AI companions and digital health tools pose mental health risks

Researchers at Northeastern University have identified potential mental health risks associated with the use of AI companions. A peer-reviewed paper published in JMIR Mental Health suggests that the tendency for users to anthropomorphize chatbots—attributing human qualities to non-human entities—can lead to emotional dependence, the reinforcement of unhealthy beliefs, and the displacement of human social connections.

Specific concerns include the tendency for AI systems to become overly agreeable, validating users even when a more critical or cautious response might be safer. Services like Replika and Character.AI are noted for their ability to simulate warmth and encouragement, which may lead vulnerable individuals to substitute professional support or real-world relationships with digital interactions.

Complementary research into AI and digital health highlights concerns regarding bias and data generalizability. Studies examining digital depression assessment datasets—collected via smartphones, wearables, and cameras—showed significant demographic imbalances, such as a majority of female participants and varying age distributions. These gaps in dataset details raise questions about the ability to build unbiased and effective models for mental health assessment.