http://www.csoonline.com/read/020106/vendors.html

Finding Truth in a Vendor’s Pitch

By Lew McCreary

There might be no more pitiable predicament in life than being pitched a product that may not work very well by a salesman who works very well indeed.

This is never more true than when the thing being sold is technology that claims for itself nearly wizardly capabilities, as many do. For example, the ability to peer into human hearts and minds to discern malicious motives and intentions.

Who wouldn’t want something that could do that?

Thus the Media Room at Boston’s Logan Airport was packed last week with eager law enforcement and transportation minions on hand to see a demonstration of layered voice analysis (LVA) software, ably shown off by Richard D. Parton, Phd., CEO of V Worldwide, the Chicago-based U.S. distributor of software developed in Israel by Nemesysco. (Israel has such a luminous aura of security competency that if something was invented there, that fact alone endows it with added credibility. I know that doesn’t make rational sense, but there you have it: The adoption of Israeli-made security technology is clearly a best practice!)

Almost any technology is mysterious to those of us who are not software engineering wonks or creators of elegant algorithms. And no group of products has inspired more heated debate than those whose purpose is to reliably unveil deception. Legions of hostiles gather on websites dedicated to the proposition that polygraphs and other machine-based detection solutions are an unvarnished evil, capable of entrapping the innocent and sparing the guilty. Judging by the posts on sites such as www.antipolygraph.org, the participants are resistant to persuasion by even the strongest positive proofs. They swarm like piranhas around anyone nervy enough to offer even a tepid defense.

But there is little conclusive affirmative research and plenty of solid debunking for most of the deception-detection approaches so far introduced. (As for LVA, there is nothing scientifically rigorous yet, either up or down, according to Parton—though he says the Department of Defense has research in progress.) It may be that some of these tools are useful to gifted investigators as a form of validation; but the science suggests they wouldn’t save an inept Inspector Clousseau from himself.

Over the years, at this and other magazines my company has produced, I’ve sat through many hundreds of vendor meetings. Almost all are similarly architected to include a PowerPoint presentation followed by a demo of the technology at hand and a few minutes of questions and answers. A small percentage of such meetings are fascinating; many more include lively conversations about this or that marketplace and are far from a waste of time. But I’d say that roughly a third are either numbingly boring or else drag me out into deep water where the complex material overwhelms my small powers of understanding.

The best one can hope to get from a vendor briefing is a clear understanding of what the thing claims to do and why it’s good for you; how it’s different from other products already on the market; some sense of the probable costs (both overt and covert) of adopting it; the degree of difficulty or disruption likely to be suffered by staff learning how to use it (or customers having to put up with whatever process changes it triggers); and open access to others who have already bought the product and can comment honestly on its pros and cons as they’ve experienced them.

It’s important to remember that the demonstration is a well-rehearsed theatrical production scripted to make the starring technology look really good. It creates excitement, but not necessarily enlightenment.

Parton began his presentation by clarifying that LVA is not CVSA—computer voice-stress analysis. This is likely because a volume of research exists casting doubt on the efficacy of CVSA, and Parton didn’t want his product tarred with that brush. He offered this distinction: "LVA measures voice frequency, not voice energy." In fact, he said, it measures 128 vocal parameters—an impressive sounding number enunciated with evident pride.

The voice, Parton said, "contains the DNA of thought." It is not what the subject says, but how he says it that matters to the decoding algorithms that are the software’s active ingredients. The product can be used in casual conversations or purposeful interrogations; with or without the subject’s knowledge; and in live encounters or, retrospectively, on recorded audio files.

This last point prompted a wag in the front row to ask the question that might have been on many minds (well, mine at least). "Could you analyze Bush’s State of the Union address?"

"We’ve steered clear of politicians ever since Bill Clinton," Parton said, delivering his first certifiable laugh line. Maybe so. But Clinton is neatly grandfathered into Parton’s spiel as one of two poster boys for prevarication (Laci Peterson’s murdering husband Scott is the other). Parton played the tape of Clinton famously shaking his finger and intoning that he "did not have sexual relations with that woman, Miss Lewinsky." As the tape played, we watched the squiggles etch across the software interface. Parton stopped the tape at intervals and pointed to spikes that indicated stress, concealment, excitement, uncertainty, conflict, inaccuracy—indexing thoughts at odds with the president’s spoken words. In my notebook is written, "Poor Bill Clinton—case materials for deceptiveness."

It’s interesting to observe the vocal etiology of what we already know are lies. And in fact these post-dated analyses take on the trappings of a parlor trick—a tea-leaf reading of the past engineered to convince the credulous that the future can be predicted. Being credulous myself, I actually found it amazingly persuasive. (It is testament to the appeal of a promising innovation that Parton kept everyone’s attention for nearly two and a half hours.)

Then Parton asked for a volunteer from the audience. Silence ensued. At the start of the session, I was identified to all in attendance as a member of the media—like being the skunk at a garden party—and Parton cautioned everyone else to speak with extra care because of my presence. So I briefly considered raising my hand, figuring that this particular audience would relish seeing the journalist on the hot seat. But the truth is I was uneasy. What if this thing actually works? I thought. Any of my various misdemeanors might be exposed to people with guns and badges!

Fortunately, Dennis Treece leapt into the breech. Treece is director of corporate security for Massport, the agency that runs Logan Airport and the Port of Boston, among other facilities (see "Safe Harbor," CSO, April 2003). He organized this meeting and invited the various agency people and other parties to attend.

Treece has been in security for more that three decades. Just prior to coming to Massport in 2002, he worked for ISS, an Atlanta-based security solutions vendor. So he knows how this works, and he’d already done some homework on V Worldwide by the time he sat down to be questioned. Treece wisely confined his session with Parton to professional matters. Even so, the software interface—projected onto a screen visible to the audience but not to Treece, who faced us—showed an overwhelmingly undeceptive profile marred only by a few clear readings of stress or inaccuracy that indicated the need for further inquiry. That questioning allowed him to provide plausible answers, labeled as truthful by the system, that explained the spikes benignly.

Parton is clearly an experienced questioner; he is also intimately familiar with the most advantageous use of the tools he sells. And he has surely given this presentation many times before. It is easy to be swept up. A system such as this is the holy grail of both prevention and detection. If you are an airport security leader, a district attorney or a fraud investigator, this kind of technology is about the most succulent red meat there is. If something helps you spot the shoe bomber, eliminate the innocent to concentrate on the guilty, or nip a phony insurance claim in the bud, you win and the bad guys lose. The momentum of your own urgent hope and desire to prevail over wickedness can easily be turned to the vendor’s advantage.

I want to be clear that I intend no critique of LVA as a technology category, or of Richard Parton and V Worldwide in particular. I don’t know nearly enough to render a judgment. And for all I know, this product may be the genuine breakthrough that many in the field of detection have been waiting for. It certainly seemed clear to me from the reactions of those in the room that interest in at least trying out the system was very high indeed.

But I was reminded anew how incredibly hard it is to separate claims about products from their provable capabilities. Good marketing offers food to the starving and drink to the parched. Done well, it brings a vendor within striking range of the customers’ wallets. But it shouldn’t be confused with neutral information, because it isn’t. Obviously, it’s in the best interest of buyers to let euphoria pass and then get down to due diligence. Check references (preferably those you can pick on your own from a long list of customers); call on independent analysts and trusted peers; and have a good set of metrics for the kind of improvement you would need to see in order to justify the cost of a system.

Only then are you likely to arrive at a final judgment that amounts to "Probable Truth," as Parton’s system interface would put it. Or else you could buy Parton’s system and henceforth use it on all of your vendors.