The “synthetic validity” problem in AI-generated science - PNAS

We’re sharing this article early to provide faster access to peer-reviewed, accepted research. It is citable and carries a permanent DOI. This version is subject to further edits and will be replaced automatically by the final Version of Record. All legal disclaimers apply. AbstractTo investigate the validity and test–retest reliability of the One-Minute Sit-to-Stand Test (1STS) in individuals with Parkinson’s disease (PD). Twenty-eight individuals with PD performed the 1STS twice with a 7-day interval. Concurrent validity was examined using correlations with the 2-Minute Walk Test (2MWT) and the Hoehn and Yahr (H&Y) stage. Test–retest reliability was assessed using the intraclass correlation coefficient (ICC). The 1STS showed a strong positive correlation with the 2MWT (r = 0.69, p < 0.01) and a moderate negative correlation with the H&Y stage (r = − 0.58, p < 0.01). Test–retest reliability was excellent (ICC = 0.94, 95% CI: 0.86–0.97). The 1STS is a reliable and potentially valid measure of functional performance in individuals with PD. It may be considered a practical option in clinical settings; however, further research is needed to confirm these findings.

This writer argues that psychology, like other social sciences, faces rampant cultural disparagement as a discipline of science. As a third year psychology student, she shares insights of her realization that science isn’t exactly as objective as it’s understood. Better understanding that nuance though, she argues, could actually douse the flames of scientific skepticism. As I sit in my psychology lectures at Drake University for the third year, I’m noticing a theme in Drake’s Bachelor of Arts in Psychology curriculum — an emphasis on the validity of psychology as a science.

Google DeepMind researchers who developed the method say that this will help uphold scientific integrity and prevent misuse of protein designs Google DeepMind has developed a way to watermark proteins designed by artificial intelligence (AI) during the design process itself, which could help scientists better track the origin of these structures.

The University of Waterloo is developing a Synthetic Data Network (SDN) as part of a new cross-Canada consortium with institutions from Ontario, Quebec and New Brunswick. “Canada’s health-care system produces an immense amount of data with the potential to drive better decisions and better outcomes,” said Dr. Helen Chen from the School of Public Health Sciences. “However, vital privacy requirements can make it difficult to share and use person-level data at scale. This limits collaboration and leaves important opportunities for discovery untapped.”