). However, no significant correlations were found between other covariates and rejection rates in the UG or contribution in the PGG. Critically, supplementary analyses showed no significant associations between any covariates and MFN amplitudes (ps > 0.05) and P300 amplitudes (ps > 0.05). These results suggest that dispositional trust and empathic concern may influence inequity aversion or cooperation during the interaction, and thus should be controlled as covariates in subsequent analyses.

To test our primary hypothesis regarding how generalized unfairness processing (vs. fairness) shapes cooperation, we employed a binary classification approach within a neural-behavioral integration framework. This consolidation serves three key purposes: (a) maximizes statistical power for mediation analyses, (b) reduces model complexity given our focus on broad unfairness effects, and (c) aligns with established methods examining global inequity aversion (e.g., Sanfey et al., 2003). Unfair experiences were dummy-coded (0 = fair, 1 = moderately or extremely unfair) and correlated with cooperation (T2), Δ Cooperation (T2–T1), UG rejection rates, amplitudes of MFN and P300.

Results revealed significant negative correlations between unfair experiences and both T2 cooperation (r = −0.48, p < 0.001) and Δ Cooperation (r = −0.36, p = 0.004). As illustrated in Figure 6A, UG rejection rates were also negatively correlated with T2 cooperation (r = −0.38, p = 0.002) and Δ Cooperation (r = −0.31, p = 0.009). Notably, as illustrated in Figure 6B, P300 amplitudes showed a significant positive correlation with Δ Cooperation (r = 0.34, p = 0.006), while MFN amplitudes exhibited no significant correlations with any behavioral measures (ps > 0.05). These findings indicate that participants exhibiting reduced P300 amplitudes during unfair (vs. fair) conditions showed diminished adaptive adjustments in cooperation. This pattern of correlations is consistent with a potential mediating role of P300 in the neurobehavioral pathway linking unfairness to cooperation, though formal mediation testing requires future investigation.