Statistical prediction of parole success outperforms subjective judicial decisions

Evidence

The results were stunning. Models with random weights—as long as they were constrained to be positive—were as accurate or more than the judges themselves. Models with equal weights were even better. — Brian Christian, The Alignment Problem, 441 (#corroborates, via linked claim)

Inductive generalization: Also known as induction by enumeration, this is the process of arriving at general or universal propositions from the particular facts of experience, relying on the principle of induction. The premises are instances of certain facts with accompanying legal consequences which repeatedly accompany one another (legal rules) from which it is concluded that the legal consequence will always accompany the facts (legal principle). — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 546 (#qualifies, via linked claim)

A proper analogy should identify the number of respects in which the compared cases, or fact scenarios, resemble one another (let us call these resemblances positive analogies) and the number of respects in which they differ (negative analogies). In analogy, unlike the method of enumeration, the quantity of cases is not significant. Instead, what is important is relevancy—whether the compared facts resemble, or differ from, one another in relevant respects. — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 343 (#qualifies, via linked claim)

Categorization

• corroborated: 2 independent passages support this claim
• XREF: connects to: Science only recently began predicting parolee behavior after release; Inductive generalization derives universal legal rules from repeated particular case outcomes — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 546; Legal outcomes depend on variable choices and value judgments rather than absolutes. — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 752 (+5 more)

Related

• Parole prediction is possible only as aggregate statistics, not certainty (similarity 0.83)
• Science only recently began predicting parolee behavior after release (similarity 0.83)
• Even randomly weighted linear models beat human expert judgments (similarity 0.82)

Broader Connections

• Even randomly weighted linear models beat human expert judgments #corroborates (0.95) — Brian Christian, The Alignment Problem, 441
• Inductive generalization derives universal legal rules from repeated particular case outcomes #qualifies (0.45) — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 546
• Valid legal analogies depend on relevant similarities rather than case quantities #qualifies (0.40) — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 343
• Legal outcomes depend on variable choices and value judgments rather than absolutes. #qualifies (0.35) — Ruggero J. Aldisert, Agatha D. Aldisert, Logic for Lawyers- A Guide to Clear…, 752
• Stated precedent rules often fail to determine later case outcomes #exemplifies (0.30) — Burton, Steven J, An Introduction to Law and Legal…, 216