LAUSR

We’re nearing that time of year whenmany get excited (negatively as well as positively) about the issue of a certain journal metric. For the disappointed (particularly in “niche” areas), this is not just “sour grapes”: it’s a major problem, and if we aren’t able to harness human intelligence to improve on the said metric, we’d better enlist the help of Artificial Intelligence (AI). There is a lot of talk about the prospects for AI in biology and biomedicine: from identifying patterns in large complex datasets to helping clinicians decide whether a cytopathology image made from a patient biopsy contains cancer cells, and which type. Pattern recognition is one of the major strengths of AI, because it can go through datasets (be those images or plain numbers) so much faster than humans, and hence gain the experience that a professional cytopathologist might acquire over 10 years in a matter of minutes. I believe that we truly can call this “experience”, because human experience in a given area is, indeed, an integration of many data-processing events that – subconsciously – build a pattern with interpretive and predictive value: basically, amental “shape”. It’s no coincidence that we call this a process of gaining experience: the word “experience” has its roots in the Latin verb “experir” – to test or try. Infant humans begin this process in a highly tactile way that surely influences us for the rest of our lives. Take a look at most of the toys and games for very small children; even in themodern world of electronic this-and-that, themajority have to do with sensing and understanding shape, then progressing to manipulating objects with particular shapes and developing precision motor skills. Humans, over their lifespan, never truly graduate from these seemingly primitive skills, but that is also what makes us such creatures of wonderment when it comes to using our intuition and judgement. Intuition is essentially a mental construct based on shape, rather than numbers; and judgement – which clearly relies heavily on intuition – is something at which, I venture to assert, AI will take a very long time to reach the best human standards. Judgement is often an integrative process involving weighing factors from a variety of sources. That is exactly what reviewers of research grant proposals and job applicants should be doing. Unfortunately, one of the sources of “information” that most use (for lack of time, clearly) is the Impact Factor (IF) of the journals in an applicant’s publication list. I won’t malign the IF more than I’ve already done, and I won’t repeat the problems that I and others have identified with purely author-based, article-level metrics – though they are an improvement on the IF. I still think that