The Role of Data in Applying AI to Mental Health Practice

• In Event: The Intersection of AI: Entertainment, Mental Health, Media Violence, and Health Care Disparities

Abstract

Mental health in particular is a unique field of medicine, compared to others like radiology or cardiology, because of its innate subjective nature. Since there is “no unique and efficient clinical set of indicators” (Yan et al. 2023) for certain Major Depressive Disorders (MDDs), it is impossible to approach patient diagnosis as a matter of a checklist. Furthermore, even if someone exhibits symptoms of a MDD, it does not necessarily warrant a clinical diagnosis. An individual’s mental health is not binary and is better represented as a range or spectrum. Oftentimes, it is ultimately up to the psychiatrist’s discretion and professional experience to diagnose a patient. Another emerging issue is that psychiatrists also also becoming decreasingly effective at work. A study by Graham et al. in 2019 states that “Physician time is progressively limited as mental healthcare needs grow and clinicians are burdened with increased documentation requirement and inefficient technology.”
This paper will consider the application of artificial intelligence (AI) in mental health diagnosis and research to address these problems, as well as the importance of the fundamental requirements in data for implementation to be accurate and effective. ScaleAI is an artificial intelligence company that created Scale Rapid, a product that can utilize both Supervised and Unsupervised Machine Learning (UML/SML) to annotate customer datasets. Scale Rapid, in particular, is being proposed as a means for improving detection and accuracy in MDD diagnosis by detecting large patterns in patient communication (writing, conversations, clinical interviews) over large datasets. The basis for the claim is examining its numerous past applications in medical research and other language based fields, and the adaptability of the product to mental health practices. Furthermore, Scale Rapid can provide bulk annotation of patient data. This may remove the burden from clinicians to manually sort through and process each document, thus increasing their time to provide personalized patient care.
However, a fundamental requirement for a successful implementation of AI into mental health is the data itself. Thus, the conclusion is that ScaleAI has potential to be an effective tool but since it relies on our current flawed databases, more development in data is required. For example, there have been many concerns regarding the security of the databases as well as the privacy of its patients. Numerous hospital database hacks or leaks have resulted in diminished patient trust, causing them to withhold information. This not only impedes the treatment efficacy for the patient, but provides inaccurate data for machine learning. Furthermore, data fed to UML/SML is always subject to human bias, which can see itself re-emerge in the product.

Author

• Justin Liu, New Jersey Scholars Program