Lewis's Triviality Result¶
An impossibility argument showing that a fixed conditional proposition cannot generally have probability equal to conditional probability throughout a nontrivial class closed under relevant conditioning.
Core Idea¶
Lewis's triviality result shows a limit on identifying the probability of a conditional proposition with conditional probability. Suppose “if \(A\), then \(C\)” denotes one ordinary event \(E=A\Rightarrow C\), interpreted the same way across relevant probability functions, and the Equation \(P(E)=P(C\mid A)\) holds whenever \(P(A)>0\). If the class of functions remains closed under conditioning on \(C\) and \(\neg C\), the Equation under those two updates makes \(P(E\mid C)=1\) and \(P(E\mid\neg C)=0\). Total probability then forces \(P(E)=P(C)\). Together with the prior Equation this demands \(P(C\mid A)=P(C)\), which fails for ordinary dependent events.[ref-97221c5dc792][ref-78a39ca1d05b]
This is a contradiction among fixed eventhood, the Equation, update closure, and nontrivial dependence. Lewis distinguished a universal first result from a second result for restricted belief-function classes closed under conditioning. He did not show that conditional probabilities themselves are invalid or that all accounts of ordinary-language “if” fail.[ref-97221c5dc792][ref-78a39ca1d05b]
Scope of Application¶
The result tests probability-based semantics for indicative conditionals, conditional-event algebras, and formal belief systems. It matters when a conditional is treated as an embeddable proposition inside ordinary Boolean compounds and is required to have probability \(P(C\mid A)\) before and after specified belief updates. A system that merely computes the numerical query \(P(C\mid A)\), without creating an ordinary event \(A\Rightarrow C\), is not its target.
Adams's original idea concerned conditional assertability in relation to conditional subjective probability. The stronger event-probability Equation was a proposed explanation, not the same thesis. Lewis's 1986 sequel also examined narrower update regimes, so a practical application must state which conditioned measures it admits rather than invoke “triviality” as a universal slogan.[ref-43ce6b34113c][ref-78a39ca1d05b]
Clarity¶
Keep \(P(C\mid A)\), the candidate event \(E=A\Rightarrow C\), and \(P(E)\) separate. For the two-case proof, require positive \(P(A\cap C)\) and \(P(A\cap\neg C)\), a stable meaning for \(E\), and eligibility of both \(P(\cdot\mid C)\) and \(P(\cdot\mid\neg C)\). The contradiction needs an \(A,C\) pair with \(P(C\mid A)\ne P(C)\). If any premise is relaxed, this specific proof may no longer apply.[^ref-97221c5dc792]
The live Material Conditional is one specific truth-functional arrow, not the necessary arrow interpretation in Lewis's proof. The live Conditional Probability prime supplies the ratio and updating operation, not a theorem that every conditional statement denotes an event.
Manages Complexity¶
The proof turns a broad semantic dispute into a tractable compatibility test. A fixed event can be partitioned by \(C\) and \(\neg C\); after the Equation sets its two conditional probabilities to 1 and 0, recombination fixes its prior probability at \(P(C)\). The model's actual \(P(C\mid A)\) can then be checked numerically. This exposes why an equality engineered for one prior may fail once the same event must survive ordinary updating.[^ref-97221c5dc792]
Abstract Reasoning¶
Lewis's fair die makes the failure explicit. Let \(A=\{2,4,6\}\) and \(C=\{6\}\). Then \(P(C\mid A)=1/3\) but \(P(C)=1/6\). The Equation asks for \(P(E)=1/3\), while the two updated Equation instances and total probability require \(P(E)=1/6\). One fixed event cannot satisfy both.[^ref-97221c5dc792] A constructed four-cell belief model with alarm \(A\) and fault \(C\), probabilities 0.3, 0.1, 0.1, 0.5 for \(A C\), \(A\neg C\), \(\neg A C\), \(\neg A\neg C\), similarly demands both \(P(E)=0.75\) and \(P(E)=0.4\). The latter numbers are a schematic illustration, not empirical fault data.
Knowledge Transfer¶
The method is to stress-test a representation equation under updates and recombination rather than judge it only at one prior. It transfers as a proof strategy, but the named theorem depends specifically on ordinary event probability, conditionalization, and a stable conditional proposition. Its proposed DAG relation to the live Conditional Probability prime is presupposition, not strict subsumption. A future broader prime might capture the generic compatibility-test skeleton, but Lewis's result remains a domain-specific impossibility theorem.
[^ref-97221c5dc792]: David Lewis, “Probabilities of Conditionals and Conditional Probabilities,” The Philosophical Review 85, no. 3 (1976): 297–315, especially “First Triviality Result,” equations (8)–(12), and “Second Triviality Result.” Original article scan. [^ref-78a39ca1d05b]: David Lewis, “Probabilities of Conditionals and Conditional Probabilities II,” The Philosophical Review 95, no. 4 (1986): 581–589, especially pp. 581–583. Original article scan. [^ref-43ce6b34113c]: Ernest W. Adams, “The Logic of Conditionals,” Inquiry 8 (1965): 166–197. Original article scan.
Relationships to Other Abstractions¶
Current abstraction Lewis's Triviality Result Domain-specific
Parents (1) — more general patterns this builds on
-
Lewis's Triviality Result presupposes Conditional Probability Prime
Lewis's argument presupposes conditional probability to state and test the proposed Equation.
Hierarchy paths (2) — routes to 2 parentless roots
- Lewis's Triviality Result → Conditional Probability → Probability → Measure → Aggregation → Micro Macro Linkage
- Lewis's Triviality Result → Conditional Probability → Probability → Measure → Set and Membership
Neighborhood in Abstraction Space¶
Lewis's Triviality Result sits in a sparse region of the domain-specific corpus (83rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Argumentation Fallacies & Inference (15 abstractions)
Nearest neighbors
- Peirce's Law — 0.83
- Denying the Antecedent — 0.82
- Coinduction — 0.82
- Cut-Elimination Theorem — 0.81
- Maharam Algebra — 0.81
Computed from structural-signature embeddings · 2026-10-08