Investors increasingly can obtain assistance from "robo-advisors," artificial intelligence - enabled digitalized service agents imbued with anthropomorphic design elements that can communicate using natural language. The present article considers the impact of anthropomorphized robo-advisors on investment decisions, with a focus on their ability to mitigate investors' behavioral biases. We study the well-documented disposition effect, which reflects investors' greater propensity to realize past gains than past losses. In two induced-value laboratory experiments, the availability of a robo-advisor reduces (i.e., mitigates) investors' disposition effect. This relationship is mediated by two simultaneous (indirect) effects: the extent of requests for the robo-advisor's investment advice and perceptions of its socialness. These findings resonate with cognitive dissonance theory, which predicts that assigning responsibility to the advisor helps investors resolve a sense of discomfort that may arise after a financial loss. Anthropomorphic design elements alone are not sufficient to reduce the disposition effect, but they decrease investors' propensity to seek advice, which offsets the positive (indirect) effect of perceived socialness. These results have implications for the ongoing automation of advisory services, as well as for improving decision making, and suggest some further research directions.