Lecture Notes: OkCupid's Matching Algorithm

Introduction

• Speaker: Christian Rudder, founder of OkCupid
• Main Topic: Explanation of OkCupid's matching algorithm

Importance of Algorithms

• Definition: A step-by-step way to solve a problem
• Importance: Essential for computers, which excel at step-by-step processes
• Ubiquity: Algorithms are everywhere due to the prevalence of computers

OkCupid's Matching Algorithm

• Objective: Facilitate better matches for users on a dating site
• Foundational Math: Addition, multiplication, square roots
• Challenge: Translating human attraction into computable components

Data Collection

• Method: Ask users questions to collect data
• Example Questions:
  • "Do you want to have kids?"
  • "How often do you brush your teeth?"
  • "Do you like scary movies?"
  • "Do you believe in God?"
• Matching Like with Like: People with similar answers have better matches

Enhanced Data Collection

• Issue: Some questions require different pairings
• Solution:
  • Ask users their own answer
  • Ask what they want in a match's answer
  • Rate the importance of the question

Importance of Questions

• Weighting: Not all questions are equally important
• Example: Politics vs. movie preferences
• Rating Scale:
  • Irrelevant: 0
  • A little important: 1
  • Somewhat important: 10
  • Very important: 50
  • Absolutely mandatory: 250

Algorithm Process

• Steps:
  1. Collect data through user answers
  2. Compare user preferences using math
• Example Calculation:
  • Two questions: Cleanliness and attention-seeking
  • Numerical assignment for importance
  • Satisfaction assessments
  • Calculation of match percentage using geometric mean

Results Interpretation

• Geometric Mean: Combines wide-ranging values
• Intuitive Understanding: Mutual affection is reflected in higher satisfaction percentages
• Correction Factors: For small data sets, adjust for margin of error

Conclusion

• Skill Importance: Translating real-world phenomena for computers is crucial
• Metaphor: Sentences tell stories to people; algorithms tell stories to computers
• Encouragement: Learn the language of algorithms to tell your own stories.