A Case for the Human in Science

When a machine can do nearly everything between a question and its answer, what is left for the researcher to do? This essay argues that the human contribution does not shrink; it concentrates - into judgment: choosing what is worth knowing, recognizing what is worth pursuing, and staking something on the choice. These acts cannot be delegated, because a machine must be told what success means - and there is no telling it what makes a question worth asking. But judgment can go unseen, and today it mostly does. The essay makes the case for a science that learns to see it. Research done with machines leaves, for the first time, a recordable path - the questions, turns, and dead ends behind the finished work - and the path is where judgment shows. What science needs are measures that can read such records and credit the judgment in them: not effort, not polish, but skin in the game. The essay poses this as an open problem to researchers in every field, spells out what any answer must do, and, practicing what it preaches, releases the complete, attributed record of its own making - written in dialogue with an AI - as first raw material. The aim is a science that remains ours: machines doing ever more of the work, and people still choosing where it goes. The record of the essay's making - the full human-AI dialogue, a typed and attributeddecision log, and a machine-readable trace - is openly archived athttps://doi.org/10.5281/zenodo.21217762 (repository:github.com/avseror/a-case-for-the-human-in-science), released under CC-BY-4.0 foranyone who wishes to build on it.