Overview
- A study by McGill University and Mila AI Institute used a large language model to analyze over 4,000 clinician reports on autism diagnosis.
- Findings suggest that repetitive behaviors, special interests, and sensory differences are more indicative of autism than social skills deficits, which are emphasized in current DSM-5 criteria.
- The study introduces AI tools to help clinicians identify sentence-level factors most relevant to a diagnosis, aiming to reduce subjectivity in the process.
- Researchers hope the findings will encourage revisions to autism diagnostic guidelines and inspire broader applications in mental health diagnostics.
- Limitations of the study include a lack of demographic analysis and geographical diversity, which may affect its applicability across populations.