@ziegler_md
I think it would be hard to teach an algorithm what a “perfect” paper is, especially if it deviates from the composite of all past published papers. Maybe more feasible to start with AI for a small component of review, like determining adequate reporting of statistics?