Estimating others' pain is a challenging inferential process, associated with a high degree of uncertainty. The present study exploited models of probabilistic decision-making to investigate how uncertainty influences the assessment of another’s pain. We engaged sixty-three dyads (43 strangers and 20 romantic couples) in a task where individual choices affected the pain delivered to either oneself (the agent) or the other member of the dyad. At each trial, agents were presented with cues predicting a given pain intensity with an associated probability of occurrences. Agents chose either a sure (mild decrease of pain) or risky (50% chance of avoiding pain altogether) management option, following which they were asked to bid on their choice. A heat stimulation was then issued to the target (self or other), whose intensity was rated. We found that, the higher the expected pain, the more risk-averse agents became, in line with findings in value-based decision-making. Furthermore, agents gambled less on another individual’s pain (especially strangers) and placed higher bids on pain relief than they did for themselves. Most critically, the uncertainty associated with expected pain dampened ratings made for strangers’ pain. This contrasted with the effect on an agent’s own pain, for which risk had a marginal hyperalgesic effect. Overall, our results suggested that risk selectively affects decision-making on a stranger’s suffering, both at the level of assessment and treatment selection, by 1) leading to underestimation, 2) privileging sure options, and 3) altruistically allocating more money to insure the treatment’s success.