LLM councils show groupthink

Rohit Krishnan · 15 Jun 2026 · strangeloopcanon.com ↗

3 korrents from this piece

Rohit Krishnan did not write this page.

Every claim below was made in this piece, quoted word for word and numbered in the order the piece makes them, so you can read it there rather than take our word for it. The sentence above each quote is our reading of the claim, not their wording. Each quote was checked against a stored copy of the page at build time; where the two differ, the quote is the fact.

Our reading
Group committees tend to flatten distinctive or unconventional viewpoints into blander, more conventional output.

Their wordsWith people, the problem with committees is that they "smooth out" all idiosyncrasies. They take out any "spiky" points of view, and make things much more normie.
Our reading
Combining multiple AI models' outputs into one final answer tends to discard most of the good ideas that only one model produced.

Their wordsthe council does not simply keep the best bits from everyone. It keeps a minority of the good ideas, while peer review seems to give consensus ideas an extra push.
Our reading
Giving LLMs explicit instructions, rather than assuming alignment, prevents emergent problems in multi-model systems.

Their wordsthe best way with LLMs usually is to be explicit, since otherwise even if they're aligned they cause emergent problems.