Break the loop

1 · Read the main result

Beginner

What did the simulations find about the most accurate recommender?

It mixed incomes the most It left people the most segregated by income It kept every restaurant visited

2 · Compute a segregation score

Experienced segregation: Sᵢ = (2/3) × the sum, over the four income quartiles, of |τ_q − 1/4|, where τ_q is the share of a person's exposure to quartile q.

Intermediate

A person's exposure is 0.40, 0.30, 0.20 and 0.10 across the four quartiles. What is Sᵢ?

About 0.40 About 0.60 About 0.27

3 · Pick at random

In Boston, the team also simulated a random choice among the same reachable venues that still had room.

Expert

What does this random control show?

Segregation comes from how venues are ranked Random choice is the most accurate algorithm Distance limits are what create the segregation

4 · Defend the fix

Expert

Why would a platform that judges only by offline accuracy reject tail-boost?

Because tail-boost increases venue disappearance in most cities Because its benefits appear only in visits people make after following it Because it sends people where their own income group is a small minority

Loop broken

—