Fewer than 10% showed the selfish-serving pattern. But if it spreads and the prosocial one doesn't, that minority can shift collective behaviour. The task is abstract, but we hope it will help illuminate distortions in domains like consumption, climate action, foreign aid, leaving Twitter (or not).

Now what if you see the estimates of a past participant (no additional authority) before you make your choice? Only the selfish-serving distortion (think: "my choice didn't do much harm" / "the alternative wouldn't have helped much") caught on. Followers adopted it and behaved more selfishly.

Just post-hoc rationalization? Unlikely. Distortions showed up before people chose, and under forced choice they tracked what people freely chose, not what they were assigned. Moreover, social distortions didn't carry over to people's choices under ambiguous *personal* payoffs (double standard?).

Selfish actors reported Receiver outcomes as closer together, reducing the social cost of their choice. Prosocial actors comparatively reported the prosocial option as better for the Receiver and sometimes also inflated the social outcome of the alternative, increasing their implied sacrifice.

9 preregistered online studies (N = 8,769) built on one task. Your own payoff is certain (7 vs 5); the other person's is only known as a range (1–3 vs 3–5). You choose one and estimate what they actually got/would have gotten (accuracy incentivized). Deviations from range midpoints = distortion.

Congratulations!!