Computational topic modeling, and human bias

Submitted by Peter (United States), Jul 19, 2022 at 16:20

Interesting premise for a book (I'm currently considering working with a publisher on one where I've done the same kind of thing to analyze narratives in Russian-language social media). You hit the nail on the head though with your assessment of "Bleich and van der Veen's more profound problem". It is a mistake (and a conceit) ever to think that computers can supplant the "human in the loop" in this type of analysis. As CIA writer Richards Heuer wrote in "The Psychology of Intelligence Analysis" all human analysts have bias, and the best we can do is look critically at blind spots in our analysis. Sounds like these authors failed to recognize one very obvious blind spot, perhaps precisely because they were dazzled by their own computational analysis.

Note: Opinions expressed in comments are those of the authors alone and not necessarily those of Daniel Pipes. Original writing only, please. Comments are screened and in some cases edited before posting. Reasoned disagreement is welcome but not comments that are scurrilous, off-topic, commercial, disparaging religions, or otherwise inappropriate. For complete regulations, see the "Guidelines for Reader Comments".