Engineering better voting systems

Democracy's fatal conceit is that it assumes aggregating individual preferences will somehow yield collective wisdom. The trouble is, we've spent thousands of years arguing about how exactly to do this aggregation, and the dirty secret of social choice theory is that there's no perfect way to do it.

Our most common voting system—plurality voting, where you cast a single vote for your favorite candidate—is perhaps the worst possible design we could have selected.¹ It's as if aliens visited Earth, observed our voting mechanisms, and concluded: "These humans have deliberately engineered the optimal system for producing dissatisfaction and perverse outcomes."

Consider the current American presidential election system. A voter in California who prefers the Green Party candidate but considers the Democratic candidate acceptable faces a cruel dilemma: vote their conscience and potentially help elect their least-preferred Republican candidate, or strategically vote for their second choice. This is the infamous "spoiler effect," and it traps third parties in a perpetual catch-22: they can't win because people won't vote for them, and people won't vote for them because they can't win.

But as social choice theorists have known since at least the 18th century, we have alternatives.

The Impossibility Theorem

Before diving into solutions, we should acknowledge the elephant in the democratic room: Arrow's Impossibility Theorem. In 1951, Kenneth Arrow proved that no rank-order voting system can simultaneously satisfy all of the following reasonable properties:²

1. Non-dictatorship: The preferences of a single voter shouldn't determine the outcome
2. Pareto efficiency: If every voter prefers candidate A over candidate B, then B shouldn't win
3. Independence of irrelevant alternatives: If a voter prefers A to B, introducing candidate C shouldn't make them suddenly prefer B to A
4. Universal domain: The voting method should work for any possible combination of voter preferences
5. Ordered preferences: Every voter must be able to rank all candidates

This means that every voting system must sacrifice at least one of these desirable properties. We're not looking for perfection here—we're engineering tradeoffs.

Beyond plurality: alternate voting systems

Ranked Choice Voting (Instant Runoff)

The most prominent reform option in the United States today is Ranked Choice Voting (RCV), sometimes called Instant Runoff Voting. Rather than selecting a single candidate, voters rank candidates in order of preference. If no candidate receives a majority of first-choice votes, the candidate with the fewest first-choice votes is eliminated, and those votes transfer to each voter's next-ranked choice. This process continues until someone achieves a majority.³

RCV has been adopted in Maine, Alaska, and various municipalities. Its advantages include:

• Eliminates the spoiler effect, allowing third-party voting without "wasting" your vote
• Ensures winners have majority support (of those who ranked them)
• Tends to elect more moderate candidates who have broad appeal beyond their base

But RCV isn't without flaws. It fails the monotonicity criterion—sometimes, ranking a candidate higher can paradoxically hurt their chances of winning.⁴ It also fails the Independence of Irrelevant Alternatives criterion from Arrow's theorem.

Approval Voting

Approval voting is elegantly simple: voters can vote for (approve of) as many candidates as they wish, and the candidate with the most approvals wins.⁵

The benefits:

• Extremely simple to implement and understand
• Eliminates the spoiler effect
• Passes the favorite betrayal criterion (you're never penalized for voting for your true favorite)
• Can be implemented with existing voting machines

Approval voting has been adopted in St. Louis and Fargo. Its main drawback is that it doesn't capture the strength of voter preferences—approving your favorite and your barely-acceptable compromise count the same.

STAR Voting

STAR (Score Then Automatic Runoff) combines elements of score voting and runoff systems. Voters rate each candidate on a scale (e.g., 0-5), and the two candidates with the highest total scores advance to an instant runoff where the candidate preferred by more voters wins.⁶

STAR voting attempts to thread the needle between expressiveness (allowing voters to indicate preference strength) and majority rule (ensuring the winner is preferred by a majority when comparing the top two candidates).

Condorcet Methods

Condorcet methods determine the winner by conducting a series of virtual head-to-head matchups between all candidates. If a candidate would beat every other candidate in a one-on-one election, they are declared the Condorcet winner.⁷

While mathematically elegant, Condorcet methods face two problems: 1. They're difficult to explain to the average voter 2. Sometimes no Condorcet winner exists (we get "rock-paper-scissors" preference cycles)

Various Condorcet methods (Schulze, Ranked Pairs, etc.) offer different solutions to handle these cycles.

Beyond Mechanism: Breaking the Two-Party Duopoly

The engineering of voting systems extends beyond the mechanism used on election day. Other structural reforms matter just as much:

Proportional Representation

Most democracies use some form of proportional representation, where parties receive seats proportional to their share of the vote.⁸ This allows multiple parties to coexist and form coalitions, rather than forcing a two-party system.

Multi-Member Districts

Instead of electing a single representative per district, multi-member districts elect several at once, often using proportional methods. This reduces wasted votes and improves representation for political minorities.⁹

Algorithmic Redistricting

Gerrymandering—the manipulation of district boundaries for partisan advantage—undermines democratic legitimacy. Algorithms could draw district boundaries without partisan bias,¹⁰ though the parameters fed into these algorithms would still involve value judgments.

The Meta Problem: Coordination

The fundamental issue with voting reform isn't technical—it's a coordination problem. Existing political parties benefit from the status quo and resist changes that might undermine their duopoly. Voters, rationally ignorant about the mechanics of voting systems, don't demand change.

This creates a chicken-and-egg problem: we need political will to change the system, but the system itself prevents the formation of that political will.

The solution? Start locally. Voting reforms have gained traction in cities and states across America. As voters experience better systems at the local level, demand for reform grows upward.¹¹

The Information Problem

A voting system is fundamentally an information aggregation system. Its quality depends not just on its mechanism, but on the quality of inputs it receives—that is, how informed voters are.

Robin Hanson has proposed Futarchy, where we "vote on values, but bet on beliefs."¹² The core insight is that voting is good for expressing what we want, but markets are better at figuring out how to get there. Prediction markets could supplement democratic systems by providing better information about the likely effects of policies.

Conclusion: Engineering for Robustness

When engineering physical systems like bridges or airplanes, we don't just optimize for performance under ideal conditions—we design for robustness under stress and failure modes. The same principle should apply to voting systems.

Our current plurality system fails catastrophically under predictable conditions (when third parties gain significant support). A well-engineered voting system should gracefully handle edge cases and be resistant to strategic manipulation.

No voting system is perfect. But some are so obviously terrible that continuing to use them represents a failure of collective rationality. The question isn't whether we can design a perfect voting system—we can't. The question is whether we can summon the collective will to implement the vastly better systems we already know how to build.

Democracy isn't just about counting votes—it's about making those votes count. And that requires better engineering.

---

Note: When considering voting systems, we should remember Goodhart's Law: "When a measure becomes a target, it ceases to be a good measure."¹³ Any voting system we design will inevitably be gamed by strategic actors. The goal isn't to create an ungameable system, but one where the incentives align with producing outcomes that genuinely reflect the public's preferences.