Previous research suggests that social comparisons affect decision making under uncertainty. However, the role of the length of the social interaction for this relationship remains unknown. This experiment tests the effect of social comparisons on financial risk taking and how this effect is modulated by whether social encounters are sporadic or repeated. Participants carried out a computer task consisting of a series of binary choices between lotteries of varying profitability and risk, with real monetary stakes. After each decision, participants could compare their own payoff to that of a counterpart who made the same decision at the same time and whose choices/earnings did not affect the participants’ earnings. The design comprised three between-subjects treatments which differed in the nature of the social interaction: participants were informed that they would be matched with either (a) a different participant in each trial, (b) the same participant across all trials, or (c) a “virtual participant”, i.e., a computer algorithm. Compared to the non-social condition (c), subjects in both social conditions (a and b) chose lotteries with lower expected value (z = –3.10, p < .01) and higher outcome variance (z = 2.13, p = .03). However, no differences were found between the two social conditions (z = 1.15, p = .25 and z = 0.35, p = .73, respectively). These results indicate that social comparison information per se leads to poorer and riskier financial decisions, irrespective of whether or not the referent other is encountered repeatedly.