I tend to distrust studies entirely based on models. Modelling is good, but you should use it together with some substantive experimental data.
When LLMs help researchers quickly identify the best projects, as in some technical fields, researchers will become more discriminating about which projects they develop. On the other hand, if LLMs mostly accelerate writing and analysis, as in fieldwork-based disciplines, the fraction of investigations that are published may increase.
Ok, do we have any evidence of which of the two phenomena is actually happening at the moment?
Research has always been full of people publishing a lot of trash. If anything this should lead to changing the incentives for researchers, such as not evaluating them on the publication count. Not really on demonizing LLM use in a research setting.
I tend to distrust studies entirely based on models. Modelling is good, but you should use it together with some substantive experimental data.
Ok, do we have any evidence of which of the two phenomena is actually happening at the moment?
Research has always been full of people publishing a lot of trash. If anything this should lead to changing the incentives for researchers, such as not evaluating them on the publication count. Not really on demonizing LLM use in a research setting.