Why we may be wrong to think we’re bad at reading people, new ESMT Berlin study finds

The study, Are frogs more forgiving than acorns anticipate? How response pressure confounds can generate (apparent) biases in metaperceptions, conducted by Rebecca Schaumberg (ESMT Berlin) and Joseph Simmons (University of Pennsylvania), has been published in the Journal of Experimental Psychology: General. 

We tend to assume others are angrier about our mistakes than they really are, and less grateful for our good deeds than they actually are, a pattern that has produced now-familiar advice: give more compliments, don’t over-apologize, be bolder about asking for things. The new findings show, for the first time, that this advice may rest on a methodological flaw baked into how the underlying studies were designed, rather than on a genuine psychological effect. 

The problem, the researchers explain, lies in how these effects are typically studied. In the standard design, one group of participants is asked to predict how another person will react to their actions, while a second group is asked to report how they themselves react to someone else’s actions. Although these appear to be comparable questions, they are not: predicting that another person will feel “extremely grateful” for a good deed can appear boastful, whereas reporting that one feels “extremely grateful” oneself appears perfectly ordinary.

Similarly, predicting that another person will react with “extreme anger” to a mistake can appear responsible, whereas reporting that one is “extremely angry” oneself can appear unforgiving. Because the two groups are subject to opposing social pressures, differences between their answers can resemble a genuine bias in social perception even when no such bias exists. 

To test this, Schaumberg and Simmons ran 13 rigorously designed experiments. In one, participants imagined being an acorn that falls on a frog’s head: some predicted how negatively the frog would judge the acorn, while others took the frog’s perspective and reported the frog’s own reaction. Despite the fanciful scenario, the same predictor-reporter gap emerged as in studies of real-world actions, such as forgetting to reply to a text or giving a colleague critical feedback. 

In another study, when participants were explicitly instructed to answer as either a very polite or a very impolite person, the apparent “bias” grew larger under polite instructions and fully reversed under impolite ones, showing that the social pressure built into the question, not any real misjudgment, was driving the result. 

“You cannot call something a bias without an accurate benchmark to compare it to,” says Rebecca Schaumberg, assistant professor at ESMT. “In this research area, that benchmark has always been people’s self-reports, treated as unbiased truth. Our findings show these self-reports are shaped by social pressures, too, which means the case for many of these ‘biases’ is much weaker than it appears.” 

The authors stress that their findings do not prove that people are accurate judges of how others see them, but only that current methods cannot reliably tell us whether they are or not. They call for research designs that verify predictions against measurable outcomes, such as what people actually say or do, rather than against unverifiable self-reports of internal feelings.