AI Faces Challenges in Complex Social Science Research Tasks
"most social science research tasks" does a hell of a lot of lifting here. Which tasks? Because I can't see AI doing interviews, in depth observations of group dynamics, ethnographic research, work in phenomenology, anything that requires immersion in an experience is out.
Political philosophy, normative ethics and any field in which a summary of - say - a moral framework is not sufficient, and we must choose one it's also not something easily coded (if at all). Can't do symbolic interactionism, because AI has no idea of what context is.
It's going to be a hoot coding reflexive analysis into a model, and just as much of a party to convince political dissidents or marginalized communities to trust AI and talk to it. So, yeah, I'm not really sure most tasks in the social sciences are covered.
I mean, it can do intro HR or structured interviews, but I'm not sure I'd trust it to get the context and subtle nuance of high-stakes, cultural-fit ones. Anyways, obviously things will change in research production. I just think we need to be very mindful of how.