Step 7: assistant

Calls place_argument_passage

{
  "reading": "premise_unstated",
  "argument_id": "aabfd608-54bf-4a9e-a901-87a3c94c8108",
  "instance_id": "e797a766-3c17-4d8f-8abf-be8ac776cf2b"
}

Calls place_argument_passage

{
  "reading": "premise_unstated",
  "argument_id": "920417c3-d150-4426-bfe8-9fb9add56d4c",
  "instance_id": "bd792ba7-3b51-4560-b2a6-6add146a939b"
}

Calls record_argument_instance

{
  "url": "https://www.lesswrong.com/posts/pH3eKEAEupx8c2ep9/update-on-ought-s-experiments-on-factored-evaluation-of",
  "title": "Update on Ought's experiments on factored evaluation of arguments",
  "stance": "advances",
  "context": "Summary of Ought's human experiments on factored evaluation of arguments about reading-comprehension questions.",
  "reading": "premise_unstated",
  "argument_id": "26705e2a-1bae-4ef2-af33-3eb5852f67b2",
  "source_date": "2020-01",
  "source_kind": "blog_post",
  "verbatim_text": "However, experiment participants disagreed a lot about whether steps were valid or invalid."
}

Calls set_claim_importance

{
  "claim_id": "a4bfe1a1-0d33-4876-b6ab-40989cb38bac",
  "reasoning": "A supporting premise inside the scalable-oversight debate (whether amplification-style training is supervision-efficient); consequential locally, moderately disputed, and largely overtaken by pretraining-based methods.",
  "importance": 0.35,
  "contestation": 0.4
}