Two people produce text of identical quality. One spent six hours on it, the other ten minutes with help. Should they be assessed the same way? Our instinct says no, and yet we would struggle to explain exactly what we are rewarding: the result, or the effort?
We have covered the practical consequences of AI at school and in recruitment. Here is the underlying question that runs through both.
What we actually reward
Our intuition about merit mixes two distinct things, and this confusion went unnoticed as long as they moved together.
The result. What is produced, its quality, its usefulness. That is what matters when you buy an object: you do not care how long it took.
The effort. What the person invested, the difficulty overcome. That is what a medal, a grade, a merit-based promotion rewards.
For a long time, these two dimensions were correlated: a good result generally presupposed effort. You could reward one while believing you were rewarding the other, without having to choose.
AI decouples these two things. And we are discovering that we did not know which one we valued.
The uncomfortable argument
There is a serious objection to the very idea of merit, and it predates AI. A large part of what produces a result already does not stem from merit: background, upbringing, health, luck, natural disposition. Nobody deserves to have had parents who read to them.
From that standpoint, AI does not introduce a new injustice: it adds a factor to an already long list of things that explain outcomes without involving personal virtue. One could even argue that it reduces certain inequalities, by giving those who lacked the right codes access to what others inherited from their environment. We observed this for accessibility and for writing job applications.
The argument holds. Yet it is not enough, for one precise reason.
We developed this in our article on saved time: some efforts produce only their result, others also transform the person doing them. Writing an essay is not about producing an essay. When you grade an assignment, you are not rewarding the text: you are checking that a transformation has taken place. If the text arrives without the transformation, it is not cheating on the result — it is that the result no longer proves anything.
Three possible responses
Assess differently. If the finished product no longer proves competence, you have to check competence directly: oral examination, practical scenarios, the ability to defend and critique one's own work. This is what is emerging in schools, and it requires more teacher time, which is the real obstacle.
Own up to valuing the result. In many professional contexts, this is already the case and it is legitimate. A client wants a problem solved, not proof of suffering. Acknowledging this frankly is better than maintaining a fiction about effort.
Shift merit towards what remains scarce. If producing becomes easy, what distinguishes people lies elsewhere: knowing what to produce, judging whether it is good, deciding under uncertainty, taking responsibility. These are exactly the criteria we identified in our article on jobs.
The risk to watch
One drift deserves to be named. If only the result counts and the tool is available to everyone, assessment shifts towards increasingly fine differences: who has the best tool, who knows how to use it better, who has time to iterate more.
Those differences are no more meritorious than the previous ones, and they risk overlapping with existing inequalities. A tool that equalises access to production can very easily recreate a hierarchy based on access to quality.
What to take away
This question has no single answer, and the fact that it is being asked is rather good news. We operated with a comfortable notion of merit because it was never examined.
What seems solid: in learning contexts, effort is the very object, and production without effort proves nothing. In production contexts, the result matters, and there is no shame in using what exists.
The problem comes from the fact that we conflate the two. A degree claims to certify competence and is judged on productions. A job interview claims to assess a person and is judged on a CV. These mechanisms rested on the assumption that the result testified to the rest. That assumption no longer holds, and that is what needs to be rebuilt, rather than looking for who cheated.