Debate Rekindled Over Decision Theory When Agents Simulate Predictors

LessWrong users revisited the classic "Agent Simulates Predictor" problem, debating how agents should reason when they can simulate a weaker predictor. Richard Ngo questioned whether existing decision theory frameworks apply to agents with vastly different intelligence, while jimrandomh pointed to prior FDT discussions of superintelligent cooperation.

2026-10-03 ~ 2026-10-05 · 4 related posts

• jimrandomh: FDT literature already covers superintelligence cooperation across time-slices — jimrandomh · 2026-10-03
• Richard Ngo: existing frameworks don't apply to agents with vastly different intelligence — RichardMCNgo · 2026-10-03
• The Agent Simulates Predictor problem: should one-boxing survive a weaker Omega? — jessi_cata · 2026-10-05
• Why FDT is hard: Newcomb variants where the agent simulates a weaker predictor — jessi_cata · 2026-10-05