Why a Court Can Agree on Every Fact and Still Be Wrong
The Strange Case of the Three Judges
Imagine a court case about a broken contract. The law says the defendant is liable only if two things are true: (1) the contract forbade a certain action, and (2) the defendant did that action. Three judges must decide. Each judge votes on three statements: P (“the contract forbade the action”), Q (“the defendant did the action”), and R (“the defendant is liable”). The legal rule is that R is true exactly when both P and Q are true.
Now look at the votes. Judge 1 says P true, Q true, so R true — that fits the rule. Judge 2 says P true, Q false, so R false — also perfect. Judge 3 says P false, Q true, so R false — again, no mistake. Each judge is completely logical. But what happens when we take a majority vote on each proposition separately? A majority (two out of three) says P is true. A majority says Q is true. Yet a majority says R is false. The group’s verdict becomes: the contract forbade the action, the defendant did it, but the defendant is not liable. That contradicts the law’s own rule. The group result is logically broken even though every individual was consistent.
This is called the doctrinal paradox, first studied in detail by legal scholars Lewis Kornhauser and Lawrence Sager in the 1980s. (The mathematician Siméon Denis Poisson noticed a similar problem way back in 1837, but it only became famous recently.) The paradox shows that when you vote on linked issues one by one, a perfectly sensible group of people can produce a nonsensical group answer.
Why Voting on Each Part Can Break the Whole
The trouble is not just for courtrooms. Suppose you and two friends are planning a movie night. You agree: “We’ll order pizza only if we all watch a film, and we’ll have ice cream only if we order pizza.” You vote on three things: watch a film (yes/no), order pizza (yes/no), get ice cream (yes/no). Each friend votes logically. But majority voting on each question can say “yes to film,” “yes to pizza,” “no to ice cream” — breaking the rule you all agreed to. The group ends up with an impossible plan, even though nobody was illogical alone.
Philosophers and mathematicians study these situations by looking at an agenda — the set of statements a group must decide — and a judgment aggregation rule, which is a method for turning everyone’s individual judgments into one group judgment. They ask: what rules can guarantee that the group’s set of answers is both complete (every question gets a clear yes or no) and consistent (no contradictions)? And what rules are fair?
A fair rule might have these natural-sounding properties. Universal domain says the rule must work for any possible collection of individual opinions. Unanimity preservation says that if everyone agrees on something, the group adopts it. Independence says that the group’s answer on each single question depends only on what people said about that question — not on other questions in the agenda. Majority voting on each question satisfies independence and unanimity, but it can fail consistency, as we just saw.
In 2002, philosophers Christian List and Philip Pettit proved a remarkable impossibility theorem. When the agenda is rich enough — like the court’s agenda or any set with logical connections — any judgment aggregation rule that satisfies universal domain, collective consistency and completeness, independence, and unanimity preservation must be a dictatorship. That means the group’s judgment would simply copy one fixed person’s judgment every time. No fair, democratic rule can both respect each question’s independence and always produce a rational group answer. This result echoes the famous Arrow’s impossibility theorem for ranking candidates, proved by Kenneth Arrow (1921–2017). Arrow showed that no voting system for preferences can satisfy a similar list of reasonable conditions. Judgment aggregation turns out to be even wider: it’s about votes on any logically connected statements, not just rankings.
The Escape Routes: Giving Up Fairness or Logic?
If perfection is impossible, we have to look for escape routes. One way is to give up independence. In the court case, instead of voting on the conclusion R, the judges could vote only on the premises P and Q. Then the conclusion is worked out logically from the majority on those premises. That’s called the premise-based procedure. It guarantees consistency, but it breaks independence because what the group says about R is now forced by what happened on P and Q, not by a direct vote on R.
Another idea is to use sequential priority rules. The group votes on issues in a fixed order, and each decision ties the next one. For instance, first decide P, then Q, and then R must follow the legal rule. Consistency is preserved, but the result depends on the order you chose, which can feel arbitrary.
You could also restrict what kinds of opinions are allowed — relaxing universal domain. If all judges’ views line up in a simple pattern (like everyone who says P true also says Q true), majority voting can stay consistent. But real disagreements rarely line up so neatly. Or you could drop the requirement that the group must have a complete answer on every single proposition; maybe the group can honestly say “we can’t decide this one.” That avoids contradictions but leaves decisions open. Each escape route trades one appealing quality for another. There is no one perfect fix; every real group must choose which principle matters most.
When Computers Try to Merge Beliefs
Computer scientists face a similar puzzle: how to combine information from many different databases or sensors when they disagree. This is called belief merging (or fusion). Imagine three friends deciding on a shared birthday present. Two friends want to buy a book and take the person out to dinner. The third friend doesn’t want either. What is the fairest joint plan?
One method, the majority operator (or minisum rule), tries to minimize the total amount of disagreement. It would choose to buy the book and go to dinner, because that satisfies the majority — even though the third friend is completely unhappy. Another method, an arbitration operator, aims to spread the dissatisfaction evenly. It might select either buying the book or going to dinner, so each person has exactly one complaint. That feels fairer, even though nobody gets everything they wanted.
These methods use a mathematical idea: Hamming distance. It counts the number of atomic statements on which two opinions differ. If one person’s wish list is {book yes, dinner yes} and a possible group plan says {book yes, dinner no}, the distance is 1. The merging operator adds up (or combines in another way) all the distances from each person’s wish to each possible outcome. Then it picks the outcome that is “closest” to everyone overall.
Crucially, merging operators can respect integrity constraints — fixed rules that any acceptable result must follow, like a budget limit or, in the court case, the legal rule that R is true exactly when P and Q are both true. When we apply a minisum merging operator to the three judges under that integrity constraint, inconsistency is avoided. But the result is often not unique: in the court example, three different consistent sets of judgments tie for being closest to the profile. That forces the group to use a tie-breaking rule.
Why Can’t We Have It All?
The tension between fairness and logic shows up everywhere. A student council votes on a new rule that links a dress code and a club budget. A family decides on a vacation only if both parents get time off and the cost is low. A video-game team picks a strategy where certain moves depend on others. In every case, voting on each piece separately can produce a plan that contradicts the very logic that ties the pieces together. Philip Pettit coined the term discursive dilemma for this situation — a dilemma any group faces when it must take a collective stand on multiple, logically linked propositions.
The impossibility theorems are not just about mathematics; they teach us that no voting system can be perfectly fair and perfectly logical at the same time when issues are interconnected. Recently, philosophers have even asked: what if new information arrives after a group has already decided? Can a group update its collective judgment in a rational way, just as a single person would revise their beliefs? That’s the problem of dynamic rationality. Some distance-based methods can help, but a new impossibility result already threatens: under reasonable conditions, no combination of aggregation rule and revision method can guarantee dynamic rationality.
Knowing these limits doesn’t mean giving up. It means choosing procedures wisely. Maybe a group should agree on a logical order of issues before voting. Maybe it should allow a “no opinion” option. Maybe a chairperson must break ties. The puzzle has no single answer, but seeing it clearly makes you a sharper participant in any group decision — and helps you understand why democracy can sometimes deliver results that feel crazy even when everyone was reasonable.
Think about it
- Your class votes separately on three connected rules for a field trip. The combination is logically impossible, but each rule won a majority. Should you change the plan to make it consistent, even if that means ignoring a majority vote on one rule? Why?
- Is it ever fair for one person to make the final decision for a group if that’s the only way to keep the plan logical?
- If you had to design a voting system for a small club where issues are often linked, what trade-offs would you be willing to make — fairness, consistency, or something else? Explain.
