Exercise 3: What Additional Data to Request Before Accepting the Conclusion — Possible Solution
===================================================================================================

Before accepting that the company's aggregate promotion-rate gap
proves the process itself is biased, the Berkeley case suggests
requesting the data broken down by a real, relevant subgroup - most
directly, department or role/level within the company, since that's
the exact factor that explained the Berkeley reversal.

Specifically, worth requesting:

1. Promotion rates for men and women WITHIN each individual department
   or job category, not just the company-wide aggregate - to check
   whether the gap actually exists at that more granular level, or
   whether it disappears or reverses the way the Berkeley department-
   level data did.

2. The real distribution of which departments or roles men and women
   are concentrated in - specifically, whether women are
   disproportionately represented in departments or levels with
   structurally lower promotion rates (for reasons unrelated to the
   promotion process itself, such as newer teams with less overall
   turnover at senior levels).

3. Whether the departments/roles with lower promotion rates are lower
   for a reason connected to the promotion decisions being questioned,
   or for some other structural reason entirely (e.g. team size, how
   recently the department was created, natural turnover rates).

If the per-department data shows no gap, or a reversed one, the honest
conclusion becomes closer to the real Berkeley finding: the aggregate
number reflects where different groups are concentrated, not how
individual promotion decisions are actually made - a genuinely
different (though still potentially worth investigating) real problem
than direct bias in the promotion process itself.

ANSWER: Request promotion rates broken down by individual department
or role, along with the real distribution of which departments/roles
men and women are concentrated in - directly mirroring the Berkeley
case's own department-level breakdown. If the gap disappears or
reverses once broken down this way, the real explanation is likely
about where different groups are concentrated, not necessarily bias in
individual promotion decisions.

WHY THIS WORKS AS AN ANSWER
------------------------------
This directly applies the Berkeley case's own specific diagnostic
approach (breaking aggregate data into real subgroups) to a new,
analogous real-world scenario, rather than offering only a generic
"get more data" response.