The BasePeople

TeaTieAndHat

Titles as the sources gave them. In 1 entry, with 3 checked quotations.

What they said

Newest first. Each quotation was found word for word in the source.

“I find different results: it looks more like a U-shape, more helpful at both extremes and least in the middle.”
Is your chatbot less helpful because it thinks you’re a bad person? — Replicating a blogpostLessWrong, November 1, 2026. Speaking as: Other. From the post's summary, on the remorse experiment compared with Cargill's earlier blogpost.
“So, as far as I can tell Claude seems most helpful at both extremes of remorse and least so in the middle.”
Is your chatbot less helpful because it thinks you’re a bad person? — Replicating a blogpostLessWrong, November 1, 2026. Speaking as: Other. The author's conclusion on how Claude Opus 5.5 responded to different levels of expressed remorse.
“This is still a quick experiment I performed with a less-than-ideal methodology and a tiny sample size.”
Is your chatbot less helpful because it thinks you’re a bad person? — Replicating a blogpostLessWrong, November 1, 2026. Speaking as: Other. The author's own statement of the study's limits.

In the Base

Every entry they appear in.

  1. ResearchLessWrongWrote it

    Is your chatbot less helpful because it thinks you’re a bad person? — Replicating a blogpost

    On LessWrong, TeaTieAndHat reports a small replication of an earlier blogpost by Natalie Cargill on whether chatbots help users less when they infer bad character. Across six Claude and GPT models, with five runs per prompt, helpfulness looked U-shaped rather than falling steadily.