IEEE Technology and Society Magazine - June 2017 - 80

were the double blind series to fail, as the conditions
during both series would be essentially identical.

Results
The results of the calibration stage were that the
GT200 had correctly located the sample in four out
of four attempts. The results of the double-blind
stage were that the detector located the sample in
three out of twenty attempts and failed in the
remaining seventeen.
The calibration stage showed that the GT200 was
perfectly capable of finding the sample when the operator knew beforehand the location where it had been
hidden. The probability of obtaining this result by

The detector located the sample in
three out of twenty attempts and
failed in the remaining seventeen.
chance is (1/8)4 ≈ 0.00024, negligible. Thus, this stage
also showed that it couldn't be claimed that the experimental conditions (sample, boxes, envelopes, bags,
neighbors, presence of researchers and witnesses,
position of boxes, room, walls, weather, etc.) hampered
the search.
Given a detector with efficacy p, i.e., with a probability p of success in any one trial, the probability of having
m successful trials out of N = 20 independent search
attempts would be described by the binomial distribution b(p, m, N). A completely ineffective detector would
have chosen randomly between any of the 8 boxes, so
would correspond to p = 1/8. The corresponding distri− = 2.5 and a standard deviabution has an average of m
tion of Δm ≈ 1.5, and therefore is completely consistent
with our experimental result m = 3 for the double blind
stage. This contrasts even with a poorly effective detector, with an unacceptably low efficacy of p = 80%, for
− = 16 and Δm ≈ 1.8, so our experimental result
which m
would be more than 7 standard deviations away from
the mean, a result with a negligible probability
b(.8,3,20) < 10−9.
The results may be summarized by Fig. 0 which
displays the normalized likelihood l(p) = b(p,3,20)/
b(3/20,3,20) that the detector had an efficacy p, given
that there were 3 successful searches in 20 trials.
With a likelihood l(1/8) = 0.95 we can conclude that
the detector was completely useless, while the likelihood l(p) of any p ≥ 0.5 is negligible, less than 0.005,
and very rapidly decreasing (see inset).

Further Developments
The Mexican government payed about 350 million
pesos for the acquisition of more than 940 molecular
detectors [2], though there are claims [16] that the actual number was 1112 units for about 450 million pesos.
According to press releases issued during the years
2010 and 2011 (see for example [4]) these devices were
successfully used on an almost daily basis. They were
used to justify searches in homes and schools, many
times without judicial warrants [17], and were employed
at roadblocks yielding numerous detentions, trials, and
imprisonments. The number of illegal searches
increased dramatically since the GT200 was adopted,
so the National Committee of Human Rights issued a
recommendation declaring its use violates human rights
[17]. According to Lemus [18], more than 5000 people
were imprisoned using the GT200 as main evidence
and by mid-2015, 1980 of them remained in jail, 652 of
them in federal prisons.
The resolution [11] agreed upon by the Senate in
June, 2012, urging the President to order a scientific
evaluation has not been complied with yet. Related resolutions presented to the House of Representatives and
the Senate [16], [19], [20] urged the Attorney General to
investigate the possibility of corruption in the purchase
of these devices and to investigate the company that
sold them (SegTec) for possible fraud. They also urged
the Ministry of the Interior to publish the number of
people that had been charged, judged, and imprisoned
using evidence provided by the GT200 and the Foreign
Relations Ministry to investigate the role played by the
U.K. government and its former ambassador Paxman in
the sale of the devices. These resolutions were ignored
and were not even put to a vote [21].
In Aug. 2012, the Mexican Supreme Court was
involved in a case to evaluate the use of the GT200 to
provide evidence: Two women were accused of trafficking drugs. The judge in charge of the case freed
them using information contained in [7] to argue that
the GT200 didn't provide scientific evidence. The
Attorney General appealed the case and the Supreme
Court stepped in. In Mar. 2013, the Attorney General
gave up on the case before the Supreme Court could
set a precedent.
In July 2012, Gary Bolton was charged in the United Kingdom for dishonestly representing the GT200
as capable of detecting explosives [22], [23] and his
case was taken to court. Similar accusations were
brought at the time against the manufacturers of
other similar detectors.
In Dec. 2012 the Mexican Attorney General announced it would discard its detectors [24]. We inquired
for official documentation through IFAI but were told
that it was non-existent.

IEEE Technology and Society Magazine

∕

june 2017

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