Columbia Researchers Develop AI Tool Aimed at Predicting Teen Depression

Led by postdoctoral research scientist Isaac Treves from the Department of Psychiatry at Columbia University Vagelos College of Physicians and Surgeons, the research team is working on a smartphone application designed to convert these predictions into proactive interventions. The app would nudge at-risk teens toward healthier thought patterns.

Treves explains, "Given how quickly and efficiently AI can analyze smartphone text entries, an AI-powered phone app could deliver just-in-time interventions to teens to prevent depressive episodes."

To better predict and potentially prevent depression in teenagers, Treves and his advisor Randy P. Auerbach have focused on rumination—a common issue characterized by an excessive preoccupation with negative thoughts. They hypothesized that smartphone text data could serve as an effective means of detecting such negative self-talk, which often signals mood changes.

The study involved over 200 teenagers who used a smartphone app that recorded all their typed communications, collecting over 4.5 million individual entries during a year-long study. Each participant recorded an average of more than 17,000 entries. Treves noted that many teens were open to this monitoring, especially since it was framed as a support tool for their mental health and was automated without real-time oversight.

Previous research utilizing smartphone data to identify depression typically relied on simplistic algorithms that identified certain keywords. However, Treves pointed out that these methods often fall short in interpreting the more nuanced language of teens. In contrast, the AI model developed in this study agreed with human experts on sentiment analysis 84% of the time and effectively predicted the escalation of depressive symptoms in subsequent days.

Although human expertise was essential in refining the AI model, Treves stressed the importance of maintaining human involvement even if the app is implemented. The sensitive nature of monitoring teens' mental health raises ethical concerns, especially regarding privacy and the risk of trauma associated with automatic responses to potential crises. Treves indicated that many schools currently monitor shared device language to address suicide risks, but this approach can lead to inappropriate emergency responses.

The envisioned app aims to complement traditional therapy rather than replace it, offering nudges towards positive activities based on detected negative self-talk. Clinicians could utilize insights from the app during therapy sessions to provide tailored interventions.

Looking ahead, the researchers are seeking funding to develop the app, which would include two main intervention strategies: distraction techniques and problem-solving methods. Treves stated that the app would help users engage in fulfilling activities when experiencing self-criticism and provide specific strategies to tackle their challenges. Unlike existing mental health apps that frequently prompt responses, this tool would operate more subtly, catering to the preferences of teens who might resist more intrusive suggestions.