Exercise 3: How a Single-Blind Design Can Still Produce Bias — Possible Solution
===================================================================================

A genuinely plausible scenario: a researcher testing a new pain
medication against a placebo, where patients don't know which they've
received but the researcher administering the treatment and recording
outcomes does know.

When assessing a patient's own reported pain level - especially where
the measurement involves any real judgment call, such as observing
body language, asking follow-up questions, or deciding how to phrase a
question about pain - a researcher who knows a given patient received
the real medication might unconsciously ask more encouraging or
leading questions ("You're feeling better, aren't you?"), record
borderline responses more favorably, or spend more time reassuring
that patient specifically, none of which happens deliberately or
dishonestly, but all of which can genuinely skew the recorded outcome
in the treatment group's own favor.

The researcher's own real knowledge specifically affects: how
questions are phrased, how ambiguous or borderline responses get
interpreted and recorded, how much attention or reassurance a patient
receives, and potentially even subtle nonverbal cues that could shape
a patient's own reported experience after the fact - none of which the
participant's own blindness to their treatment group does anything to
prevent.

ANSWER: A researcher who knows which patients received a real pain
medication (versus placebo) could unconsciously phrase questions more
encouragingly, interpret ambiguous pain reports more favorably, or give
treated patients more reassurance - genuinely biasing the recorded
outcome even though the patients themselves remain blind to their own
group. The researcher's own knowledge affects data recording and
interaction quality, a source of bias single-blinding alone doesn't
address.

WHY THIS WORKS AS AN ANSWER
------------------------------
This describes a real, concrete, plausible mechanism by which
researcher-side knowledge specifically introduces bias, rather than a
vague general statement that researcher bias "could" occur.