Exercise 2: Why "74% Conformed At Least Once" and "64.3% of Responses Stayed Correct" Don't Contradict Each Other — Possible Solution
=================================================================================================================================================

These two real numbers are measuring genuinely different units, which
is exactly why they can both be true at once without any contradiction.
The 74% figure counts PEOPLE - specifically, the share of the roughly
50 real participants who gave at least one incorrect, conforming
answer somewhere across their own 12 critical trials. A person only
needs to conform ONE single time out of twelve real opportunities to
be counted inside that 74% - the figure says nothing at all about how
often that same person conformed beyond that one instance.

The 64.3% figure, by contrast, counts individual RESPONSES, not
people - it's the share of all the individual trial-by-trial answers,
pooled together across every participant and every critical trial,
that stayed correct despite the group pressure. This is a completely
different unit of measurement: one person contributes twelve separate
responses to this total, not just one.

So a real, fully consistent scenario looks like this: most people
gave a wrong, conforming answer only occasionally rather than on
every single trial - meaning a large share of individual people (74%)
can each have contributed just one or two conforming responses to the
pool, while the large majority of all the individual responses put
together (64.3%) still land on the correct side. Both numbers describe
the same real dataset accurately - they're simply counting two
different things: how many PEOPLE conformed at least once, versus how
many total RESPONSES were correct.

ANSWER: These numbers aren't contradictory because they count
different units - 74% is the share of PEOPLE who conformed at least
once across their twelve trials, while 64.3% is the share of all
individual RESPONSES, pooled across everyone, that stayed correct. A
person only needs one conforming answer to count toward the 74%, so
it's fully consistent for most individual responses to still be
correct even though most people conformed at least once somewhere
along the way.

WHY THIS WORKS AS AN ANSWER
------------------------------
This identifies the real unit-of-measurement distinction (people vs.
individual responses) that resolves the apparent contradiction between
the two statistics.