Summary

The Taxicab Problem, presented by Amos Tversky and Daniel Kahneman in 1982, is a classic example in probability theory that illustrates base rate neglect, a cognitive bias where people overlook the actual statistical base rate in favor of specific information.

What does it mean for everyday problem-solving?

The probability of an event occurring must take into account all known variables. There is no such thing as an independent event with an absolute likeliness.

The Problem

In a city, 85% of the taxis are green and 15% are blue. A witness claims they saw a blue taxi involved in a hit-and-run accident. The witness’s ability to correctly identify the color of a taxi is tested, and they are correct 80% of the time, but they also make mistakes 20% of the time.

Question

What is the probability that the taxi involved was actually blue?

Solution

This is a Bayesian probability problem. The key is to combine the base rate of the taxis (green vs. blue) with the reliability of the witness:

  • Base rate (prior probability): 15% for blue taxis.
  • Witness reliability: 80% correct for identifying blue taxis.

Using Bayes’ theorem, the correct answer shows that the probability of the taxi actually being blue is about 41%, not 80%, demonstrating how people often misjudge probabilities by focusing on the witness’s accuracy rather than integrating the base rate.

Here’s the step-by-step calculation using Bayes’ theorem:

1. Definitions:

Let’s define:

  • B: Event that the taxi is blue.
  • G: Event that the taxi is green.
  • W: Event that the witness says the taxi is blue.

We need to find the probability that the taxi is actually blue given that the witness said it was blue, .

2. Bayes’ Theorem Formula:

Source: https://www.geeksforgeeks.org/real-life-applications-of-bayes-theorem

P(B|W)=[P(W|B)∗P(B)]/P(W)

Where:

  • P(B|W) is the probability the taxi is blue given the witness says it’s blue.
  • P(W|B) is the probability the witness says the taxi is blue when it actually is blue.
  • P(B) is the base rate (prior probability) of the taxi being blue.
  • P(W) is the total probability of the witness saying the taxi is blue.

3. Known Values:

  • P(B) = 0.15 (15% of the taxis are blue)
  • P(G) = 0.85 (85% of the taxis are green)
  • P (W|B) = 0.8 (80% accuracy when the taxi is actually blue)
  • P (W|G)= 0.2 (20% error when the taxi is actually green)

4. Compute P(W)

P(W) = Total Probability of Witness Saying “Blue”

Plugging in the values:

P(W)= P(W|B)xP(B) + P(W|G)xP(G)=0.8x0.15 + 0.2x0.85=0.29

5. Apply Bayes’ Theorem

P(B|W)=(0.8x0.15)/0.29=~0.414

Result

The probability that the taxi is actually blue, given the witness said it was blue, is approximately 41.4%.

You might also wanna check out a cool simulation of this problem