Mike Thompson-Brusstar

@mthompsonbrusstar.com

history and bureaucracy in the sinosphere | mthompsonbrusstar.com

what methods are people using for studying agenda setting/taking in text data lately? vectors + cosine similarity seems … not right. topic prevalence seems fraught and also maybe conceptually distinct. bracketing llm-coded pairs for now. surely someone smarter than me has thought about this?

Read the proposed justification and rebut it too, if you can. IANAL, but my read of the proposal is that they argue that altering/removing these programs would do is reduce “administrative burden,” which of course is an abuse of the term as they are talking about ED, not applicants.

Tim Longman 🏳️‍🌈@timlongman.bsky.social · 4w ago

The Department of Education has now officially submitted proposals to eliminate International Education and Fulbright Hays programs, which have funded thousands of students and scholars to study languages and conduct research abroad. I encourage you to make public comments and contact your reps.

Things get interesting when we ask about the likelihood of Japan coming to aid Taiwan in the event of a conflict. Here we see a somwhat more familiar patern of partisan breakdown, but it is not nearly as polarizing as you might think! KMT&TPP are largely split on the question!

Bild

New working paper With Torben Johansen and @ChristianVedel It's on a methodological issue that we think is crucial for the discipline of economic history: Whenever we use machine learning tools in our workflow, they can introduce structural bias. So, what should we do?

ML tool performance can decline when applied to archaic text

My question here is: while many of us paying attention to ML (but, speaking for myself, not “of ML”), have been wrong directionally, investors have not. They pour more and more money into the research and development of these tools, it seems.

Ted Underwood@tedunderwood.com · last mo.

Yeah, it's not that a small group of people were wrong. Most of us have been wrong -- repeatedly, in the same way. In that context, doubling down on 2020-era skepticism is ... definitely a choice.