IEEE Technology and Society Magazine - June 2017 - 79

Remarks
This protocol was designed to account for the expected
distrust among the participants. Information exchange
between SH and SS had to be avoided; three bits of
information could subvert a trial. Expected excuses for
june 2017

∕

The number of illegal searches
increased dramatically since the
GT200 was adopted.

failures, such as accusing the experimenters of contaminating the boxes, had to be precluded. It was also
important to convince the soldiers that the experimenters wouldn't trick them.
The number of boxes was chosen as 8 = 2³, so that
three coin tosses could select a box randomly. The
number of trials was chosen as 20, so that a detector
not better than randomly choosing boxes would be
unlikely to succeed in more than half of the trials (probability less than 10−4), and a moderately efficient detector (with efficacy of 85% in each trial), would be as
unlikely to fail in more than half the trials. Thus, success or failure would be clear cut.
The purpose of the calibration series, which we
expected to be successful, was to eliminate excuses

N
E

W
S

Figure 1. Schematic drawing of the ballroom, indicating the
entrances, the cardinal directions, and the positions chosen for
the boxes in the initial (green) and the following (red) parts of
the test.

1
Likelihood

1e+00

0.8
Likelihood

no kind of contact took place between the members of both teams.
2) SS used his GT200 to identify the box presumably
containing the sample, and, without opening it,
placed an indicator on top of its lid. Then, the S team
left the ballroom through its main entrance. The
experimenter ES of the S team knocked on the door
of the waiting room, to which the S team entered
after allowing the H team time to enter the ballroom
through the lateral entrance.
3) The H team recorded and signed forms indicating the
actual location of the sample and the location identified by the GT200.
4) The steps above were repeated until 20 iterations
were completed.
After completion of all the iterations, the results were
written out to a final form where all the participants
signed their agreement.
By the end of the fifth iteration SS felt exhausted and
claimed it would be impossible for him to proceed with
the test. In order to be able to finish the test we decided
to swap operators after each fifth trial. To reduce the
distance to walk during each search, making it less tiresome, the soldiers suggested a new disposition for the
boxes, making a zig-zag pattern with approximately 6 m
between neighboring boxes, as shown in Figure 0, and
signed a statement declaring that the new disposition
would not hinder the efficacy of detection.
Unfortunately, swapping operators allowed them to
know the ongoing results before the test was finished,
violating its strict double-blind character and leading to
complaints and excuses for the numerous failed trials.
However, the excuses were quite lame and we were able
to deal with them and complete the test. For example,
they complained about possible contamination with
marijuana vapors (they were using a cannabis card in
their GT200 to presumably enhance its detecting
power), they claimed that climbing the stairs between
the waiting room and the ballroom left them agitated
(they were allowed unlimited rest periods and several
breaks), they stated they required a trained companion
dog to assist in their searches (we were testing the
device, not the dogs), and they argued that the GT200
was not expected to point at a precise location (the
manual stated the uncertainty was 2 m, less than the
distance between boxes).
The times taken for each trial fluctuated from 4 min
to 24 min, and we took four rest breaks [13].

0.6

1e-20
1e-30

0.4
0.2
0

1e-10

0.2 0.4 0.6 0.8
Efficacy

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Efficacy

Figure 2. Likelihood of the detector having an efficacy p given
the experimental result of 3 successful hits out of 20 tries. The
vertical line corresponds to complete randomness, p = 1/8. Inset:
Same but with a logarithmic scale.

IEEE Technology and Society Magazine

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - June 2017