In a perfect democratic world, the collective preference of a group should depend only on the individual preferences of its members. However, in many real-world scenarios—from choosing a three-course meal to passing legislation in Parliament—we don't vote on every possible combination of outcomes at once. Instead, we break the decision down into a sequence of smaller choices.
This decomposition introduces a subtle but powerful vulnerability: the order of the votes can determine the winner. This phenomenon, known as agenda control, suggests that the person who sets the schedule often has more power than the people casting the ballots.
The Combinatorial Explosion Problem
Why do we vote sequentially at all? As Lirong Xia and Jérôme Lang point out in Strongly Decomposable Voting Rules on Multiattribute Domains, voting on multi-attribute outcomes (like a "main dish, side, and drink" combo) leads to a "combinatorial explosion." If there are only 10 options for each of the three categories, a voter would have to rank 1,000 possible meals.
To make this cognitively manageable, we vote on the main dish first, then the side, then the drink. But this sequence creates dependencies. If you win your preferred main dish (steak), your preference for the side might change (from rice to potatoes). Because voters have these conditional preferences, the final outcome becomes tethered to the path taken through the decision tree.
The Power of the Agenda Setter
When decisions are made one after another, the "agenda setter"—the chairperson or committee that decides the order of business—gains strategic leverage. In How Hard Is It to Control Sequential Elections via the Agenda?, Vincent Conitzer, Jérôme Lang, and Lirong Xia demonstrate that by simply reordering the issues, a chairperson can often engineer a specific desired outcome.
This isn't just a theoretical concern for multi-attribute choices; it applies to any elimination-style process. Jérôme Lang, Maria Silvia Pini, and their colleagues explored this in the context of sequential majority voting, such as knock-out tournaments. They found that an agenda setter can often "rig" a tournament by carefully selecting the bracket, ensuring their favorite candidate faces easy opponents early on while rivals eliminate each other in a different part of the bracket.
Parliamentary Procedures as Sequential Games
Legislative bodies use specific sequential protocols to manage complex bills. Robert Bredereck, Jiehua Chen, and Toby Walsh have analyzed these in Parliamentary Voting Procedures. They highlight two common patterns:
- The Amendment Procedure: Alternatives are pitted against each other in pairs. The winner of the first pair faces the third alternative, and so on.
- The Successive Procedure: Alternatives are voted on one by one. The first to receive a majority is the winner, and the process ends immediately.
In both cases, the "early" or "late" position of a bill in the sequence is critical. In a successive procedure, being first is a massive advantage; in an amendment procedure, being last (the "challenge" position) is often better. This sensitivity allows for strategic manipulation, though Jiehua Chen and Rolf Niedermeier later noted in their 2017 follow-up that while "hacking" these elections is computationally easy, real-world data suggests these systems are surprisingly robust against such manipulation in practice.
Designing for Fairness and Order Independence
If the order of voting is the problem, can we design rules where the order doesn't matter? Lirong Xia and Jérôme Lang introduced the concept of order-independent sequential composition. These are specific voting rules designed to produce the same result regardless of the sequence in which attributes are decided. This effectively strips the agenda setter of their power.
Other researchers focus on "bottom-up" sequential processes to protect minority interests. Egor Ianovski and Aleksei Y. Kondratev investigated the proportional veto core, where voters simultaneously "consume" or eliminate their least-preferred options. By focusing on what is least liked rather than what is most liked, this sequential elimination ensures the winner is the candidate who is the "last to be hated" by any significant group.
Similarly, Tristan Cazenave and Ulysse Pavloff explored sequential elimination games, where voters take turns removing candidates. While these protocols reduce the cognitive load on voters, they reinforce the central theme of this research: in any system where choices are made one by one, the structure of the sequence is just as important as the votes themselves.