I spent last week at PolMeth, presenting a poster on my paper “Stance Is Not a Construct” — an argument about what we’re actually measuring when we ask language models to detect stance. The feedback was generous and remarkably consistent: sharpen the conceptualization of the attitude construct and add more studies to the experiments. That’s now the heart of the revision I’ll present at APSA in Boston this September.
The best parts happened between sessions: conversations about time series, dynamic networks, and — of course — LLMs; the IPPSR session on open congressional data (my thanks to IPPSR for supporting my travel); and dinners where senior scholars recalled what PolMeth was like for grad students in their day. One dinner line stuck with me: as social scientists, our job is counting the social world. That’s more or less why my newsletter is called The Social Computist.
Now: revising the stance paper for APSA, moving a collaborative project on congressional speeches toward its first draft, and preparing my network paper — how do networks respond to shocks? — for journal submission.