AI has much promise for research wrt qualitative coding, potentially taking a process that provides a rich source of data but is expensive and time-intensive, and making it more affordable and timelier. BUT we have to keep humans in the loop (and ensure it's humans that we're actually studying).
Can large language models stand in for human participants? Many social scientists seem to think so, and are already using "silicon samples" in research. One problem: depending on the analytic decisions made, you can basically get these samples to show any effect you want. THREAD 🧵