Glossary · Term

multi-agent debate

← all terms

Definition

Plain language

A setup where several AI agents argue or discuss with each other to reach a better answer.

As stated in the literature

A multi-agent inference pattern in which LLM agents exchange natural-language arguments to refine a shared answer; used as a no-market baseline against economic coordination in Economy of Minds, where it underperformed.

Why it matters: It matters as a coordination approach to compare against, showing where simply having agents argue falls short of better-structured teamwork.

For example, three AI agents might each propose an answer to a tricky question and then argue back and forth until they settle on the strongest one.

Heard on the show

“It's going up against systems that were fine-tuned, or built around multi-agent debate, or trained as dedicated code retrievers.”
Episode 169 — Why Better Bug Reports Can Make AI Coding Agents Worse

Mentioned in 4 episodes

  1. 169
    Why Better Bug Reports Can Make AI Coding Agents Worse
  2. 107
    How a Market of Crippled AI Agents Outscored One Unrestricted Model
  3. 095
    Seven Wins to Zero: How Organizing AI Agents Like a Lab Changes the Search
  4. 012
    Why AI Coding Agents Keep Trying to Debug Without a Debugger

Related terms