# A Study on the Performance of Generative Pre-trained Transformer (GPT) in Simulating Depressed Individuals on the Standardized Depressive Symptom Scale

**Authors:** Sijin Cai, Nanfeng Zhang, Jiaying Zhu, Yanjie Liu, Yongjin Zhou  
**arXiv:** [2307.08576](https://arxiv.org/abs/2307.08576)  
**Date:** 2023-07-17  
**Summary page:** https://www.summarizepaper.com/en/paper/2307.08576v1/

## Key points

- Study focused on evaluating GPT technology in simulating depressed individuals on standardized depressive symptom scales
- Aimed to address challenges in diagnosing depression using current self-report and assessment scale methods
- Three depression-related assessment tools used: HAMD-17, SDS, and GDS-15
- Two experiments conducted to simulate GPT responses from normal individuals and those with depression
- Results showed promising performance of GPT in assessing depression, aligning well with scoring criteria for both groups
- Performance variations observed based on severity of depression, better on scales with higher sensitivity
- GPT demonstrated potential for developing more effective depression scales
- Further research warranted to explore leveraging GPT technology for enhancing diagnostic accuracy and supporting mental health care efforts

## Summary

In a study on the performance of Generative Pre-trained Transformer (GPT) technology in simulating depressed individuals on standardized depressive symptom scales, researchers aimed to address the challenges in diagnosing depression using current self-report and assessment scale methods. The study evaluated GPT's potential in accurately diagnosing depression by assessing its ability to simulate individuals with depression and examining the impact of different depression scales. Three depression-related assessment tools, including HAMD-17, SDS, and GDS-15, were utilized in the study. Two experiments were conducted to simulate GPT responses from both normal individuals and those with depression. The researchers compared GPT's responses with expected results, evaluated its understanding of depressive symptoms, and analyzed performance differences under various conditions. The results indicated that GPT demonstrated promising performance in assessing depression. It aligned well with scoring criteria for both individuals with depression and normal individuals. However, some performance variations were observed based on the severity of depression. Notably, GPT performed better on scales with higher sensitivity. In conclusion, the study found that GPT accurately simulated individuals with depression and normal individuals during assessments related to depression. While deviations occurred when simulating different degrees of depression, particularly limiting understanding of mild and moderate cases, GPT showed potential for developing more effective depression scales. This suggests that GPT technology has significant promise in aiding clinicians and patients in the assessment of depression. Further research is warranted to explore how this technology can be leveraged to enhance diagnostic accuracy and support mental health care efforts effectively. References
1. Thapar A., et al., Depression in young people. Lancet 2022; 400(10352): p. 617-631. 2. Takeda M., Tanaka T., Depression in the elderly. Geriatrics & Gerontology International 2010; 10(4): p. 277-279. 3. Greenberg P.E., et al., DEPRESSION - A NEGLECTED MAJOR ILLNESS. Journal of Clinical Psychiatry 1993; 54(11): p. 419-424. 4. Goldman L.S., ...