Throughout our daily lives we make multitudes of decisions. Occasionally, these decisions may lead to favorable outcomes, such as gains, while at other times they may lead to unfavorable outcomes, such as losses. We generally tend to repeat those decisions that have led to favorable outcomes in the past, and we tend to refrain from repeating those decisions that previously led to unfavorable outcomes. How relevant information is integrated over time in order to form optimal decisions is likely influenced by NMDA receptor activity. Here we tested the effects of ketamine, an NMDA receptor antagonist, which has been suggested to affect a wide array of cognitive functions, on choice behavior in a biased matching pennies task in monkeys. In this task, monkeys had to collect six tokens in order to receive a reward. With every binary choice, monkeys would either gain tokens, lose tokens or the number of tokens would remain unchanged. Monkeys generally tended to switch choice targets following losses and to repeat choice targets following gains. Low doses of ketamine temporarily changed choice behavior in particular following unfavorable (loss or neutral) outcomes, but not after favorable (gain) outcomes. Specifically, monkeys were less likely to switch away from a choice target after it had led to an unfavorable outcome under ketamine conditions. Reinforcement learning models suggest that ketamine independently and to varying degrees increased the value of the three possible outcomes in this task. The strongest value increase occurred for loss outcomes. This was the case when ketamine was administered intramuscularly or intranasally, whereby intranasal ketamine induced substantially less side-effects on eye movements. Overall, we found evidence that ketamine may temporarily mitigate in particular the negative valuations of unfavorable outcomes, potentially allowing some insights into its underlying cognitive effects.