Vous êtes sur la page 1sur 8

Nemesysco Press Release

16th March, 2009

“Great ideas often receive violent opposition from mediocre minds” – Albert Einstein.

We wish to clarify our position with regard to the so-called 'scientific research' writ-
ten by Professors Lacerda and Eriksson and published in the International Journal of
Speech, Language and the Law in December 2007, but later withdrawn.

Our response is divided into five sections:

I: The Real Story

‘Scientific research’ does not mean carte blanche to write whatever one pleases.

The article accuses Nemesysco's founder and CEO Amir Liberman of charlatanry, i.e.,
of knowingly and deceitfully inducing his customers to use a technology he himself
knows doesn’t work. While we gladly comment below about the technology and de-
fend its scientific basis and accuracy, we firmly believe that it is perfectly legitimate
for us to take reasonable legal steps to defend both our company's and Liberman's
integrity and reputation from these unfounded malicious attacks – a reputation be-
ing one’s essential asset in the marketplace. Indeed, Liberman was personally
quoted in the respected US magazine Science as saying: ‘[t]he objection was not in the
publication of their study results, it was in their calling us charlatans’1; this was the
sole reason for our demand that the article be withdrawn, but there should be no
doubt that we also strongly contest the article's 'findings'.

It could be said that the article – which Lacerda refers to as being "provocatively
written"2 (an understatement, no doubt), intending to show how "the technology be-
hind the lie detector is a scam"3 – indeed touched a raw nerve, but we accept the
slings and arrows of a free market and free speech; what we do not accept is the ad
hominem attacks and unfounded accusations – as one blogger put it, "…you can see
why they took offence. The paper is unusually lively for an academic article. … in parts
the thing reads more like a blog post or an op-ed than a scientific paper"4, another
commented, "the researchers shouldn't be throwing around words like 'charlatanry'
unless they're ready to prove that Nemesysco is purposely promoting technology that
[it doesn’t itself] believe to be effective"5 and a third opined, "[t]he article is quite un-
usual for a scientific article. For example, it has a section titled 'Who is Mr. Liberman?'

1http://sciencenow.sciencemag.org/cgi/content/full/2009/210/1
2http://www.su.se/english/about/news_and_events/scientists_threatened_with_legal_action
3http://www.thelocal.se/17188/20090127/
4http://neuroskeptic.blogspot.com/2009/02/lies-libel-and-love-detection.html
5http://www.techdirt.com/articles/20090129/1934593574.shtml, comment 8
addressing a private person and claiming that he is a charlatan based on a visit by a
friend made to a private company"6.

However, we should be clear that insofar as the article features a review of a half-
century of lie detection as well as an analysis of voice stress analysis technology
(VSA and CVSA) that has nothing in common with Nemesysco’s LVA technology, we
have no claim or dispute with it.

II: Nemesysco, the Scientific Community and Our Patents:

Firstly, and most importantly, we welcome and encourage proper scientific review
of, or research about our technology's capabilities; however, as in any type of scien-
tific research, some basic criteria must be adhered to and a fundamental under-
standing of the research material must be established. Over the past years we have
either ourselves conducted, or have assisted researchers in the US, Russia, Poland,
Japan, South Africa, Sweden and Israel in doing so, mostly providing our technology
and relevant training free of charge to genuine researchers to this end. Further re-
search is currently being undertaken in Japan, Israel, Argentina and the US.

The article's claims that our technology does not work and cannot work are based
on assumptions made by Eriksson and Lacerda, selectively (as explained below) cit-
ing from only three published US research studies, while completely ignoring all the
other studies that have been published in diverse scientific publications, or pre-
sented at conferences, many of which are available for review on our web site.

Contrary to the claimed lack of academic support for our technology, a growing
number of academic works have focused on Nemesysco’s core technology and
products7 and found in their favour. The most recent one, a 457-page Polish-

6http://www.mindhacks.com/blog/2009/01/legal_threat_for_cri.html, comment by Joseph Lehmann


7E.g.:

 “Forensic Criminology and Psychophysiology: Truth Verification tools, with a special study of Trus-
ter Pro” , Drs. Guy van Damme , Crime Research in South Africa (CRISA), Volume 2, Number 2
(April 2001)
 “Reliability and Validity of Layered Voice Analysis technology in the detection of mental stress”,
Nemoto K, Tachikawa H, Takao T, Sato H, Ashizawa Y, Endo G, Tanaka K, Ishii R, Ishii N, Hashimo-
to K, Iguchi T, Hada S, Hori M and Asada T (presented at the WORLD CONFERENCE OF STRESS,
August 2007, Budapest, Hungary and published at Seishin Igaku (Clinical Psychiatry) Magazine,
October 2008, ISSN: 0488-1281, ISBN: 05627)
 “The Power of Voice: Managerial Affective States and Future Firm Performance”, William J. Mayew
and Mohan Venkatachalam, Fuqua School of Business, Duke University, July 21, 2008 (Available
at SSRN: http://ssrn.com/abstract=1171102).
 “Robotic KANSEI Communication System Based on Emotional Synchronization”, Hashimoto, Mino-
ru (Shinshu University), Yamano, Misaki (Shinshu University), Usui, Tatsuya (Shinshu Universi-
ty), 2008 IEEE/RSJ International Conference on Intelligent Robots and Systems, Sept, 22-26,
2008, Acropolis Convention Center, Nice, France
 “The sexual arousal factor in pedophiles”, Prof Herman Conradie, Department of Criminology,
University of South Africa, Codicillus, Volume 48, Issue 1, 2007
language doctorate written by a Warsaw University student, which was published
only few months ago8, actually covered, unlike Eriksson and Lacerda, many methods
of ‘lie-detection’ available today. The author actually conducted experiments on dif-
ferent versions of the LVA technology, and found that there is a strong and consis-
tent correlation between LVA's emotional parameters and the expected emotional
state of those who display what he describes as a 'lack of honesty'.

Perhaps the most outrageous fact is that Eriksson and Lacerda decided to ignore one
specific piece of research of which Eriksson at least is certainly well aware. In Swe-
den itself, as yet unpublished research conducted in the psychology department of
Umea University (2004-2005) on the then-latest version of the technology (on
which those conducting the research had received proper training), yielded excel-
lent results, and as repeatedly reported in the Swedish press, Eriksson is well aware
of this research and its findings. To the best of our knowledge, this research has not
been published to date mainly due to Eriksson's highly aggressive attack on the
Umea research team, e.g. his calling the lead research professor "an unscrupulous
joker" or "an unscrupulous rascal"9.

One can’t help but wonder if Eriksson's passionate defense of academic freedom re-
lates only to negative findings.

In contrast to all the positive pieces of research mentioned above, in preparing their
article, Eriksson and Lacerda never made any contact with Nemesysco10. Rather
than seeking access to the actual up-to-date technology, they instead attempted to
somehow ‘re-construct’ the technology from only one of our two published patents,
filed over 10 years ago, which they should have realized, as intelligent and good
faith researchers, was unlikely to produce a fair reflection of the technology, and
particularly the technology in its 2007 state.

To be clear: Nemesysco is not an 'academic institute', nor a research institution. Ra-


ther, Nemesysco is a privately funded commercial company with natural profit-
making aims, dedicated toward the aim of producing better and more accurate
technology for voice analysis as well as tools for fighting crime and wrongdoing. As
one of the researchers in the first official research by US Air Force Rome Labs
(commissioned by the NIJ) stated, following a year’s field trial: “I believe in this sys-
tem’s capability of becoming a valuable investigative tool for the law enforcement of-
ficers on the streets of our cities, town and villages across the nation”11.

8Not yet available in English.


9 Free translation from Swedish, found at http://www.vof.se/forum/viewtopic.php?f=17&t=66.
10 Not, as some have suggested, to seek Nemesysco's 'approval' for their research, which in any event

it is not (nor should it be) in our power to give or withhold, but rather in order to get access to the
latest technology.
11 Page 17, “Investigation and Evaluation of Voice Stress Analysis Technology”, AFRL, March 20,

2002.
Patent applications are by law required to provide just enough information about
the present state of the invention and sometimes, a 'preferred embodiment' or a
'proof of concept'. This is what our patent applications (from 1998 and 1999) are
designed to do, and we stand firmly behind the basic ideas presented in the patent
papers. Any mathematician with basic knowledge of signal processing will easily see
the connection between our patent's 'thorns' and the presence of high frequencies
in the analyzed wave form as claimed, and our 'plateaus' and the presence of low
frequencies. It's surprising to note that while, according to his résumé, Eriksson has
a background in mathematics12, he appears not to have made this connection.

If the science of phonetics had any useful parameters for the determination of false-
hood, given the amount of research devoted to it (of which the public is aware of on-
ly a small fraction), no doubt these parameters would have been established by now.
Rather, we firmly believe that we have discovered something completely new.

In fact, the technology intentionally makes use of what Lacerda and Eriksson refer
to as 'artifacts', as a sophisticated (not complicated) way of evaluating the 'outline'
of the frequency range in the voice segment. The first discovery we made (hence, the
first patent) related to the fact that these artifacts, notwithstanding the fact they are
often unstable and influenced by hardware and the surrounding environment, nev-
ertheless still very well reflect the mental state of the speaker.

Much of our confidential know-how relates to the way we filter extraneous 'noises'
out (as well as additional vocal indicators which have been discovered since the pa-
tents were first filed), but that part of the patent that relates to the 'calibration pe-
riod' (which is only one of the filters we use today) was completely ignored by
Eriksson and Lacerda.

Funnily enough, only two of the three research studies cited by Lacerda and Eriks-
son were indeed unfavorable towards our technology. While we object to the re-
search protocols in these two cases (our detailed comments on them are available
on request), the researchers in those cases at least had the integrity to state that the
technology was not used according to our instructions or to add our comments in
the body of their pieces. The third piece of research mentioned by Eriksson (Had-
dad, 2002) is the US Air Force research a quote from which was cited above. This re-
search in general produced a very favorable conclusion about the ability of our
technology to detect stress (which it was the sole objective of the research to as-
sess).

Having done their research, such as it was, Eriksson and Lacerda neither sought our
comments on or response to their findings before publishing, nor made the slightest
reference in their article to any of the positive research cited above – despite the fact
that they must have been aware of it, given not least that some of it is cited on our

12 http://imej.wfu.edu/articles/1999/1/09/aboutauthor.asp
website at www.nemesysco.com/studies.html. Rather, they simply and conveniently
state that the studies they selected are of 'excellent quality', and one need look no
further.

III: Nemesysco’s LVA Technology and Lie Detection

Nemesysco today distributes various solutions for various needs, from lawful inter-
ception, to quality assurance, from entertainment applications to investigation focus
tools; each is designed to work in its own natural environment taking into out psy-
chological and circumstantial factors.

It is vital to make the distinction between our technology used for investigation
purposes and so-called ‘lie detectors’. Although the Professors deride the distinction
between our technology in general and a traditional 'lie detector', it is crucial to un-
derstanding how our technology works, which is by quantifying and classifying dif-
ferent states of stress and other emotions underlying speech which suggest, at times
and in different combinations, that the subject's current state of mind is drastically
different to his normal emotional baseline in a specific portion of his speech– not by
detecting 'yes' and 'no' lies.

Lacerda and Eriksson apparently worry that genuine UK benefit applicants or insur-
ance claimants will have their claims turned down solely due to our technology. Le-
gitimate as this concern may be, it has no basis in real life (even were it permitted
by law, which it is not). Our training, contracts and end-user licenses make it clear
that users of the technology should never assume that it can stand alone in their
fight against fraud and crime or that it is free of possible analysis errors, but rather
that it is designed to be used as an additional tool in that fight, by indicating which
subjects should be further investigated and on what topics. Indeed, this point is
made specifically in the Science magazine article cited above, by a representative of
the UK’s Department of Work & Pensions who is quoted as saying that the Depart-
ment "wouldn't just say from one phone call [that] we're stopping the benefit"13.

We have never suggested that the technology is a magic pill. It is not and never will
be. It is, however, to our belief, the most accurate and efficient technology of its type
in the marketplace, yet even so, we are committed to refining it even more and thus,
Nemesysco's research and development continues.

Furthermore, and crucially, our ‘best practice’ training advises, in so far as this per-
mitted by local law, against informing the tested party about the test14 and the
technology can be and has been used to analyze recorded speeches, sometimes from
distant history. In both of these cases, by definition, the ‘bogus pipe-line effect’ men-

13http://sciencenow.sciencemag.org/cgi/content/full/2009/210/1
14However, in jurisdictions where conducting this type of test without the consent of the subject is
unlawful, consent must of course be requested and obtained.
tioned so widely in the article and press cannot apply, since it can by definition only
apply where the subject is aware that the technology is being used on him, and de-
cides as a result to 'come clean'.

Deterrence is not to be discounted and is a vital element in fighting crime but there
must be something real behind it for it to continue to be effective over a prolonged
period; our solutions have been in use for several years in the same places yet have
continued to consistently generate accurate results time after time. Indeed, where
our new solutions are used to help increase call center quality15 – where by defini-
tion that is no 'deterrence' factor since there is nothing to deter – they are regularly
reported to generate very good results in detecting various positive and negative
emotional states during conversations.

Over 6 million dollars have been invested by Nemesysco and its founders in re-
search and development of the core technology and applications since the patents
were filed. We employ a full-time certified clinical psychologist, as well as investiga-
tors, psychiatrists, mathematicians and other professionals – either full time or as
need requires – and have even employed a full-time ‘devil’s advocate’ specifically to
find loopholes in the technology and new applications. Nemesysco is not merely a
‘one-man band’ in some small nondescript out-of-the-way office16.

Nemesysco’s global network today consists of several hundred people directly in-
volved in technology, market research and sales. Our products and technologies are
used in many parts of the world, by governments and corporations alike. Nemesysco
is of course not permitted to disclose the identity of those customers who have re-
quested confidentiality, but as published in the UK press, UK government agencies
and various insurance companies have used the technology for a few years now
with very favorable results .

Given natural and expected skepticism, especially in the light of the very negative
impression left from 'lie-detection' technologies of the past, it is unlikely that any
other technology of this scale has been subjected to – yet passed – such aggressive
and intensive acceptance tests. In fact, to date, not a single corporate or governmen-
tal body has deployed our technology without first testing it for a significant dura-
tion to ascertain its effectiveness. And yet, we continue to encourage and in fact of-
ten demand that new users perform their own testing – because testing conducted
in genuine 'real-time' situations where the emotions in play are necessarily real and
there is genuine jeopardy is naturally preferable to testing in a laboratory mock-up
situation.

15http://www.nemesysco.com/press/QA5_press_release_120207.pdf
16Not that it would matter if it were, but it seems to be germane to the tone of the article to portray us
as such, as if this is necessarily relevant to the quality of the technology. In fact, our office (the office
visited by Swedish journalist Arne Lapidus back at 2005) sits in the same building as MSN Israel, and
several departments of Elbit, a multi-billion dollar company, are situated next door
In fact, we have observed a simple yet distinct line between the research that found
LVA technology highly accurate and those which didn’t. Testing on mock-up cases,
on actors, or in cases where either the 'stick' or the 'carrot' is missing, find against
the technology. But field studies conducted in airports (both in Russia17 and in Eu-
rope) show almost flawless operation, where personnel involved were able to suc-
cessfully identify subjects requiring further investigation and distinguish them from
unremarkable passengers in ‘real time’. Such a distinction is, as we have said, entire-
ly to be expected. (Referring back to the alleged 'bogus-pipeline effect', it is instruc-
tive that in the Russian test, for example, despite being warned that the technology
was to be used on them, none of the passengers tested voluntarily admitted wrong-
doing until confronted with evidence of it.)

IV: Nemesysco and Amir Liberman

As far as the article's personal attacks on Liberman are concerned, Liberman rep-
lies:"I have personally spent the past 12 years researching emotions in the human
voice outside of the academic world which – it's no secret – I was never really part of.
In fact I am thankful for that, as being outside of the academic world has given me the
opportunity to keep my mind fresh and lead the way in this exciting new field, a field
whose potential the academic and business world are only now beginning to realize,
hence the interest in this article and Nemesysco's response to it. It goes without saying
that we wouldn't have got any patents if our technology were not novel.

"To the best of my understanding, our technology has no relationship to the science of
phonetics in which Professor Eriksson specializes, and the audible characteristics of
the voice wave form are not of our concern and interest. The phonetic characteristics
of the human voice are highly controlled by the speaker, but LVA technology inten-
tionally ignores anything in the voice that can be controlled – like pitch and amplitude,
and anything that relates to the voice 'envelope' – hence the claim that the technology
could be fooled by a deliberate change in voice can be discounted 18. For example our
call center quality assurance systems: where the LVA technology generates expected
results on genuine customers’ calls, it fails to do so on actors' voices usually used for
training purposes in such scenarios, since the actors – unlike real callers – are not ex-
pressing their true emotions.

"The LVA technology has to do with psychology, psychiatry, neurology and criminolo-
gy, and indeed the professionals with whom we work in these fields are very receptive
to our theories and products. Yet, so that there can be no doubt, I have never pretended

17http://www.eastline.ru/eng/news/news.asp?ID=957
18This is not to deny that, as the Professors rightfully state, distinguishing truth from lies is never a
simple task, and that, as the Professors failed to mention, it is highly influenced by the circumstances,
the relationship between the tester and the tested party, the jeopardy of the situation and the inten-
tions behind the lies. Liberman has published a paper (http://www.nemesysco.com/article-
voiceanalysis.html ) which speaks about this complexity in more details.
to be a psychologist, psychiatrist, neurologist or criminologist or indeed a university-
trained mathematician or anything else I am not. (We are currently running research
for the detection of pre-menstrual syndrome in conjunction with an Israeli hospital,
and for the record, I’m not a gynecologist either…).

"Contrary to the professors' patronizing assumption that the academy is the exclusive
font of knowledge, many of the greatest researchers and inventors of the past – and
particularly of the hi-tech present – have or had no or little formal education."

V: Conclusion

In our view, as we have never denied, it is legitimate to question the reliability of


new technology which makes bold claims in a field which itself is a scientific fron-
tier, and one ought even to be skeptical about it. However, we believe that Eriksson
and Lacerda started with the prejudicial view that a ‘lie detector’ cannot work and
then set out to prove that preconception. By failing to distinguish between our tech-
nology and the classic ‘lie detector’, failing to acknowledge that our technology is not
to be used as a sole arbiter of truth or falsehood in any case and failing to do their
research on up-to-date technology rather than a re-constructed version of the tech-
nology they have produced a fatally flawed analysis which we strongly and absolute-
ly refute.

We would like to thank our customers and affiliates for their support, and others
who have been decent enough to contact us over the past weeks since this story
broke to seek our side of the story.

We invite genuine researchers in the relevant fields to contact us with a view to per-
forming additional serious research on the LVA technology. We may be contacted
via our website or at bizdev@nemesysco.com.