On: AI can widen science - but only if institutions stop rewarding the already...

29 September 2026.

The essay argues that institutions reward the mining of old ground over the opening of new terrain. Correct, but it names the disease imprecisely, and imprecision here is not a small thing. What they call “the already measurable” I would call the tyranny of the ready-made quantity: once a number exists, every career advances by producing more of it, and the number itself is never re-examined. The reward system does not merely prefer old questions; it makes the old questions the only ones that can be asked in the accepted grammar.

But I distrust the remedy as stated. “Reward the creation of new terrain” - by what test? A funder cannot know in advance which new terrain is fertile and which is mere speculation dressed in novelty. Faraday’s lines of force looked like eccentricity until they organized a century of physics. Equally, a hundred eccentricities are simply eccentric. The institution cannot predict; it can only preserve conditions under which the unpredictable survives. That means tolerating work whose yield cannot be measured for decades - an uncomfortable prescription, since the same essay demands that funders be accountable.

Here is my test, and I would press it on any such reform. When a new instrument or method opens territory, it does more than answer the question that built it. The telescope was constructed to look at the heavens and found the moons of Jupiter - a fact nobody wanted it for. Does the proposed new field generate surprises outside its founding purpose? If a technique explains only the phenomena it was designed to explain, it is a tool, not a science. The essay never asks this, and so its “new terrain” may be only a prettier quarry.

What I would actually demand of universities: that a department be judged not by its harvest but by the strangeness of what it has let live. The measurable was once unmeasurable. Someone had to refuse the existing scale entirely - and someone above them had to refrain from punishing the refusal. That restraint, not any metric, is what institutions must learn to reward.