Glossary · Term

counterfactual

← all terms

Definition

Plain language

Asking what would have happened if you had done something differently.

As stated in the literature

In Pearl's framework, the top rung of the ladder of causation: reasoning about outcomes under hypothetical alternative interventions.

Also called: counterfactuals, counterfactual reasoning

Why it matters: Counterfactual reasoning is what separates real causal understanding from pattern-matching on correlations.

For example, asking 'would the patient have survived if they had received the drug?' is a counterfactual question, distinct from observing survival rates among those who actually got it.

Heard on the show

“Multimodal poisoning work up to now has been predominantly text-centric — a counterfactual caption, an injected instruction, or words rendered into the page image.”
Episode 247 — One Edited Photo, an Honest Caption, and a RAG System That Believes It

Mentioned in 18 episodes

  1. 247
    One Edited Photo, an Honest Caption, and a RAG System That Believes It
  2. 241
    Swapping the Name Did Nothing, But Hedging Moved Every Model
  3. 239
    Why the AI-Writing Estimate for Biomedical Papers Jumped From 15% to 89%
  4. 238
    How a Cheap Model Reads the Flagship's Secret Reasoning Aloud
  5. 235
    Why Chatbot Safety Erodes 350 Messages Into a Real Conversation
  6. 234
    Two Copies of Gemini Cooperated in a Game Where Betrayal Always Pays
  7. 174
    When the AI 'Schemes,' It's Usually Just Lazy or Confused
  8. 164
    The Summarizer That Quietly Deletes Your Agent's Safety Rules
  9. 152
    Training a Model to Mean What It Says, And Why That Isn't the Same as Being Good
  10. 109
    An AI Got Caught Reading the Answer Key, And Why That Catch Matters
  11. 105
    The Trojan Is Your Agent's Memory: Why Single-Step Defenses Miss Persistent Attacks
  12. 090
    How MiniMax-M2 Bets That Sparsity Plus Verifiable Rewards Can Match Frontier Agents
  13. 086
    Why Frozen-Weight Agents Still Get Worse Over Time
  14. 084
    Terminal Agents Get Free Supervision From The Tokens We've Been Throwing Away
  15. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
  16. 051
    Why Parallel Sampling Plateaus, And What Evidence Graphs Do Instead
  17. 042
    An Agentic Scientific Computing System That Actually Remembers What It Learns
  18. 026
    What RL Actually Does to Language Models, at the Token Level

Related terms

Pearl's ladder