A tiny probability per pair may become large at system level when the number of possible pairs grows quadratically.
The Birthday Paradox feels surprising because people often focus on 23 individuals, whereas the relevant quantity is the 253 possible pairs. The case is primarily a compact demonstration of aggregate risk and scale.
A tiny probability per pair may become large at system level when the number of possible pairs grows quadratically.
Under the standard model of independent, uniformly distributed birthdays over 365 days, excluding leap day, 23 people have about a 50.73% chance of at least one shared birthday. There are 23 × 22 / 2 = 253 pairs, but pair-match events are not mutually independent. The exact probability is 1 − ∏(1 − k/365), for k = 0 through 22.
The probability is elementary and established; a single historical origin is unclear. Cryptographic birthday bounds are a mature application.
The number of pairwise opportunities grows as n(n−1)/2. Aggregate probability may therefore rise rapidly while the per-pair probability remains small.
AI transfer to investigate: AI does not create the combinatorics, but it can increase the number of generated artefacts very quickly.
Proposed test, not yet conducted: Ask participants to estimate collision risk using only per-pair probability versus explicit pair counts or simulation. Measure calibration.
The birthday paradox helps explain how the number of combinations affects probability. Applications to AI require separate research.
The standard model uses independent, uniformly distributed birthdays over 365 days, excluding leap day. With 23 people there are 253 pairs; pair-match events are not mutually independent.
Diaconis and Mosteller (1989, §7.1, pp. 857–858) describe the probability model. The stated standard probability and pair count follow from that model. This does not validate the proposed teaching intervention or AI application.
Help improve this case
We review your contribution before updating the text or sources.