What Online Dating Taught Me about Algorithms
summary
Begins with a 1727 newspaper ad and traces 1964 St. James and 1965 Operation Match which mailed lists after encoding questionnaires onto punch cards run at 2–4am for $3, showing proto-algorithms that survive today. Describes implementation steps: data encoding, weight attribution, k‑nearest neighbor similarity via Euclidean distance, and output generation; shows a worked example giving a 94% match using the geometric mean. Covers OKCupid factors (match%, unread messages, attractiveness, distance, random “just in case”), Tinder’s ELO‑style score with a 0.4 decay, and Hinge’s Gale–Shapley matching, with mixed real‑world effectiveness.
AI-written summary · about
chapters (17)
- 0:00Dating app algorithms
- 1:281700s newspaper dating ads
- 2:111964 computer dating service
- 3:36Operation Match’s punch cards
- 6:44Encoding user data in algorithms
- 7:50Compatibility with Euclidean distance
- 9:15Early pair matching systems
- 10:20Modern apps like OKCupid
- 13:27OKCupid’s match percentage
- 15:37Unread messages and attractiveness
- 17:00Attractiveness bias experiments
- 18:24Tinder’s ELO score system
- 20:53ELO score adjustments
- 22:38Hinge’s Gale-Shapley algorithm
- 24:04From Operation Match to Hinge
- 24:48Lessons on algorithms and math
- 25:10Workshop and closing
related talks
28mWhat Online Dating Taught Me About Algorithms
35mDating Algorithms: What Are They Really Optimizing?
31mHow to Find Your Other Half or Why Algorithms Matter
22mCommunicating Algorithmic Process in Online Behavioral Advertising
47mAlgorithmic bias is everywhere (especially at Breeze) - what can we do about it?
34mAlgorithms to live
43mKeynote: Consequences of an Insightful Algorithm
21mEvan Estola, Lead Machine Learning Engineer, Meetup, NYC 2017