Results of the Social influence task in Experiment 1 (a1,b1,c1) and Experiment 2 (a2,b2,c2). In Experiment 1, participants believed to be interacting with a human partner, while in Experiment 2 they were aware of interacting with a computer. The partner’s responses were identical across experiments. (a1,a2) Average influence (i.e., shift towards the partner’s response in the final decision) plotted across participants in Experiment 1 and Experiment 2, respectively. (b1,b2) Standardized influence plotted as a function of agents’ response distance. To obtain this graph, for each participant, we standardized trial-by-trial normalized distance between the estimates of the two agents [distance (cm)/stimulus length (cm)] to express each distance value in terms of deviation from the individual subject mean. We clustered standardized distances in three bins: low (distance < − 0.5 SD); medium (− 0.5 SD ≥ distance < 0.5 SD); high (distance ≥ 0.5 SD). We also standardized, for each participant, trial-by-trial influence to express influence values in terms of deviation from individual influence means. Eventually, for each distance range, we averaged individual influence means across subjects. Error bars represent between-subject standard errors. (c1,c2) Scatter plot of individual average influence as a function of the difference between self and partner rating of perceptual inference accuracy, assessed at the end of the Social influence task in both experiments. The two measures are not correlated in Experiment 1, whereas we found a strong correlation in Experiment 2.