The Mating Ritual: Stable Matching in Algorithmic Game Theory

Irving's algorithm solves the stable roommates problem by following three phases: (1) using a modified stable marriage approach where each person proposes to their next preferred roommate, accepting proposals only if they improve the current match; (2) removing all rejected proposals from preference lists; and (3) generating P and Q lists to detect cycles and reject unstable pairings. The algorithm runs in quadratic time complexity (O(n²)) and will find a stable matching if one exists, or determine that no solution is possible when someone is universally rejected.

Irving's stable roommate matching algorithm solves the problem of pairing an even number of people into roommates such that no two unmatched people prefer each other over their current matches; the algorithm consists of two main phases: a proposal-rejection phase similar to the Gale-Shapley algorithm where individuals propose to their most preferred partners and accept/reject based on preference rankings, followed by a cycle detection and diagonal removal phase that eliminates unstable pairings by identifying cycles in the preference table and removing conflicting matches, ultimately producing a stable matching or determining that no stable matching exists.

Unlike the stable marriage problem with two distinct groups, the stable roommates problem (where everyone ranks others for rooming assignments) does not always have a solution. This is demonstrated by constructing specific examples where no stable configuration exists, showing that the heteronormative assumption of separate male and female groups is essential for guaranteed solutions.

The stable roommates problem extends stable matching to single populations where everyone must be paired with someone else. Unlike stable marriage, this problem may have no solution at all—no stable pairing might exist. The solution requires two phases: first applying a modified stable marriage algorithm, then decycling the results to remove cycles that prevent stable pairings. The decycling phase identifies chains of people who can rotate around while maintaining stability, removing these cycles to achieve a valid solution when one exists.

This comprehensive lecture covers two fundamental problems in matching theory. The stable marriage problem involves finding stable matchings between two sets of people with preference lists, where a matching is stable if no blocking pair exists. Stable matchings always exist and can be found using the Gale-Shapley algorithm. The set of all stable matchings forms a lattice structure with maximum and minimum elements. To find all stable matchings, one can make artificial rejections and run the algorithm repeatedly. The stable roommates problem generalizes this to a single set of people who must be paired up, where stable matchings may not exist. Irving's algorithm (1985) solves this through two phases: Phase 1 runs Gale-Shapley-style proposals, and Phase 2 handles failures by finding cycles and performing artificial rejections. The algorithm maintains invariants about preference lists and terminates when a stable matching is found or proven impossible.

Many-to-one merge allows multiple observations in the master dataset to match with a single observation in the using dataset. This is useful when applying the same treatment or value from the using dataset to multiple master observations based on a key variable. The video demonstrates merging Section 1 (master) with Section 3 (using) using 'gender' as the key variable, where male observations receive Treatment 1 and female observations receive Treatment 2. The merge indicator shows zero non-matching observations when all master observations find matches in the using dataset.

MatchPy is a Python library for pattern matching that enables efficient symbolic computations by matching nested term structures with placeholders, supporting features like sequence wildcards, associativity, and commutativity; its key innovation is many-to-one matching, which exploits similarities between patterns to achieve 5-20x speedup over individual one-to-one matching, making it particularly valuable for applications like linear algebra optimization and abstract syntax tree manipulation.

Many-to-one matching is a generalization of one-to-one matching where one side (such as companies, research labs, or hospitals) can accept multiple participants while the other side (such as students or workers) can only be assigned to one entity. The key difference is the introduction of 'capacity' (CJ), which represents the maximum number of participants each entity can accept. A matching is stable if it is individually rational (no participant would prefer to leave) and no blocking pair exists (no student prefers a lab over their current assignment while that lab has capacity and would prefer this student over current members). The Gale-Shapley algorithm can be adapted to find stable matchings in many-to-one settings by having students propose to their preferred labs and labs tentatively accepting up to their capacity limit.

A many-to-one function is a function where multiple elements from the domain can map to the same element in the codomain. This is a valid function type because each element from the domain still has exactly one image, even though different elements from the domain may share the same image in the codomain. This differs from one-to-one functions where each codomain element has at most one pre-image.

A function f is many-to-one (not injective) if there exist x1 ≠ x2 such that f(x1) = f(x2). This means at least two different elements in the domain map to the same element in the codomain.

School choice is a market design problem where parents submit ranked preferences and schools have fixed capacity. When over-demand exists, applicants are ranked by government priorities (neighborhood, economic disadvantage). Research shows how rejected applicants are reassigned significantly impacts what parents apply for. The Boston mechanism causes systematic misrepresentation—parents apply only where they have priority. The Gale-Shapley mechanism, while incentivizing truth-telling, assigns families to neighborhood schools regardless of preferences when failing schools exist. Schools are not fixed objects; assigned students determine quality, and mechanisms affect segregation. The Boston mechanism without neighborhood priorities leads to school segregation through coordination processes. When private schools are incorporated, families who can afford them play more aggressively, gaining better public school access—creating a novel inequality source where wealth buys better public school access. School choice must be embedded into the broader school market reality.

Alvin Roth, a Nobel Prize winner in economics, pioneered market design through his work on matching theory, allocation theory, and market design. Markets are human inventions that help coordinate and cooperate, but they don't always work optimally. Commodity markets like the New York Stock Exchange use prices alone to allocate resources anonymously. Matching markets, like universities and labor markets, involve selection processes where both parties must be mutually chosen. Most markets are hybrid—combining price mechanisms with matching considerations. School choice systems aim to help children avoid poor schools and attend good ones, addressing intergenerational poverty from residential segregation. Traditional systems faced major problems: congestion where 30,000 students weren't assigned despite enough seats; strategic behavior where students couldn't reveal true preferences because schools only considered first-choice applicants; and schools concealing capacities to keep options for last-minute decisions. The deferred acceptance algorithm (developed by Gale and Shapley) solves these problems by having students privately submit preferences. Each student applies to their first choice, and schools tentatively assign seats in priority order. Rejected students apply to their next choice, and schools reassign all applicants in priority order. Acceptances are deferred until no more applications come in. This makes it safe to reveal true preferences because your chance of getting your second choice if rejected from your first is the same as if you had listed it first. The algorithm produces stable matchings where no unmatched student-school pair would both prefer each other. In New York City, implementing this algorithm solved congestion—only 3,000 students weren't matched (vs. 30,000 before). Over four years, more students got their first choice, first or second choice, and first through fourth choices each year. Schools learned the algorithm worked well and put more places back into the centralized system, improving outcomes.

Market design is economic engineering—the application of economic principles, particularly incentives and market mechanisms, to redesign market institutions. Unlike traditional economics that analyzes existing markets, market design actively builds markets from scratch, combining theory with practical implementation. Key concepts include strategy proofness (incentivizing truthful preference revelation), market thickness (ensuring sufficient participation), and flexibility (adapting to changing policy goals). Impossibility theorems reveal that perfect mechanisms satisfying all properties don't exist, forcing trade-offs. In school choice applications, students (or parents) are strategic agents submitting preferences over schools, while schools have fixed priority rankings over students. Each student has unit demand, and each school has capacity constraints. Priority structures typically include factors like sibling attendance, walk zone residence, and diversity considerations. Key design goals include individual rationality, elimination of justified envy (stability), strategy proofness, Pareto efficiency, and respect for unambiguous priority improvements. A fundamental impossibility result states that no mechanism can simultaneously achieve Pareto efficiency, strategy proofness, and stability. The deferred acceptance algorithm (student optimal stable mechanism) works iteratively: students apply to top-choice schools, schools tentatively hold highest-priority applicants up to capacity, rejected students apply to next choices, and this continues until convergence. The resulting matching is stable, individually rational, and strategy proof.

Market design is an economic field where economists apply theoretical models to real-world problems to improve markets. School choice systems are a key application, where allocation mechanisms assign students to schools without using prices. The field has successfully matched doctors to hospitals, students to universities, and kidney patients to donors. School choice systems require careful design because they involve strategic behavior, congestion, and the need to balance individual preferences with societal goals like equity and diversity.