Wired and Unwell

This semester, I conducted a Sunset Talk centered around the relationship between artificial intelligence and psychology. Our Sunset Talks are events where we can discuss topics related to MEDlab that interest us personally. My Sunset Talk was based on the article “Use of Artificial Intelligence in Mental Healthcare, Health Psychology, and Related Research: A Narrative Review to Address Challenges and Opportunities” by Rahman, et al. One of the big issues we discussed was whether therapy from a chatbot is better than no therapy for those who can’t afford it or simply can’t access it.

A major caveat with this is that there are no widely provided clinical chatbots, so many people use the most common ones, like ChatGPT. Chatbots like ChatGPT are not FDA approved for clinical use, often provide inaccurate information, and are not beholden to HIPAA. The growing relationship between AI and psychology is important to talk about because it actively affects people’s lives currently; it’s no longer just discussions about possibilities. Due to the prevalence of people using chatbots for therapy, efficient and safe alternatives must be offered to minimize any harmful effects of misinformed treatments.

Another issue I brought up was that individuality in the experience and expression of symptoms may be too nuanced for AI to handle. Psychological disorders do not have a simple or singular cause to identify and treat, and there is still a need for more research regarding their potential genetic influences, so symptoms are usually the point of focus from which to approach treatment of a psychological disorder. A problem also arises with the fact that besides the diversity of psychological disorders themselves, people with the same disorder can present incredibly differently. The question is whether potential treatments using AI can effectively account for this.

Besides potential issues, though, we also discussed the possible positives of AI use in the field of psychology. One of the main points made was to be precise about exactly what we mean when we say “AI”. It’s important to avoid discussing AI as a single entity because there are many possible implementations of it regarding psychology. There is research being done about the potential usefulness of AI regarding diagnosing, therapy, mapping disease progression, and more as well as in cognitive psychology through models like face attraction and affective computing.

The idea we kept coming back to was that there is a serious need for more research due to the seemingly inevitable merging of AI and psychology, but this can and even should be seen as a positive due to the potential breakthroughs; however, we also emphasized the need for addressing issues including ethics, data privacy, and bias when it comes to AI. In terms of possibly expanding this conversation into action, there needs to be some sort of regulation on existing chatbots offering therapy. For example, there should at least be a requirement for chatbots not approved by the FDA for clinical use to provide a disclaimer that they’re not the same as therapy provided by a psychologist. There should also be research directed towards an alternative that is research-proven and protects users’ personal information. With MEDlab specifically, maybe we could develop protocols that chatbot producers could follow regarding personal data safety and ensure factual clinical information is provided.