Human mate selection has undergone a seismic shift over the last few decades. Historically constrained by physical geography and small local social circles, dating habits have been radically disrupted by modern online dating platforms. Today, algorithmic matching operates as one of humanity’s largest unmonitored behavioral experiments—altering social dynamics with largely unstudied macro-level consequences.
Core hypothesis
Increasing information availability (visualized as an expanding “vision radius”) combined with lowered baseline commitment thresholds inherently destabilizes macro-level pair formation, resulting in high agent volatility, brief relationship durations, and dynamic choice paralysis.
02 / Algorithmic background
Deferred acceptance & dynamic ranking
Modern dating platforms leverage variants of deferred acceptance mechanisms like the Gale-Shapley algorithm combined with Elo-style rating systems (e.g., hyper-personalized preference weights) to rank and route users into matching pools. While these algorithms aim to maximize stable matching in theory, in real-world continuous environments with endless choice influx, they often amplify inequality and churn by continuously exposing agents to higher-ranked alternatives.
03 / Canvas mechanics
How simple heuristics create emergent patterns
This simulation uses an interactive agent-based model (ABM) on an HTML5 Canvas to demonstrate emergent social phenomena from simple cognitive heuristics:
Bounded rationality & heuristics: Particles (agents) evaluate potential partners only when entering proximity. Acceptance is determined by a weighted function of baseline attractiveness, perceived social rank, and a dynamic desperation score (which lowers standards after repeated rejections).
Interaction memory: To avoid infinite evaluation loops without physical collisions, agents retain a short-term memory matrix of past rejections. Re-evaluation occurs only when an agent’s internal desperation crosses a predefined step threshold.
Temptation & infidelity: Coupled agents continuously drift at lower speeds. If a passing single exceeds the partner’s rank beyond a stability margin, commitment decays, triggering a breakup and returning agents to single status.
Visual diagnostics: Live feedback halos indicate agent states—green for mutual match, red for rejection, yellow for cooldown/refractory periods, and dashed red lines for tempted relationship bonds.
Synthetic data potential: Interactions, evaluation outcomes, breakups, and relationship durations are natural measures for a future logging layer; the current canvas visualizes the dynamics in real time.
04 / Proof of concept & future work
From interactive model to Monte Carlo framework
This application serves as an interactive proof of concept (PoC) demonstrating how computational cognitive neuroscience can model complex social behavior without relying solely on traditional, low-TRL laboratory tasks (e.g., EEG, reaction times):
Experimental setup: Running automated Monte Carlo simulations (e.g., N = 1,000 runs across parametric sweeps of Rvision and Ccommitment) allows researchers to apply bootstrapping resampling on output logs to calculate strict confidence intervals for pair stability.
Metrics analyzed: Mean time-to-pair, breakup frequency, inequality of match distribution (Gini index of agent rank), and system volatility.
Future work: Expanding network topology beyond 2D spatial physics to small-world graph networks, integrating dynamic preference drift, and fitting parameters directly against empirical dating app usage datasets.
Live canvas // agent-based model
Change the dating environment
Mutual matchRejectionCooldownAgent
Simulation code (showcase)
Selected excerpts from the canvas model. The parameters below define the default environment; the controls above override vision radius, commitment, temptation, and rejection-driven desperation to compare the two scenarios.
Model configuration
let config = {
INITIAL_PAIR_COUNT: 5, // Starting number of pairs before the population grows.
MAX_POPULATION: 20, // Maximum number of agents allowed in the canvas.
PARTICLE_RADIUS: 15, // Agent body radius in canvas pixels.
PROXIMITY_EVAL_RADIUS: 25, // How close two agents must be to evaluate each other.
BASE_VISION_RADIUS: 50, // Default distance at which agents notice alternatives.
BASE_COMMITMENT: 0.7, // Starting commitment propensity, varied slightly per agent.
TEMPTATION_FACTOR: 0.3, // Scales the breakup risk when a higher-ranked alternative appears.
MAX_SPEED: 0.7, // Movement speed for single agents.
MIN_COUPLE_SPEED: 0.2, // Slowest movement speed for a coupled pair.
SETTLING_RATE: 0.002, // How quickly a pair's movement slows over relationship time.
DESPERATION_GROWTH: 0.015, // Desperation added to an agent after a rejection.
DESPERATION_RETRY_STEP: 0.05, // Desperation increase needed before retrying a past rejection.
COOLDOWN_TICKS: 180, // Frames an agent waits after a breakup before evaluating again.
FPS_TARGET: 50, // Target frame rate used when deciding whether to add another pair.
...userConfig, // Scenario-specific values override these defaults.
};
Recent data highlights an extreme structural asymmetry between offline and online dating environments, creating market mechanics that align directly with our simulation's parameters:
01 / Adoption & preference
Mass market adoption vs. IRL preference
27–39%new couples online
77%Gen Z meet IRL
Online dating has become the primary driver of relationship formation, with 27% to 39% of new relationships/marriages starting online. However, user burnout has triggered a cultural pushback---particularly among Gen Z, where 77% report meeting their partners in real life due to digital fatigue.
02 / Platform demographics
Demographic imbalance
~65–78%men among active users
Popular hookup and swipe-based platforms exhibit a severe gender ratio gap, with men accounting for ~65--78% of active users (e.g., Tinder sits at roughly 67--78% male globally).
03 / Swipe dynamics
The swiping asymmetry
46–53%men swipe right
5–14%women swipe right
Selectivity Gap: Men swipe right on 46--53% of profiles, whereas women swipe right on only 5--14%.
2.6%average man's match rate
30%+average woman's match rate
Match Rate Disparity: The average man's match rate drops to a mere 2.6% (with median men receiving zero to one match per hundred swipes), whereas the average woman's match rate sits above 30%.
04 / Match distribution
Extreme inequality: the “Gini” effect
0.58match-distribution Gini coefficient
According to behavioral data analyses (e.g., SwipeStats dataset across thousands of profiles), the distribution of matches among men yields a Gini coefficient of 0.58---indicating greater inequality than 95% of national economies worldwide.
05 / Ranking & choice
“Swiping up, matching down”
Top 20%most attractive / highest-ranked profiles
Research into preference distributions shows that users across both genders consistently swipe "up" toward the top 20% most attractive/highest-ranked profiles. Because high-ranked profiles have overwhelming choice abundance, they often match with slightly lower-ranked profiles for short-term encounters ("matching down"), further exacerbating relationship instability and rejection rates for the average user.
06 / Industry trends
App burnout & platform decline
Payers ↓consecutive quarterly declines
This structural asymmetry has driven widespread "dating app fatigue". SEC financial filings (10-K/10-Q reports) from industry leaders like Match Group (parent of Tinder, Hinge, OkCupid) reveal consecutive quarterly declines in paying users (Payers). To offset shrinking user bases, platforms increase monetization rates per remaining user---a corporate strategy that inadvertently reinforces churn, user dissatisfaction, and mass app abandonment.
// 06 / SOURCES
Literature & data references
Rosenfeld, M. J., Thomas, R. J., & Hausen, S. (2019).Disintermediating your friends: How contemporary couples meet. Proceedings of the National Academy of Sciences (PNAS), 116(36), 17753-17758. (Establishes the shift of relationship formation from social networks to online platforms).
Pew Research Center (2023).From Dating Apps to Marriage: How Americans Experience Online Dating. (Documented demographic splits, user satisfaction, and app adoption rates).
Tyson, G., et al. (2016).A First Look at User Activity on Tinder. IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM). (Quantified the 10x-15x asymmetry in swipe acceptance and match rates between men and women).
Bruch, E. E., & Newman, M. E. (2018).Aspirational pursuit of mates in online dating markets. Science Advances, 4(8), eaap9815. (Demonstrated that both men and women systematically swipe up toward partners who are on average 25% more desirable than themselves).
Neyt, M., et al. (2021).Long-term consequences of swipe-based dating apps on relationship commitment and mental health. Journal of Social and Personal Relationships. (Connects infinite choice architecture to user burnout and commitment erosion).
Todd, P. M., & Miller, G. F. (1999).From simple decisions to mathematical models of mate choice. (Pioneered heuristic agent-based models of mate search under bounded rationality).
Finkel, E. J., et al. (2012).Online Dating: A Critical Analysis From a Psychological Science Perspective. Psychological Science in the Public Interest. (Analyzes choice overload, evaluation mindsets, and commitment erosion in digital dating).
Gigerenzer, G., & Selten, R. (2002).Bounded Rationality: The Adaptive Toolbox. MIT Press. (Foundational framework for fast and frugal decision heuristics under environmental constraints).
Schelling, T. C. (1971).Dynamic Models of Segregation. Journal of Mathematical Sociology. (Demonstrates how simple micro-level cognitive choices create dramatic macro-level emergent patterns).