Trust in recommendation agent (RA) is essential for users to accept recommendations. Authentic social information may enhance users’ trust, but it also raises privacy concerns for users. According to information privacy theory and social role theory, we conceptualize the trust-privacy paradox that social information disclosure generates for RA users. We conduct an online laboratory experiment in China, and we find and validate some interesting findings. The mixed effects of social disclosure can be divided into direct trust mechanisms and indirect privacy mechanisms. For the trust mechanism, partial disclosure decreases trust, while full disclosure increases trust. Nevertheless, in the privacy mechanism, both partial and full disclosure may engender privacy concerns, resulting in distrust. The gender and privacy awareness levels play a moderating role. Generally, the trust mechanism is stronger than the privacy mechanism, and if social information is fully public, it will be beneficial to RAs. This study will encourage collaboration between recommendation platforms and social media.