retrodiction
Explaining or reconstructing past events after they have occurred, as opposed to predicting future ones; Cowen says macroeconomics is good at this but not at prediction.
As defined in Patrick Collison
Where it comes up · 3
- Patrick Collison2 claims
- 158 | David Wallace on the Arrow of Time1 claim
- AMA | May 20251 claim
Most important claims · 4 of 4Search all claims for “retrodiction” →
- Although retrodicting should intuitively get harder the further back in time you go, the Past Hypothesis means your ignorance about the state of affairs actually decreases the further back you go, because entropy goes down substantially—an apparent tension worth explaining.AMA | May 2025
- What gives macroscopic processes their arrow of time is not the esoteric microscopic symmetry violations (like certain particle decays) but irreversibility (many initial states funneling into one final state, losing information) and noise (de facto unpredictability of which future state is reached); these wash out information, which is why we can predict but not retrodict macroscopic processes like a melting ice cube.158 | David Wallace on the Arrow of Time
- On important macroeconomic issues, economics will not become predictive any time soon—not in our lifetimes—because we cannot rerun alternate histories and better statistics alone will never settle key disputes; there is plenty of micro (e.g., Uber's datasets) where prediction works well.Patrick Collison
- Economics has advanced in retrodiction—we can now trace where bad mortgages were made and which banks held them in the Great Recession with detail unavailable for the 1979-82 recession—but it has made no progress in controlling or predicting business cycles, crashes, or recessions, and the best theories imply such events are intrinsically difficult to predict.Patrick Collison