When Groups Have to Think as One

Imagine you and two friends are on a school committee that has to decide whether a student is guilty of breaking a rule. The rule says: a student is guilty if and only if they were in a restricted area and they were caught on camera there. Three things to vote on:

Were they in the restricted area? (call this p)
Were they caught on camera? (call this q)
Are they guilty? (call this r)

You vote Yes on p, Yes on q, and Yes on r — they’re guilty.

Your first friend votes Yes on p, No on q, and No on r — not guilty.

Your second friend votes No on p, Yes on q, and No on r — not guilty.

Now you take a vote on each question separately. On p, it’s 2–1 Yes. On q, it’s 2–1 Yes. On r, it’s 1–2 No. So the committee’s official position is: Yes, they were in the restricted area. Yes, they were on camera. But No, they’re not guilty.

That’s impossible. If both conditions are true, the student has to be guilty — that’s the rule you all agreed on. Each of you individually voted in a way that made sense. But when you put your votes together by majority on each question, the group ended up believing something self-contradictory. This is called the doctrinal paradox.

Philosophers and computer scientists have been trying to figure out what to do about this problem for the last few decades. It turns out that it’s not just about courts — it happens whenever a group of people needs to form a single position on several interconnected questions. And it’s surprising how hard it is to avoid.

The Basic Problem: Aggregating Judgments

Let’s think about what just happened more carefully. Each person in the committee had a complete, consistent set of judgments. That means each person’s answers made sense together and covered all the questions. But when you simply took a majority vote on each separate question, the group’s answers stopped making sense.

One obvious solution: don’t vote on each question separately. Instead, you could vote only on the “premises” (was the student in the restricted area? were they on camera?) and then let the conclusion (guilty or not) be forced by the rule. That would give you: Yes on both premises, so guilty. Or you could skip the premises entirely and just vote on the conclusion. That would give you: Not guilty.

Notice something strange? These two procedures can give opposite answers. The same group of people, using the same information, can reach different outcomes depending on how they vote. That’s a big problem if you care about fairness and consistency.

Philosophers Christian List and Philip Pettit showed that this paradox is just the tip of the iceberg. They proved something disturbing: for any method of combining people’s judgments on logically connected questions, if the method satisfies a few reasonable conditions, then it must be a dictatorship. One person’s views determine everything, and everyone else’s input is ignored.

The conditions they used are worth understanding, because they seem fair on their own:

Universal domain: the method has to work for any possible set of individual judgments, as long as each person is consistent.
Collective rationality: the group’s outcome has to be consistent too.
Independence: the group’s judgment on each question should depend only on what people think about that question, not on what they think about other questions.
Unanimity preservation: if everyone agrees on something, the group should agree too.

The trouble is that these four conditions together force a dictatorship. And the doctrinal paradox is just a special case: majority voting satisfies independence and unanimity, but it can produce inconsistent results, violating collective rationality.

Another Approach: Merging Beliefs

Computer scientists who work on artificial intelligence ran into a related problem. Imagine you have several robots exploring a building. Each robot has sensors that give it beliefs about where things are. Robot A thinks the door is open and the hallway is clear. Robot B thinks the door is closed but the hallway is clear. Robot C thinks the door is open but the hallway is blocked. They need to combine their information into one consistent picture so they can navigate.

This is called belief merging. It’s different from judgment aggregation in an important way: there’s no fixed list of questions to vote on. Each robot just has a set of beliefs, and they might express those beliefs in different ways. One robot might say “the door is open and the hallway is clear,” while another might say “the door is open, the hallway is clear” as two separate statements — but the content is the same.

The key idea in belief merging is to find a collective position that’s as close as possible to everyone’s individual positions, while also satisfying any constraints that have to be met. In the court example, the legal rule — guilty if and only if both conditions are true — is a constraint. In the robot example, physical laws might be constraints.

To find this “closest” position, computer scientists use distances. The most common distance is the Hamming distance: count how many individual propositions two positions disagree on. If one person thinks p is true and q is true, and another thinks p is false and q is true, the Hamming distance between them is 1 (they disagree on p).

Here’s the clever part: you can define what “close to the group” means in different ways. One way is to add up all the distances from a candidate outcome to each person’s position and pick the outcome that minimizes the sum. This is called the minisum rule, and it turns out to be equivalent to majority voting when majority voting gives a consistent result. But when majority voting would give an inconsistency, minisum picks a consistent outcome that’s as close as possible to what everyone thinks.

Another way is to minimize the maximum distance to any individual. This is called the minimax rule, and it tries to make sure no one person is extremely unhappy with the outcome, even if the total dissatisfaction is higher.

Why This Matters for Groups

Let’s return to the three friends on the committee. Suppose you use the minisum rule with Hamming distances. You have three possible consistent outcomes that respect the legal rule:

Outcome A: Yes on p, Yes on q, Yes on r (guilty)
Outcome B: Yes on p, No on q, No on r
Outcome C: No on p, Yes on q, No on r

The Hamming distance from each person to each outcome:

Outcome	Distance to Person 1	Distance to Person 2	Distance to Person 3	Total
A	0	2	2	4
B	2	0	2	4
C	2	2	0	4

All three outcomes have the same total distance! So the minisum rule can’t pick one — it gives a tie. You’d need a tie-breaking rule. But notice: the inconsistent majority outcome (Yes on p, Yes on q, No on r) has a total distance of 3, which is actually smaller. But it’s not allowed because it violates the legal rule. The group is forced to pick among the consistent options, even though each is farther from the individuals’ original positions.

This reveals a real tension: the most “popular” position among the group is contradictory. Any consistent position will leave some people unhappy. And different ways of measuring “unhappy” (sum vs. maximum vs. something else) will produce different outcomes.

Majority vs. Arbitration

The difference between minisum and minimax points to a deeper debate. Should a group aim for the outcome that makes the most people happy overall (majority, or minisum)? Or should it aim for the outcome that is most fair to everyone, even if it means not giving the majority exactly what they want (arbitration, or minimax)?

Philosophers have thought about this in the context of social choice theory, but belief merging makes it concrete and computable. Consider a different example: three friends are deciding what to get a mutual friend for their birthday. Two want to buy a book and take them out to dinner. One doesn’t want to contribute to either. A majority approach would go with both the book and the dinner, making the third person very unhappy. An arbitration approach might pick only one of the two — say, just the book — so that everyone is equally dissatisfied (two people don’t get what they wanted on dinner, one doesn’t get what they didn’t want on the book, but no one is completely overruled).

This is a real, practical question. When schools make policies, when clubs make decisions, when any group has to form a single position on multiple interconnected issues, the choice between majority and fairness matters.

Still an Open Debate

Nobody has found a perfect solution to the doctrinal paradox. Every method for combining judgments on connected questions has some flaw — it can give inconsistent results, or it can be manipulated by strategic voters, or it can be extremely hard to compute, or it doesn’t treat everyone equally.

Some philosophers argue that independence is the wrong requirement to begin with. When questions are logically connected, why should the group’s judgment on one question be independent of what people think about other questions? The premises and conclusion in the court case aren’t independent — they’re linked by a rule. So maybe the group should take the links into account.

Others argue that groups shouldn’t try to have consistent beliefs at all. Maybe a committee can live with the fact that its policies contain contradictions, as long as it does something reasonable in practice.

And some worry about strategic manipulation. If you know how the group is going to combine your judgments, you might be tempted to lie about what you really think in order to steer the outcome your way. For example, if you know the group will use majority voting on each question, you might vote against a premise you actually believe in, just to avoid an outcome you like less.

The study of judgment aggregation and belief merging is still active. Computer scientists build algorithms that can handle these problems for specific cases. Philosophers argue about what it means for a group to “believe” something. And the rest of us — when we serve on committees, vote on school policies, or even plan group projects — face the practical version of the same puzzle.

Key Terms

Term	What it does in this debate
Doctrinal paradox	Shows that majority voting on separate questions can produce a contradictory group position, even when everyone is individually consistent
Judgment aggregation	The problem of combining people’s yes/no answers on a fixed set of logically connected questions
Belief merging	The problem of combining people’s sets of beliefs into one consistent set, without requiring answers to a fixed list of questions
Independence	The rule that a group’s stance on each question should depend only on what people think about that question
Collective rationality	The requirement that the group’s final position must be logically consistent
Hamming distance	A way to measure how much two positions disagree by counting the number of individual items they differ on
Minisum	A merging rule that picks the outcome with the smallest total distance from everyone’s positions (majority-like)
Minimax	A merging rule that picks the outcome that minimizes the largest distance to any individual (fairness-like)
Integrity constraints	Fixed rules that the group’s final position must satisfy (like the legal rule in the court case)

Key People

Christian List and Philip Pettit — Philosophers who formalized the judgment aggregation problem and proved that any procedure satisfying a few reasonable conditions must be a dictatorship
Lewis Kornhauser and Lawrence Sager — Legal scholars who first identified the doctrinal paradox in court decisions
Sébastien Konieczny and Ramón Pino Pérez — Computer scientists who developed the standard framework for belief merging, including distance-based methods
Kenneth Arrow — Economist who proved a famous theorem showing that no fair voting system can avoid paradoxes when there are three or more options

Things to Think About

Is it fair to treat a group’s “belief” as something that can be computed from individuals’ beliefs, or does it make more sense to think of groups as making decisions without having beliefs of their own?
The minisum rule can give ties (multiple outcomes equally good). Is that a problem? How should ties be broken in a fair way?
If you knew that someone on your committee was going to vote strategically, would you do the same? Is it ever okay to vote against what you really think in order to get a better group outcome?
Some philosophers argue that the group should base its decision on the premises alone and let the conclusion follow logically. But what counts as a premise versus a conclusion might depend on how you frame the question. Does that mean the leader who sets the agenda has too much power?

Where This Shows Up

School committees and student councils — Any time a group votes on multiple connected issues (budget, rules, events) that affect each other
Juries — Real courts have to decide on multiple facts that together determine guilt or innocence
Wikipedia and other collaborative editing — Multiple editors have conflicting views on how to present information; there are systems for resolving disagreements
Robot teams and AI systems — Multiple sensors or algorithms need to combine their data into a consistent picture of the world
Social media moderation — Platforms have to decide what content violates rules based on multiple factors (context, intent, harm), and different people see the same content differently