Abstract
Depression is considered one of the leading causes of disability and ill health. People with depression tend to interpret everyday events through a negatively biased lens, which contributes to symptomatology. Cognitive bias modification (CBM) is designed to reduce negative bias through reading scenarios common to depression but enforcing positive interpretations to them. CBM offers low-cost delivery and greater accessibility for combating depression. Despite this, the development of CBM is onerous and costly, with an estimated cost in the hundreds of thousands. Here, we aimed to automate CBM’s development to lower cost by validating artificial intelligence (AI)-generated CBM scenarios against human-generated scenarios.
People with current or previous experience of depression (N = 13) created 55 items, which included scenarios and interpretations, based on their everyday accounts; CoPilot, a free AI-powered chatbot, was instructed to generate 55 items common to depression. Next, 30 participants with experience of depression (Mean years of depression = 11) were blindly presented all 110 items (which included both AI-generated and human-generated items) in randomised order. Using a 7-point scale, participants rated all items on readability, commonality of experience to depression, and degree of negativity-to-positivity for each biased interpretation.
To investigate whether participants’ ratings of AI-generated and human-generated items were equal, a series of Two One-Sided Test and Paired-Sample t-test showed statistical and practical equivalence of all rating criteria between human-generated and AI-generated items, all p <0.001. Results suggest that CoPilot may be capable of generating scenarios that are reflective of human experiences of depression.
Our findings demonstrated that CoPilot has potential to generate scenarios akin to human experiences of depression, with consideration to sample size as a limitation of generalisation. The implications of our findings entertain the possibility of significantly lowering cost and time in developing CBM to enable greater proportions of society to access evidenced-based depression treatment.