I think that the lottery discussion is somewhat loose from the AI discussion. In my view there are good arguments for the lottery, but AI is not one of them. In fact, a badly-set-up lottery with insufficient quality check might act as an incentive to flood the system with AI proposals (which may even be of poor quality) because someday one will be lucky if one tries often enough.
As for the lottery, one issue in the Netherlands is that, given that it's a small country, one cannot have a different call and committee for all small fields. This means that it can happen that a proposal is graded by a committee in which nobody is familiar with the topic and can really judge novelty and feasibility and the probability of interesting outcomes. Asking external reviewers (experts) does not fully resolve this if the committee does not understand the reviewers' reports and the rebuttals. So the current system is a lottery in the sense that if you're lucky, there is someone in the commission who understands enough of the topic to get enthusiastic about your proposal, and pushes it, or can see through an overly negative review.
In my view, a lottery should, however, still weed out bad proposals.
I expect that it is much less time-consuming to weed out those proposals that are not worthy of being funded (infeasible, not novel, or simply written so confusingly that one cannot follow it...), than to rank those which are. In my limited experience in committees, there were some obviously weak proposals, and quite a few of the kind "I'd like to fund that, but maybe some other proposal will get even a few more points". It is in my view much harder to rank the "worthy" proposals, which all deserve funding, than to kick out the not worthy ones.
So my suggestion: Let there be a vigorous quality check to get rid of obviously weak proposals, and then hold a lottery over the rest.
Maybe the quality check will also help to weed out the worse "AI slop".
Dit heeft ook te maken met het verwijderen van de numerus fixus in Nijmegen voor Biologie (wij hebben natuurlijk ook studenten die van ver komen die nu dichterbij kunnen studeren als ze willen) en al helemaal voor Biomedische Wetenschappen aan de VU volgens mij. Daar zijn ze gegroeid met +400%. Dus ja het is interessant, maar als we realistisch zijn is de kans groot dat vele studenten bij Biologie normaal BMW willen doen of dus in Nijmegen gaan studeren. En dit is maar 1 opleiding, maar toch zie je hoeveel invloed een numerus fixus kan hebben op keuzegedrag en dus aanmeldingen of niet in deze situatie.
Zoals te lezen is bovenaan de column, is dit een echte chat met ChatGPT. Als auteur van de column probeer ik te laten zien hoe ChatGPT met mij als gebruiker 'interacteert'.
Did someone do a test and check with the same method (Pangram) some older speeches dating from before AI? I know that some German guy did this for fictional writing. The tool they used (not sure if it was Pangram) "detected" quite a lot of AI in some old stories.
Maybe AI detection tools are better in detecting "hot air" phrasing (which existed before AI already, though maybe less frequent) than AI-generated text itself.
I think that the lottery discussion is somewhat loose from the AI discussion. In my view there are good arguments for the lottery, but AI is not one of them. In fact, a badly-set-up lottery with insufficient quality check might act as an incentive to flood the system with AI proposals (which may even be of poor quality) because someday one will be lucky if one tries often enough.
As for the lottery, one issue in the Netherlands is that, given that it's a small country, one cannot have a different call and committee for all small fields. This means that it can happen that a proposal is graded by a committee in which nobody is familiar with the topic and can really judge novelty and feasibility and the probability of interesting outcomes. Asking external reviewers (experts) does not fully resolve this if the committee does not understand the reviewers' reports and the rebuttals. So the current system is a lottery in the sense that if you're lucky, there is someone in the commission who understands enough of the topic to get enthusiastic about your proposal, and pushes it, or can see through an overly negative review.
In my view, a lottery should, however, still weed out bad proposals.
I expect that it is much less time-consuming to weed out those proposals that are not worthy of being funded (infeasible, not novel, or simply written so confusingly that one cannot follow it...), than to rank those which are. In my limited experience in committees, there were some obviously weak proposals, and quite a few of the kind "I'd like to fund that, but maybe some other proposal will get even a few more points". It is in my view much harder to rank the "worthy" proposals, which all deserve funding, than to kick out the not worthy ones.
So my suggestion: Let there be a vigorous quality check to get rid of obviously weak proposals, and then hold a lottery over the rest.
Maybe the quality check will also help to weed out the worse "AI slop".