Base Rate Fallacy (False Positive Paradox)
When an event is rare, false positives can form a large share of a model’s positive predictions even if its overall accuracy is high. Ignoring that base rate can lead people to overestimate what a positive prediction establishes.
THE IDEA
DOES FACEBOOK KNOW YOU’RE A TERRORIST?
So if corporations with access to Big Data are still pretty limited in what they “know” about you, what’s to worry about?
Try worrying about this. Suppose a team at Facebook decides to develop a method for guessing which of its users are likely to be involved in terrorism against the United States. Mathematically, it’s not so different from the problem of figuring out whether a Netflix user is likely to enjoy Ocean’s Thirteen. Facebook generally knows its users’ real names and locations, so it can use public records to generate a list of Facebook profiles belonging to people who have already been convicted of terroristic crimes or support of terrorist groups. Then the math starts. Do the terrorists tend to make more status updates per day than the general population, or fewer, or on this metric do they look basically the same? Are there words that appear more frequently in their updates? Bands or teams or products they’re unusually prone or disinclined to like? Putting all this stuff together, you can assign to each user a score,* which represents your best estimate for the probability that the user has ties, or will have ties, to terrorist groups. It’s more or less the same thing Target does when they cross-reference your lotion and vitamin purchases to estimate how likely it is that you’re pregnant.
There’s one important difference: pregnancy is very common, while terrorism is very rare. In almost all cases, the estimated probability that a given user would be a terrorist would be very small. The supplied passage ends here without completing the example. Assessing the positive predictions would require the base rate and the model’s false-positive and true-positive rates.
When to use it
Medical testing for rare diseases
Fraud detection in banking
Spam email filtering
Pre-employment screening for rare skills
FROM THEORY TO EVERYDAY LIFE
In practice
Facebook terrorism prediction
Ellenberg imagines a Facebook team trying to predict terrorist ties from user data. Because the target event is rare, even a relatively small false-positive rate could make false alarms outnumber true detections. The story is hypothetical.
Target pregnancy prediction
Ellenberg contrasts the imagined terrorism classifier with pregnancy prediction from purchasing patterns. A higher base rate can change the share of positive predictions that are correct, but the comparison also depends on each model’s error rates.
A MOMENT TO REFLECT
Put it to work.
Consider a scenario where you are evaluating a medical test for a rare disease. How does the base rate affect your interpretation of a positive result?
Think of a time when you made a prediction based on a model. Did you account for the base rate of the event?
How can understanding the base rate fallacy improve your decision-making in business or everyday life?
CONTINUE THE CONVERSATION
Source texts
- How Not to Be Wrong The Power of Mathematical Thinking (Ellenberg, Jordan)