People do not always know exactly how their choices affect others. A purchase may support jobs while also carrying environmental costs. A personal convenience may save time but shift burdens elsewhere. When those consequences are uncertain, people have room to interpret what their actions mean.
New research suggests that this ambiguity is not interpreted neutrally. Across nine preregistered online experiments involving 8,769 adults, researchers found that people who made selfish choices and people who made prosocial choices tended to distort uncertain social consequences in different directions. More strikingly, interpretations that made selfish behaviour appear less harmful were more likely to influence other people’s choices.
The study, published in Nature Human Behaviour on 1 October 2026, provides unusually extensive experimental evidence for a simple psychological possibility: people may disagree not only about how much they value the welfare of others, but also about what they believe their own actions actually do to others.
Turning social ambiguity into an experiment
Marius C. Vollberg and Jörg Gross designed a series of experiments around choices that affected both a decision-maker and another person, described as the Receiver. The key feature was uncertainty. Participants knew a range within which the Receiver’s outcome would fall, but they did not initially know the exact value.
This allowed the researchers to compare people’s estimates with an objective midpoint. If a Receiver could obtain between one and three points, for example, the midpoint was two. An estimate above or below that midpoint could therefore reveal whether participants systematically interpreted the same ambiguous range more favourably or unfavourably.
The experiments distinguished between people who selected an option that maximised their own payoff and those who accepted a personal cost to improve the Receiver’s outcome. Across the programme, the researchers also introduced neutral conditions, explicit explanations of how outcomes were generated, estimates made before choices, forced choices, personal ambiguity tasks, third-party judgements and social influence conditions.
That progression matters. A single experiment could show that selfish and prosocial people report different estimates, but it could not easily distinguish whether those estimates reflect deliberate excuses after a choice, pre-existing social preferences, confusion about uncertainty, or a broader tendency to like or dislike ambiguous outcomes. The nine-experiment design was intended to separate these explanations.
Selfish and prosocial choices produced different interpretations
In the first experiments, selfish decision-makers tended to make the consequences of the selfish and prosocial options look more similar than prosocial decision-makers did. In experiment 1a, the difference between these groups was estimated at b = 0.24, with a 95% confidence interval from 0.14 to 0.34.
The direction of the estimates was revealing. Selfish participants tended to inflate the social benefit associated with their chosen option while downplaying the benefit that the alternative could have delivered. In effect, the perceived social cost of choosing for themselves became smaller.
Prosocial participants also showed distortions, but in the opposite direction. They tended to portray their choice as comparatively more beneficial to the Receiver and sometimes exaggerated the sacrifice implied by choosing the prosocial option. These prosocial distortions were less consistent across experiments than the pattern among selfish decision-makers.
Importantly, the findings do not mean that every selfish or prosocial participant distorted outcomes in the same way. The researchers observed substantial individual variation. In pooled data from experiments 1a and 1b, for example, fewer than 10% of decision-makers produced the exact aggregate distortion pattern that most clearly compressed the difference between the selfish and prosocial outcomes. The headline pattern emerged statistically across groups rather than describing every individual.
The effect appeared before people could justify a decision
A central question was whether participants were simply rationalising a choice they had already made. Experiment 2 therefore reversed the sequence. Participants estimated the Receiver’s outcomes before choosing between the options.
The distortions remained. Selfish decision-makers overestimated the Receiver outcome associated with the selfish option relative to the midpoint, with an estimated coefficient of 0.16 and a 95% confidence interval from 0.10 to 0.22. Prosocial decision-makers showed evidence of exaggerating the Receiver outcome associated with the prosocial option, with an estimated coefficient of 0.06 and a 95% confidence interval from 0.01 to 0.12.
Because these estimates were recorded before the choice, a purely post-choice explanation became less plausible. The researchers then went further by removing choice itself.
In experiment 3a, participants were randomly forced to implement either the selfish or prosocial option. If performing an action caused the distortion, people assigned to different actions should have developed different estimates. Instead, the aggregate estimates did not meaningfully differ according to the randomly imposed choice. A Bayesian comparison strongly favoured the absence of an overall condition effect, with BF01 = 657, while the Bayes factor for the condition-by-option interaction was 32.
This pattern is more consistent with people’s underlying social preferences shaping how they construe ambiguous consequences than with the chosen action mechanically causing a new belief. A conceptual replication also measured participants’ free-choice preferences alongside the forced-choice task to examine this interpretation at the individual level.
Distorted interpretations can change how others behave
The final part of the research asked whether these interpretations stay inside one person’s reasoning or can spread socially.
Experiment 6 recruited 849 participants after exclusions. Before making their own choice, each participant saw estimates attributed to a previous participant. The researchers randomly assigned them to one of four conditions: accurate midpoint estimates, a control in which both outcomes were underestimated, a prosocial-serving interpretation, or a selfish-serving interpretation that made the two social outcomes appear equally good.
The selfish-serving message changed behaviour. Compared with the no-distortion condition, exposure to the selfish-serving interpretation increased selfish choice, with an estimated log-odds difference of 0.49 and a 95% confidence interval from 0.10 to 0.88. It also differed significantly from the control and prosocial-serving conditions.
The prosocial-serving interpretation did not produce a corresponding increase in prosocial choices. Bayesian analyses supported the absence of such an effect, with BF01 = 19.3 relative to the no-distortion condition and BF01 = 10.3 relative to the control condition.
This asymmetry is one of the study’s most consequential findings. An interpretation that reduces the apparent social harm of a selfish action may be especially easy for others to adopt because it offers both a belief and a behavioural permission structure. By contrast, exaggerating the virtue or sacrifice involved in a prosocial action did not appear to pull behaviour in the opposite direction.
Why ambiguity may help disagreements grow
The findings offer a useful way to think about disputes over choices whose consequences are genuinely uncertain. People may hold different values, but they may also construct different factual interpretations of ambiguous consequences in ways that align with those values.
This distinction matters in areas such as climate action, ethical consumption, charitable giving and social welfare. Two people can appear to disagree about whether a sacrifice is worthwhile when part of the disagreement actually concerns how much benefit they believe the sacrifice produces. If those beliefs are themselves shaped by social preferences, providing more information may not always eliminate the disagreement as straightforwardly as expected.
The contagion result also suggests a route through which interpretations can become socially reinforced. If claims that minimise the harm of self-interested behaviour are more behaviourally contagious than claims that magnify prosocial benefits, groups could gradually shift even when everyone begins with access to the same underlying uncertainty.
What the study does not establish
The experiments were tightly controlled and preregistered, but that control also limits how directly the findings can be transferred to complex real-world disputes. The tasks involved relatively simple monetary trade-offs with explicit ranges. Climate policy, consumer ethics and public spending involve far more complicated evidence, identities, institutions and incentives.
The 8,769 participants were predominantly US residents and were recruited online, with some experiments also including UK participants. Cultural differences in social norms, uncertainty and self-presentation could affect whether the same patterns appear elsewhere.
The researchers also emphasise heterogeneity. Distortion was not universal, and some effects, particularly among prosocial decision-makers, varied across experiments. The study therefore supports probabilistic group-level tendencies rather than a claim that selfish people always deceive themselves or that prosocial people always exaggerate their virtue.
Finally, the experiments cannot show that participants consciously manipulated their estimates. A distorted judgement can arise without deliberate dishonesty. The results are compatible with social preferences influencing perception or interpretation itself, which is precisely why the pre-choice and forced-choice findings are important.
A broader lesson about motivated interpretation
The study moves the discussion beyond the familiar idea that people invent excuses after behaving selfishly. Across multiple experimental variations, interpretations aligned with social preferences even before a free choice was made and when choices were externally imposed.
That suggests ambiguity can become part of the decision itself. People do not merely choose between known social consequences according to their preferences. When consequences are uncertain, their preferences may also help shape what they believe those consequences are.
The final experiment adds a social dimension: some interpretations travel better than others. In these experiments, the interpretation that made selfish behaviour easier to defend was the one that subsequently shifted other people’s behaviour. That asymmetry may help explain how disagreements about the impact of everyday choices can persist and spread even when people are ostensibly discussing the same facts.
Source Information
Study: Selfish and prosocial individuals both distort their actions’ impact, but selfish distortions can be more contagious
Authors: Marius C. Vollberg and Jörg Gross
Journal: Nature Human Behaviour
Published: 1 October 2026
DOI: 10.1038/s41562-026-02585-3
Study design: Nine preregistered online experiments involving 8,769 adults
