Stories

AI is helping researchers understand mental illness

Toby Talbot / AP

Machine learning, AI, computer games and virtual reality are helping researchers study and better understand psychiatric disorders, particularly borderline personality disorder (BPD), per MIT Tech Review. Sarah Fineberg and colleagues at Yale University are using a computer game to understand feelings of social rejection in people with BPD.

Past difficulties in studying psychiatry include a reliance on subjective observations, thus making it difficult to objectively understand the unique human behavior expressed by people with mental disorders. And the cause of BPD is still unknown — researchers understand that various social, genetic and environmental factors contribute to it, but pinning it down on a large scale hasn't been easy.

That's where machine learning and AI come in.

Virtual reality has helped researchers study behavior by creating a digital environment with controlled conditions.

  • In one case study, participants with and without borderline personality disorder (BPD) controlled their own avatar in the VR world.
  • They were then asked to talk about the avatar they interacted with in the digital environment.
  • One group claimed the other avatar was being controlled by the participant's "romantic partner."
  • By using VR, researchers could observe how key details about interpersonal behavior — how closely they approached someone, the direction of their gaze in relation to the other avatar, and even posture.

A computer game in which three participants pass a ball back and forth helped researchers understand how those with BPD recognize and process social rejection.

  • The game, called Cyberball, allows the researchers to control the other two subjects in the game and how frequently they pass the ball to the subject with BPD, but the participant doesn't know that.
  • In cases where the subject passed the ball to one of the computer-controlled players who then passed it back and forth between themselves, participants felt "sadness and anger in as few as six rounds of play," Finberg told MIT Tech Review.
  • One fascinating finding: Those with BPD felt less sad when they received the ball more times than the other player, but they still felt sad.

Limitations: Some researchers argue that applying computation to psychiatry and neuroscience broadly presents challenges, particularly with data and conclusions being inexact and inconsistent. Last year, researchers in the Oxford Academic's Brain: Journal of Neurology illustrated how applying computer science statistical methods to neuroscience supported "several profoundly different conceptualizations of how hallucinations may arise." Though these two approaches are not the exact same, there are various arguments for how computational intervention in psychiatry presents unique challenges to studying psychiatry.

Bottom line: Using tech to analyze psychiatry has allowed researchers to better understand a range of behaviors exhibited by those with and without BPD, which they hope will ultimately lead to the discovery of what causes this disorder. Recruiting participants remains a difficult task, but as computational psychiatry becomes more prevalent that could change.