Optimal action under ideal assumptions can be useless under real conditions
Evidence
Economists since the mid-twentieth century have discussed the “theory of the second best,” which argues, in effect, that knowing the right thing to do in a theoretical version of the economy, which obeys a number of mathematical assumptions, may have virtually no bearing in an economy that deviates from those assumptions even slightly. The “second-best” policy to follow, or action to take, may have virtually no resemblance to the best. — Brian Christian, The Alignment Problem, 1033
Categorization
- corroborated: 2 independent passages support this claim — Brian Christian, The Alignment Problem, 1033
- XREF: connects to: Design choices optimizing for one task can cripple performance on the real task — Brian Christian, The Alignment Problem, 1221; An agent should plan around its own future errors, not ideal execution — Brian Christian, The Alignment Problem, 1031; Choice of action should hinge on what will actually happen, not ideal possibilities — Brian Christian, The Alignment Problem, 1018 (+16 more)
Related
- An agent should plan around its own future errors, not ideal execution (similarity 0.82)
Broader Connections
- Choice of action should hinge on what will actually happen, not ideal possibilities
#corroborates(0.80) — Brian Christian, The Alignment Problem, 1018 - Convergent correlational evidence should justify action absent perfect scientific certainty
#corroborates(0.70) — Harriet A. Washington, A Terrible Thing to Waste, 50 - The precautionary principle should override demands for exhaustive proof before acting
#corroborates(0.70) — Harriet A. Washington, A Terrible Thing to Waste, 1032 - Deterrence works only when fighters can reliably predict battle outcomes
#qualifies(0.70) - Judicial rules should not be applied to dryly logical extremes
#expands(0.60) — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 144 - Inductive legal generalization produces workable probabilities rather than absolute truths.
#corroborates(0.50) — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 339 - A belief is an accepted assertion lacking absolute proof but prompting action
#expands(0.50) — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 239