A talk with Andrew Thompson, a New York City-area private investigator and a researcher of criminal behavior and interrogation techniques. Andrew shares how he gets people talking while lacking the authority and command of a police badge, and what studying FBI crime cases and interrogations taught him about linguistic clues to deception in people’s stories. We also discuss his research into criminal personality categories and how that may be tied to linguistic patterns. We discuss how mistaking signs of anxiety for guilt or deception can send investigators down very wrong paths; and how aggressive questioning can contribute to false confessions and other negative outcomes. We also discuss why appearing uninterested in an important revelation from someone can sometimes encourage them to reveal more, and other tips on influencing and manipulating people in order to achieve investigative results.
Episode details:
- YouTube (coming soon)
- Apple Podcasts
- Spotify
Resources related to or mentioned in this talk:
- Paper by Thompson and Hartwig, 2022: The language of high-stakes truths and lies
- A talk with Maria Hartwig and Christian Cory on science-based interviews
- A talk with Bennett Kleinberg about how AI/machines detect deception
- The work of Gabriel Salfati, whose criminal personality profiling work we mention
TRANSCRIPT
(transcripts are generated automatically and will contain errors)
Andrew Thompson: The other thing that John Jay [College] had, I don’t think it has anymore, but they had a collaborative, uh, agreement with the FBI’s Behavioral Science Unit, the sort of profiling unit, where certain researchers were given access to this massive vault of FBI files, closed FBI cases. It was housed on, in the college in a secure room. And so some of us, and I was lucky to be among them, got access to that to conduct research using those files. And they were just raw closed case files, um, of everything from terrorism cases, serial killers, uh, contract killers, all, all sorts of, you know, uh, child, uh, molestation cases, all sorts of stuff And so my big, my thesis, and that turned into some published research, was on looking at sort of the linguistic behaviors of sexually motivated homicide offenders during their interrogation when they were lying and telling the truth.
So we found that a lot of the techniques that, and a lot of the, the potential cues for deception that had been developed and empirically supported didn’t hold water in our sample. That being said, a couple of potential deceptive cues did come out in those studies as relevant.
Zach Elwood: That was a clip from my talk with Andrew Thompson, a New York City private investigator and a criminal behavior researcher.
This is the People Who Read People podcast, a podcast aimed at better understanding people. It’s hosted by me, Zachary Elwood. You can learn more about this show and subscribe to get updates, at behavior-podcast.com.
I had this talk with Andrew Thompson back in June, four months ago. I’ll say that this was one of my favorite talks I’ve done for this podcast. One reason I wanted to talk to Andrew was because I’m currently doing research for my book, which will be called Read People For Real, nad which I hope will be a helpful resource for getting quick reads of people. It will largely focus on the things people say, and it will push back against false and misleading claims and myths in the “behavior expert” space. As part of my efforts for this podcast, and for that book, I’d been wanting to talk to someone in the private investigation space for a while, and so this talk with Andrew got me thinking about a wide range of things that I found very interesting, and which helped me define what I wanted to talk about with my book. Some of the topics Andrew and I discuss include:
- The methods private investigators use to persuade people to reveal information, and why that approach has to be much more persuasive and subtle due to lacking the authority of a police badge.
- What Andrew’s research into real murder interrogations reveals about lying, mental effort, and the details people include in their stories.
- His research into criminal personality types and how that might link up to linguistic patterns
- How mistaking anxiety for deception or guilt can lead to confirmation bias, aggressive questioning, and false confessions.
- Why acting nonchalant when sources reveal crucial information can help an investigator learn more.
- Why statement verifiability is such a big clue to whether someone is lying or telling the truth
A bit more about Andrew: he’s a private investigator with over 15 years of international experience, across a range of criminal, civil, and corporate matters that include serial homicide and multinational securities fraud. He earned his master’s degree in forensic psychology at John Jay College, where he developed investigative interviewing techniques in collaboration with the FBI’s Behavioral Analysis Unit and US government’s joint High Value Detainee Interrogation Group. His research has been published in multiple academic and professional journals. He’s also an advisory council member on Project Aletheia, an organization of leading criminal interview researchers and practitioners whose goal is to advance evidence-based interview techniques.
On a previous episode earlier this year, I interviewed Maria Hartwig and Christian Cory, who are leaders of Project Aletheia; we talked about science-based criminal interviewing and interrogation. Also it might be worth noting that Andrew studied with Maria at John Jay College, and did some research with her there.
You can find a transcript for this talk, and get links to papers and other resources we mention in this talk, at my site behavior-podcast.com.
I did remove the first few minutes of Andrew and I’s talk. I’ll mention some of the points from that, just because Andrew does allude to that later. Andrew told me about him being from Canada, and said he had several family members who were in the Royal Canadian Mounted Police, and that was a vocation that he thought from a young age he might wind up in. But in college he developed an interest in research and that led him to pursue graduate studies in forensic psychology at John Jay College. Then he eventually settled in the U.S. and pursued private investigation work.
Okay here’s the talk with Andrew Thompson.
Zach Elwood: Hi, Andrew. Thanks for joining me.
Andrew Thompson: Hey, Zach. Thanks so much for having me on. It’s great
Zach Elwood: I know that the investigative, private investigative field can be quite wide, and, uh, having moved to New York recently, New York City, I, I learned about all of these different roles that were available, you know, private, corporate, uh, security and, and investigative related work, you know, law firm related work and, and I’m curious what kinds of things have you done over the years, uh, out of all those various things you can do in those re- realms?
Andrew Thompson: Yeah. So I mean, that’s one of the great things about PI work, right? Is that it carries such a wide swath of, of opportunities and, you know, people tend to specialize, right? So you can be the infidelity people who go out and do surveil- or insurance guys who go out and do surveillance, and you can go all the way out over to the other side, which is like more business intelligence, kinda closer to, you know, legal corporate espionage type work.
I’ve done kind of a smattering of, of almost everything. I haven’t done the, the cheating spouse stuff. Um, probably make more money if I did, but, uh, I’ve done kind of everything from sort of plaintiff civil lawsuit type investigative work, um, to working directly for law firms doing like multinational securities fraud cases, to private investigations doing primarily criminal defense work in the state and federal systems, doing stuff like, you know, murder for hire cases, rapes, robberies, um, your garden variety murders, gang stuff, that kind of thing.
Zach Elwood: What would, what would you say, uh, w- what are some of the major differences between the private investigative work and more law enforcement related in- investigative work?
Andrew Thompson: I think PI work is in an, you’re in an interesting position, right? Because you… Police officers have certain sort of implicit and explicit authorities and sort of threats that they can levy just by, you know, not even consciously, but just by way of having a badge, right?
You kind of feel compelled to talk to a police officer a lot of the time. PIs have none of that, right? Like I have no authority. I can… I have, pose no threat to you if you decide not to talk to me. So in that way, we’re probably, our tactics are more situated probably more closely to like a CIA officer. I know you’ve interviewed a, a couple of those guys.
But on that end of it, we don’t have the same resources and we don’t have the same permissions, right? We’re not allowed to go break the law. Uh, whereas like that’s what a CIA officer does, right? At least in foreign countries. So you’re sort of stuck in this middle ground where you have to be very persuasive, uh, very manipulative at times to make up for the fact that we don’t have a badge and we don’t have the resources at play.
But at the same time, you don’t have all the freedoms that, that, you know, an intelligence officer would have going out and recruiting sources- Mm … for the, for the government.
Zach Elwood: Do some examples come to mind of the types of situations where you have to be super persuasive and/or manipulative to get people to talk to you?
Andrew Thompson: One case, uh, in particular, we had a, a situation where for some kind of important but m- mostly boring reasons, we had to figure out what a hospital used for their sharps containers, right? The boxes where they, like, put needles and, and sort of other bio threats in when they’re discarding them And basically a nurse had grabbed a knife off of a, a murder suspect and thrown the knife in the sharps box.
Cops come, no one can find a key for it, uh, so the cops rip it off the wall, rip it open, and take custody of the knife into evidence. So we had to figure out sort of like, okay, well, who owns the sharps container, right? ‘Cause we’re trying to determine whether or not the cops should’ve gotten a search warrant before opening this thing up.
But the hospital was not willing to give us any information. Hmm. So you have to sort of go around it, right? The, the hospital system had a good relationship with the police, and so they were responding to subpoenas, but not in any helpful kind of way. Hmm. So ultimately what we did is I created a fake sort of medical company, like a medical practice that I was setting, I was, quote-unquote, “setting up” and I was able to figure out what the company that made the, the sharps container was, and I called them up as a potential new client and had them sell me on this sharps container and sort of said like, “Oh, you know, a couple of the practitioners at this new practice I’m standing up worked at this hospital during this time period and they really loved that sharps container, and they really loved the, how your process worked, and so can we get that exact model and how does that work, and can you tell me about it?”
And ultimately what it turned out was not only did the hospital not own those sharps containers, they were like leased essentially, they didn’t have access to them because it was like a full service kind of deal where like this company, this vendor would come in and empty them and replace them. Hmm. So like the nurses and medical staff couldn’t have consented to the cops getting into it even if they wanted to.
Hmm. Um, and so that ended up going to an evidentiary hearing and ended up with a very favorable sort of result for the, the defendant in that case. Whereas if you’re a cop, right, you just… I mean, in this situation you would just literally ask them and they would give you the information, but you know, if they were slightly more aggressive, they could get a search warrant or, or a, or a subpoena for it that they would comply with.
Zach Elwood: Yeah, so yeah, you basically have to be more persuasive and more creative in general with your approaches because of that.
Andrew Thompson: Yeah, and it, it can go either way, right? So like listen, especially, you know, we live in New York, you know, people have, are probably more anti-cop here than in lots of places in the country, so sometimes you can play on that and maybe I get the benefit of the doubt versus a police officer in certain ways.
Hmm. Mm-hmm. But as a general rule, I have to sort of be a little more creative in how I go out and get information. Same with getting access to places, right? Like we’ve had cases where we’re trying to develop information on people who live in very secure sort of high worth buildings. And so you have to get past them, right?
If you’re a cop, you just walk up and flash your badge and say, “I’m going up to floor 17 to see Mr. Smith.” Mm. If I do that- … just walk up and flash my, you know, my PI creds or whatever, the security guard’s probably gonna tell me to go pound sand, right? Mm-hmm. Mm-hmm. So I have to come up with some sort of either cover to get under there under a ruse, or I have to find some way to persuade the security guard to let me by knowing who I am.
Zach Elwood: Mm.
Andrew Thompson: So it’s a lot more of a softer touch as a general rule.
Zach Elwood: I would imagine those skills, like the skills you develop doing that, like those would be super useful if you were a cop too, but cops don’t, probably don’t, aren’t forced to develop that skill as much I would imagine. Yeah.
Andrew Thompson: Yeah, my… Listen, I have lots of friends that are cops.
I’ve, I’ve worked with lots of colleagues that are cops, both like with them and on the other side of the aisle who are very, very good at their job and very talented. My experience has been though that guys who were police officers and come over to the private sector to do PI work tend to struggle at first because of that reason, right?
They get really frustrated that people just say no to them or they don’t have the same… It’s not quite as easy, right? You have to be a little softer with people. Mm-hmm. Um, you have to be a little more creative. And so it, it is a definitely a different animal when you come on this side. But, you know, like, uh, Patrick Skinner, I don’t know if you know who he is, he was a CIA case officer and then rolled back his cover and became a cop.
He’s, he’s now a detective in Atlanta, I think. Hmm. Or Savannah. Maybe Savannah. Um, he’s big on Twitter. You know, he, his big thing he says all the time when he’s teaching new police officers is like, “How would you approach this scene if you weren’t a cop? If you didn’t have a gun and a badge, what would you do when you rolled up on scene?”
Hmm. And so I agree with you. I think like if more police officers had that attitude or had that sort of mindset, I think you would find they’d actually be a lot more effective.
Zach Elwood: And, uh, is it true part of the skill too is developing a really good network of ways to get information? Uh, and uh, it seems like that’s even more valuable too as a, a private investigator.
Am I, am I right on that?
Andrew Thompson: Yeah. In terms of like, like behaviorally, like getting like elicitation or more like developing groups of-
Zach Elwood: Like, uh, people around town that know things or in the, in the industry that can hook you up with various pieces of information basically. Sort of like more like a jour- like a journalist does almost.
Andrew Thompson: Yeah, totally. Right. Depending on the, you know, depending on the type of case you’re working, right? You know, if you’re working a murder, certainly if you can develop like, you know, you’re working a gang murder in like a housing project, you certainly as the only white guy can’t be probably walking in there and pr- being very effective at questioning people, uh, mostly ’cause they probably think you’re a cop.
Um, so you work through intermediaries and so it’s, it’s about making friends however you can. And then more on like the, the some of the other stuff I’ve done, like the securities fraud stuff, yeah, a lot of these cases are international, right? So they’re pro- the firm is probably not gonna be willing to fly me out to rural Mongolia to like knock on doors related to like a goldmine investigation there.
I’m working with NGOs and working with local workers and looking e- sometimes even with local government officials that I can sort of get my hooks into and have them do a lot of the fact-finding for me and bring people to me, both because it’s more effective and also ’cause it’s safer for the source, right?
Rather than interacting with me directly.
Zach Elwood: Taking a step, uh, backwards a bit, uh, I was curious, did you know from a young age that you wanted to do this kind of work? Was there, was there some, were there some, um, experiences that early on got you really interested in this kind of work?
Andrew Thompson: Yeah. Yeah. So like, like I already kinda touched on the fact that I, I grew up along or along or around a lot of mounties.
Um, that was always kinda what I wanted to do. And then I’ve also… I was involved in martial arts from a very young age, um, and kinda from like five or six years old, and kinda grew up still doing it, right? I sort of never stopped, um, and became fairly good at it. I was a, a national karate champion, and I was a member of the nat- the Canadian national team for five years.
Mm. And so particularly when it comes to like b- behavioral analysis and sort of reading people, you know, it’s, it’s really cliche to say like fighting’s a mental game, but it, it kind of is. Mm-hmm. And you know, I think with a lot of problems with deception detection stuff is that people don’t get good feedback, right?
So cops think they’re maybe really proficient in it because they never really learn when they’ve been lied to, or it’s harder to figure out when you’ve been lied to at like a traffic stop or something. In a fight, it’s immediately apparent when someone has faked you out. Mm. Um, and you get typically very immediate and very painful feedback when you mess up.
Mm-hmm. Mm-hmm. Um, so that makes you a very fast student in that regard. Yeah. So that was part of it, and certainly the, the sort of ethos that goes along with growing up in martial arts of wanting to defend people and, and protect people and that sort of thing also played into it- Mm-hmm … um, quite a bit.
Zach Elwood: You’ve done some interesting research at, uh, John Jay and maybe other places. Do you wanna talk about what has stood out to you as some of the more interesting, uh, research you’ve done maybe in investigative or behavior-related spaces?
Andrew Thompson: Yeah. So yeah, so like I said, I, I went to John Jay ’cause it really is the, the best in, at least in North America.
And actually I know that you just recently, uh, interviewed Dr. Hartwig. Uh, so Dr. Hartwig was my graduate supervisor. I was in– I came to New York specifically to work under her and was really- Hmm … lucky to get to. The other thing that John Jay had, I don’t think it has anymore, but they had a collaborative, uh, agreement with the FBI’s Behavioral Science Unit, the sort of profiling unit, where certain researchers were given access to this massive vault of FBI files, closed FBI cases.
Um, it was housed on, in the college in a secure room. And so some of us, and I was lucky to be among them, got access to that to conduct research using those files. And they were just raw closed case files, um, of everything from terrorism cases, serial killers, uh, contract killers, all, all sorts of, you know, uh, child, uh, molestation cases, all sorts of stuff And so my big, my thesis, and that turned into some published research, was on looking at sort of the linguistic behaviors of sexually motivated homicide offenders during their interrogation when they were lying and telling the truth.
So, you know, I’m sure you know the, the biggest issue with deception research a lot of the times is that there’s not a lot of sort of external validity, right? A lot of the research is based on college kids lying or not lying in certain experimental paradigms. This was the largest study to look at real world high stakes interrogations.
Hmm. Um, and it was also the only study, to my knowledge, to look at only a s- like, only murderers, and even within that, only sexually motivated murderers. Because the, you know, one of the ideas is that someone who’s arrested for a theft and someone who’s arrested for a sexual serial murder are probably very different people.
Both the stakes are very different inside the interrogation. But also- Mm-hmm … the type of people that would commit those crimes if they did them are probably different, right?
Zach Elwood: Right. Yeah, narrowing it down to a very specific type of situation is important, yeah.
Andrew Thompson: Yeah. And so what we found was kind of interesting, right?
So we found that a lot of the techniques that, and a lot of the, the potential cues for deception that had been developed and em- empirically supported didn’t hold water in our sample, and that could be for a number of reasons. Either maybe these types of offenders were better at lying than your average person, or maybe the, the stakes of an interrogation changes the dynamics a little bit.
So a lot of the, like, RE technique stuff, as you’d expect, did not hold much water. Um, I threw in a couple of the scan metrics just because I was curious, and those also kind of predictably failed
Zach Elwood: Yeah, the SCAN is the, the, the small, uh, verbal statement analysis type patterns. That’s what the S- the SCAN thing is.
Yeah.
Andrew Thompson: Yeah. Yeah, that really has no empirical support anyways, to, to my knowledge. Um, it’s one of those practitioner-focused or developed ones that don’t really stand up when you, when you really start to dig into it very much. That being said, a couple of potential deceptive cues did come out in those studies as relevant.
Um, the biggest ones being sort of what would, what would be a set of behaviors that were indicative of someone experiencing cognitive load. So liars have to think harder than truth-tellers. You know, you saw, like, a lot more speech errors. Um, you saw sentences that became shorter and more simple, less, like, word variability.
You saw, um-
Zach Elwood: Delayed
Andrew Thompson: response … longer… Pardon me? Yeah,
Zach Elwood: pauses. Like pauses.
Andrew Thompson: Yeah, exactly. Longer pauses. Yeah. Longer latency, so people waited longer to start to answer your question.
Zach Elwood: Mm-hmm. Mm-hmm.
Andrew Thompson: And also, in the stories themselves, those stories became more simple. So liars and truth-tellers in this, in our sample, although they reported the same number of, like- Sensory details, like I saw this, I smelled this, I heard this.
The truth tellers were able to provide a lot more details that were sort of relational to one anoth- w- relational things to one another, right? So spatial details, “Oh, she was behind the rock when I did X.” Or- Mm-hmm … yeah, you know, the dog went this way, and she went this way, and I went this way. Those things seem to be harder for liars to sort of come up with on the fly and/or also probably didn’t prepare for as well, right?
You don’t think that those are sort of important details. So those things seem to be fairly diagnostic of deception, for a better word.
Zach Elwood: Would you say, um, for that second pattern you were talking about, is it fair to say the liars’ stories were, like, more vague and more generic basically? Is, is that a way to put it?
Andrew Thompson: Yeah, exactly. And so, and so you can, you can sort of view it two ways, right? On the one hand, you’re viewing just, like, their story in and out, like, the four corners of the story, right? So you’re just looking at how complex is their story, especially as you start relating to others. The other thing that we didn’t test but is, I know, like, one of the main empirically supported cues to deception, if you will, and that I personally use a lot, is you’re looking for the number of details that you could readily go out and verify, right?
Mm. So not necessarily that you are going out to verify them, but you know, if I start giving you a vague story about my alibi and I say, “Well, I was maybe here, maybe here, but I was asleep. I was at home alone asleep,” versus, “Oh yeah, I, you know, at 8:12 I got a phone call,” or, “I called this person at this time.”
You know, stuff that you could go out and verify.
Zach Elwood: Yeah.
Andrew Thompson: Those tend to be very diagnostic of deception as well.
Zach Elwood: Yeah. I think that’s a very interesting one. And I, I didn’t, I didn’t even know that was like a aspect of study, but I, I had read about it recently. But a few months ago, an example of this was related to the Epstein stuff.
Brett Ratner, the director, had been… You know, there was some picture of him hanging out with Epstein and, and, and it had a picture of Brett Ratner with this young woman, but her face was blurred out from this party.
Andrew Thompson: Right.
Zach Elwood: And Brett Ratner, like people were saying, you know, Brett, uh, is it Brett or Brent? I can’t remember.
Brett Ratner. Uh, they were saying, you know, “Explain this, uh, photo.” And he said in this interview recently, he said something like, “Oh, that was my, uh, fiance at the time. Like, it wasn’t a, a young girl that I met at one of Epstein’s girls, basically.” And like, it immediately to me, I’m like, yeah, he wouldn’t… To me, that’s like- very likely he’s telling the truth because if he was gonna lie, he would choose something very vague that couldn’t easily be confirmed, right?
He wouldn’t- Right … say, likely say, uh, his fiance because that would be pretty easy to, uh, verify, right? And I thought that was a good example of like people saying things that like theoretically you could know to y- that, that would be possible to verify and either weak- weaken or strengthen their story, right?
Andrew Thompson: Yeah. And l- and, you know, obviously, you know, you as an investigator, you would then after the interview go out and try to verify those details, right? You do the work.
Zach Elwood: Yeah.
Andrew Thompson: Yeah. But in, in the moment, it, it is diagnostic. It also kind of goes back to some of the other findings that we had in the research, which was it seemed as if liars were much more focused on the investigator, so they tried to reference the investigator a lot.
They tried to get the investigators to agree with them a lot. Like sort of like-
Zach Elwood: Mm. “
Andrew Thompson: You know, you know what I mean?” Mm. They tried to bring them in. Yeah. Whereas truth tellers were very much just like kind of just gave the story, right? They let the story kind of stand on its own.
Zach Elwood: Yeah. That’s an interesting one, too.
‘Cause yeah, the, uh, it might be related, but I, I, I’ve seen people refer to it as like Almost like not trying to explain, but trying to, like, persuade and, like, get… You know, it’s, like, more of a focus on, like, getting them b- getting their buy-in than it is on just, like, explaining it.
Andrew Thompson: Totally. And it, and the- Yeah
it makes sense. And when you think about it in your own life, like, I mean, listen, I lie professionally. Uh, or for a lot of my professional life I’ve lied professionally, right? Maybe not so much for you. Uh, but that is a big part of it. It’s a confidence game, right? When I’m trying to convince someone I’m…
When I’m trying to convince that, you know, that sales rep that I work for a m- a midwifery practice that I’m standing up, a birth center I’m standing up, I’m doing a lot to get them to believe me, right? So there’s a part of it that’s sort of an active countermeasure or active sort of active management sort of thing- Mm-hmm, mm-hmm
that’s going on. And y- yeah, so I, I think it is a good sort of r- red flag if you’re being, talking to someone and someone is, like, trying very hard to get you to buy in with them.
Zach Elwood: Yeah. And, and, and a good liar, a good liar is gonna do that in such a way that it’s not obvious, but a bad liar, uh, or, you know, a unskilled liar might give that away that it’s, like, too actively trying to manage the, uh, the situation or the, or, or, or the other person’s perception or whatever it is.
Yeah.
Andrew Thompson: Exa- yeah. So, like, if I’m, you know, if I’m trying to recruit you as a source, Zach, I’m, I’m trying to get you to buy into me, right? I’m developing rapport. I’m trying to get you to buy in that I’m a trustworthy guy, that you like me, before I go out and feed you whatever I’m trying to feed you. Mm-hmm.
Uh, now I, you know, likely I have the advantage of time where I can do that as opposed to, you know, reacting inside an interrogation room. But that certainly is, you know… A good liar would set up that before so they don’t have to when it comes to the actual story. So you’re right in that, in that regard.
Zach Elwood: I don’t know if there’s more to say on, on that area of, of research, or did you wanna talk about, I know you did some other work related to, um, personality types, or maybe that is the personal- maybe that is the personality type
Andrew Thompson: Yeah.
No, no. That, that’s totally sep- but it kinda- That’s separate … does bleed into it. Yeah, I’m happy to talk about that as well. Um, so the, the other sort of side of the research I got to do was with a, a really preeminent sort of profiler, Dr. Gabriel Salfati, again, with this sort of BSU research collaborative relationship.
But in there, what we were looking at is there was always this premise that you could interpret someone’s actions at, like, a crime scene, and from that sort of deduce what their personality characteristics might be. Right? This is the, the basics of profiling, right? It’s the basic premise of, of offender profiling.
Several researchers have suggested you could then take that and apply it one step further and predict what their behavior would be like inside the interrogation room. But people had theorized on that, but had never sort of tried to empirically test it. Again, it’s very hard to because you, it’s very hard to get access to those kind of materials, but we had them.
So we went out to try to see, hey, if I see all these crime scene behaviors, do these crime scene behaviors, can we classify them using this, you know, various taxonomies for profiling into certain types of offender? And then when we run their interrogation transcripts through some, um, linguistic analysis software, is what we would predict their language and their behavior, would that be consistent with the characteristics we would predict from their crime scenes?
So it’s the sort of like actions to characteristics to interrogation behavior linkage.
Zach Elwood: A- and one specific model you were using was this OVP. Yeah. I can’t remember what that… What, uh, maybe you could explain that for a little
Andrew Thompson: bit. Sure. So, uh, Dr. Cantor and Young’s developed this theory based on, on narrative theory, which is sort of a general personality theory that we’re all sort of the protagonist in our own story, right?
So as we go through life, we see ourselves as the main character, and we interact with the world and with others in ways that are consistent with whatever story we’re trying to tell ourselves. The Dr. Cantor and Young’s idea was, uh, for, uh, offenders, if you look at how they behave at the crime scene, that should play through that, right?
That different offenders, depending on what story they’re telling themselves, will behave with the victim differently.
Zach Elwood: Hmm.
Andrew Thompson: And so they had developed this theory that people s- uh, offenders largely saw victims as sort of objects where the victim itself, like the person itself wasn’t important. They were just an object to use or abuse as they, as they needed.
Zach Elwood: Like for some fetish or n- need or, or whatever it was
Andrew Thompson: they had. Yeah, like the, the victim it- it was just a means to an end, right? It was just a, it was just a, a tool to get to whatever they wanted. So that was the object of the OVP. Uh, V was the vehicle. So these people, the victims were seen as important, but they were more symbolic than they were important as an individual, right?
So I don’t necessarily care that I’m raping and murdering this sex worker, this specific sex worker. It’s just sex workers generally or women generally are offensive to me in some way and I’m going to take out my aggression on this person as a vehicle for all women.
Zach Elwood: Oh, maybe that’s more of like a fetish or obsession kind
Andrew Thompson: of like focus.
Exactly. So that, that would be probably more like a- Vehicle … a fetish type situation.
Zach Elwood: Yeah, that’s more the vehicle. Yeah.
Andrew Thompson: And then the person, the P is the person where you are seeing the victim as that specific person. So this would be like your prototypical domestic homicide, right? I’m angry at my ex-wife.
I kill my ex-wife because I’m angry at my ex-wife, right? What Cantor and Young suggested was that these OVP behaviors coincided with different types of stories, right? Hmm. So the people that treated their victims like objects saw themselves as these like adventurers or sort of people on, um, these like professionals, these sort of like more high and mighty types.
And what we found was largely consistent with that. Those types of offenders acted more sort of high ego in interrogation rooms. They talked in third person a lot more. They had a lot more focus on money and power in how they talked about themselves. Mm. Whereas, like, the people who treated their victims like vehicles were more like revenge types.
They were on, like, a revenge quest essentially. And so again, consistent with that, a lot of those types of offenders inside the interrogation room were more angry. They were more sort of, again, focused on revenge, had a lot more negative sort of emotions. And then the people that saw their victims as people seemed to be a lot more sort of tragic.
Uh, they, they had this sort of tragedy motif of like, “I’m the victim here. How could she do this to me? Like, what else could I have done? I was really powerless to stop, like, this, this scenario from playing itself out.” Mm. And so we found a lot of consistency there, which was really interesting and the first time that it had been done.
But the… I think personally the really cool part about that is that because this all kind of comes back to this narrative theory, which is just a general theory of personality, right? It wasn’t made for forensics specifically. My thought was always, well, why couldn’t you apply that to a much wider array of scenarios outside of a forensic setting?
If you take this idea that we’re all the, we all wanna be the protagonist of our own story, and we will show that off in how we dress and how we interact with people and how we interact with the world, well, why couldn’t I as a private investigator use that when I walk up and try to recruit a source if I see they are acting a certain way or presenting themselves a certain way?
Could I use that to inform my pitch or my first approach?
Zach Elwood: Like, if you knew that they had, like, a personalized license plate or these various traits about them, basically profiling, that would help your approach kind of in a-
Andrew Thompson: At least at a first instance, right? So going back to the security guard idea, right?
So, like, if you’re walking up, you have to talk your way through a security guard to get to, and access to a building. The person’s got, like, a Daffy Duck tie, and they’ve got a big smile, and they’re playing with little kids. Your approach to that person might be very different from the guy who stands up and is in, like, a law enforcement style duty belt and wears his watch with the face on the inside of the wrist and, you know, has a very stern expression, all that.
Like, you’re- Your initial approach might be very different for those two people. Now it’s only a first guess, right? You could certainly be wrong. Right. But if you’re just doing a cold pitch, you kinda… It’s something to go on.
Zach Elwood: Yeah. That, that kinda relates to, I mean, something I’ve focused on with this podcast.
Uh, it’s just, it seems like the, oftentimes with this behavioral stuff, it’s like what is the use of information that we know is far from 100%? We know, we know it’s not 100%. And like a lot of times, you know, for example, in games you can use that very easily because you’re often in these situations where it could go either way, which explains why, like you can, you know, base information on poker.
You know, you can use poker tells for these close situations or you might be very confident in, in, with some tells and not others, but it just helps explain why y- you know, you can u- you can use information that’s not 100%. So in this, in an investigative or interrogation work, it, it also seems the question is like what is the value of information that is not 100%?
And, and, and you’re, the thing you’re explaining where it’s like, sure, you know some of this, these guesses about how to approach somebody ba- based on their behavior, their, their traits. It, it might be far from 100%, but if you’re in a situation where you could take, you know, it could go multiple ways and there n- there’s not much downside for guessing a way, a, a, an approach that might be better, it seems like that’s ano- that is an example o- o- of, uh, you know, some, some information that might be useful even if we know, you know, it might be wrong sometimes.
In the same way i- in the interrogation room if you get a sense that somebody’s anxious o- at a certain line of questioning, that’s a, that’s a reason to, uh, follow up with a few more questions even though you know it might not be 100%, you know, it’s far from 100%, uh, indicative of guilt or hiding something.
So, um, but yeah, that’s what occurred to me is like when, when are the situations when we can like alter our approach based on a little bit of, uh, information, and that, that seems like a good one there. Yeah.
Andrew Thompson: Right. And, and to your example, like or the opposite, right? If I, especially for someone like me who I’m not doing custodial interrogations, right?
The person can walk away from me at any time.
Zach Elwood: Mm-hmm.
Andrew Thompson: So if I’m getting a sense that someone’s anxious about my questioning, I might like mentally note that for later to come back around to it, but maybe I back off that set of questioning-
Zach Elwood: Mm …
Andrew Thompson: because I wanna keep the, I wanna keep the dialogue going.
Zach Elwood: Yeah, yeah.
Yeah. You don’t wanna drive them away.
Zach Elwood: A quick note here: this might be a good place to insert a comment about just how wide and sprawling behavior is, as a concept. The sprawling nature of it can make these things hard to talk about. A few examples of what I mean:
There’s a big difference between what’s called “body language” and the things people say. There’s also a big difference between what’s called “body language” and the actions people take; someone carrying a suspicious bag or loitering for a while in a specific area, those are much different things than what we call “body language.” There’s also a big difference between predicting physical movements in physical environments like sports, and reading body language. There’s also a big difference between “body language” and the clothing people wear, or the other choices they’ve made about how they or their physical possessions appear; our clothing and other choices about visual presentation are conscious, communicative choices; they have more in common with our verbal content than they do with small non-communicative movements of our face and body. I think there’s also a big difference between nonverbal actions that have more conscious communicative effort behind them than more non-communicative actions; for example, a smile will generally contain a lot more information than something like crossing one’s arms, simply because a smile is something someone knows they’re doing and thinks about the implications of; that is a channel where conscious communciation often happens, so it will on average contain a lot more information and be a greater source for possible deductions about intent and motives.
But all these things, despite being so different, are often wrapped up in a general “behavior” category. And this can have consequences. Because including things that are of little value in the same category as things with high value has the effect of making the things of little value seem more important.
Self-proclaimed “behavior experts” have an incentive to try to include as many things as possible in the category of nonverbal communication or “body language”. For example, I recently read this on Joe Navarro’s website: “Body language, or nonverbal communication, is more than just how your body moves. It’s about how you dress, how you talk, what you do and how you do it. It’s your character.” End quote. Lumping literally everything into a “body language” or “nonverbal behavior” bucket has the effect of making their area of focus—the body and face movement stuff—seem more impressive and more practically useful than it is.
There’s also a big difference between reading people for signs of anxiety or agitation in order to try to set them at ease, as Andrew mentioned here, and treating signs of anxiety and agitation as possible clues to deception. Getting indicators of anxiety as a clue to “back off” and take a different approach is less akin to a “high stakes” use of behavior, as it’s usually thought of, and more akin to what almost all of us do in our personal and professional settings every day, which could be termed “low stakes” situations as opposed to “high stakes” situations; most of us are always monitoring others for signs of discomfort, dislike, and similar things, for social purposes, and I think that’s a pretty banal use of monitoring and reading people and behavior; the banality of that kind of use in “low stakes” environments helps account for the many Amazon reviews on “behavior expert” books saying things like “this was all very obvious; i already knew all this stuff.”
Anyway, I thought this was a good moment to insert some thoughts on the various nuances in these behavior-related areas.
Back to the talk…
Andrew Thompson: I am a firm believer that investigation generally, and certainly like source recruitment and interrogation, it is an art. But it’s an art that can be informed by science, right? Mm-hmm. It, and it can also be harmed if you’re using sort of pseudoscience, right? So you can certainly fall into pitfalls if you, you know, are focusing on NLP or some other sort of bullshit.
It can give you, take you down the wrong path, and then you have to sort of work yourself back up from that when you’ve, when you’ve made an error. But if you’re using good techniques, empirically validated techniques, it at least gives you, you know, maybe a 60% chance of success versus just a coin flip if you were just going in purely cold and had no idea.
Zach Elwood: When I talked to Craig Silverman, the, the investigative journalist, he, we were talking about the risks of, you know, confirmation bias and, and such creeping into your a- approach and your view of people. And it’s like, you know, the, it seems like a lot of the art would be avoiding, you know, your own biases and, and trying to keep a clear head, and then B, also knowing these softer skills of how to approach people and, you know, try to, try to manage the situation to get, you know, achieve your ob- objective.
It seems like there’s so much skill in that, in that area.
Andrew Thompson: Yeah, and it’s good- Yeah … it’s good confirmation, right? It can be good confirmation, right? So like this, this sort of characteristics to linguistic style research. So, you know, the, that really stern security guard that I was talking about, I might approach him under a cer- you know, trying to bring him into the super secret spy stuff I’m doing.
But if as I’m talking to him, his language is not matching that presentation, he’s actually, you know, more of a just sort of like trying to get through the day and do his job, well, I have to adapt, right? If I just blow past all these signals that he’s showing me, then I’m, I’m really screwed. But if I can use the science to sort of be like, “Oh, wait, don’t get stuck in confirmation bias.
This guy is giving off something totally different than I first thought. Let’s, let’s try to, you know, steer the ship in a different direction and see if I can be more effective that way.”
Zach Elwood: So y- y- you had briefly mentioned it, but I was curious, in the realm of, um, bad and exaggerated ideas about what you can do with behavior, do you see people, colleagues in the investigative or law enforcement space still using a lot of those ideas?
And what comes to mind for what’s most common or, or, um, aggravating in that area?
Andrew Thompson: It, it was a lot aggravating. So, you know, I think despite what people might try to say and different agencies might try to say, Reid is still prevalent in almost every, you know, both on the private sector and in government agencies.
It’s still the way that people do things. Um, so that very aggressive, uh, sort of coercive style of interviewing and interrogation- Mm … and also a focus on anxiety and sort of these non-verbal behaviors that don’t really make any sense if you try to think about it empirically or think about it from any sort of- Historically
scientific background.
Zach Elwood: Yeah.
Andrew Thompson: You know, so I see people make mistakes all the time, and then, uh, the problem is, and then that gets into your confirmation bias, right? That’s how false confessions happen. That’s just how generally sometimes cops or PIs come to bad conclusions ’cause it’s like, “Oh, well, the person’s anxious, so clearly they’re lying to me, so I must be right, so I must have to push harder,” which makes the person more anxious.
Zach Elwood: Right. And they focus on them at the exclusion of other leads and
Andrew Thompson: such. Or focus on this one theory at, at only, at the exclusion of all others, right?
Zach Elwood: Right.
Andrew Thompson: More importantly, I think, uh, at, at least from what I, what I was doing in the defense world doing criminal defense work, is I as a defense investigator, I will use that against you 100- like, I will turn that back on you 100% of the time because what, what I’ve found is that aggressive, um, sort of threatening techniques that are used during interrogation, they get used not just in the interrogation room.
I’ve seen them used against witnesses. I’ve seen them used against victims. Um, and so if I come in with my softer skills- Hmm … and, you know, can be the first person in this whole process that’s showing you empathy- You might be more willing to talk to me, and I actually might be able to get you to recant what you said to the cops or decide that you don’t wanna talk to the cop.
I’m your friend now, you don’t wanna talk to the cops. I have your best interest in mind, not the cops.
Zach Elwood: It’s like bad cop, good investigator.
Andrew Thompson: K- kind of, right? And that’s been, you know, I know that y- like I said, I know you talked to, to Christian Corey and Maria Hartwig, who are the coolies of, of Alethea. Um, I’m, I’m also part of the advisory council of Alethea, and that’s- Hmm
my kind of soapbox to jump on, on these meetings that we have, is like, you know, yes, it’s important to sort of try to explain to law enforcement why empirically supported, non-coercive information gathering approaches like the Sue technique, like cognitive interviewing are important ’cause it gets better information.
But I’m also saying, but we also need to show them the stick, which is if you do the other thing, not only are you at risk of like getting that interrogation result sort of undermined by experts who say it was coercive, if you do that to other people, I am gonna come in and beat you over the head with it and I’m going to exploit that for my own gain.
Yeah. And then in terms of deception, like s- there’s still people that still believe in the NLP stuff. There’s, you know, the Joe Navarro type, the very like idiosyncratic, you know, crossing your arms, psychological distancing, that stuff.
Zach Elwood: Yeah. Paying way too much attention to these things that really don’t matter, that many innocent people do because they’re anxious or just are random variations in behavior.
Andrew Thompson: Yeah, exactly. And j- and again, and I think you hit it on the head, right? They’re just, they’re just random variations, right? So you can, again, you can get suckered into a line of questioning or into a confirmation bias of like a theory of the case. Um, and that happens on the private sector side, that happens in law enforcement and in the intelligence community all the time.
And it’s, it’s dangerous, right? It’s, it’s not good for outcomes. Those outcomes have real results on people’s lives. It’s easily preventable if you just like, again, even if you just think about it for a few minutes and think that like, well, why would crossing your arms really matter? Like what psychological, what brain process links to crossing your arms?
Yeah. That is indicative of lying and not 18 other things.
Zach Elwood: A quick note here. On that recent video of mine discussing what i see as the deceptive nature of what “behavior experts” like Joe Navarro and others sell, some people will say “what do you mean; Joe Navarro points out that body language is not good for detecting deception; he points out that research in his books and trainings” And yes, that is true, but you have to understand that “behavior experts” want to, as the saying goes, have their cake and eat it, too; they want to point out the research, because they have to point out the research, at least if they want to be perceived as somewhat legitimate. But at the same time what they are selling is the idea that you can see through to what people are really thinking and feeling. So whether they state it directly or communicate it more indirectly, they are still selling deception detection.
I’ll give you a couple examples of what I mean from Joe Navarro:
In his most well known book What Every Body is Saying, the promise of deception detection is obviously present in the book and book promotion itself. The back of the book reads: “He says that’s his best offer; is it? She says she agrees; does she?” The description of the book says that you’ll learn quote “The most powerful behaviors that reveal our confidence and true sentiments” unquote. Yes, in that book Navarro mentions that research shows that deception detection is very hard, but the overall message is that it’s still possible. The idea that comes across is that noticing clusters of behavior, and using proper baselining, and such, can help you become skilled at detecting likely deception. This can be seen as a more roundabout and opaque way to say that you can use various small nonverbal clues to detect deception; I see it as just laundering the idea a bit. You aren’t quote “detecting deception”, you’re detecting “signs of possible deception,” basically. And the thing is I don’t even disagree that nonverbal behaviors can help show spikes in anxiety and agitation, and that those things will sometimes be tied to guilt and deception; the difference is that I would say that the weakness of nonverbal stuff for such purposes makes them almost irrelevant in high-stakes, adversarial domains, whereas people like Navarro would say that nonverbal behavior is hugely relevant and important in such domains, and that’s where I see the falseness.
Here’s some more copy from Navarro’s website jnbodylanguageacademy.com, regarding his expert body language course:
“Truth and Lies: Use your new skills to help determine credibility. Be guided by Joe Navarro as he educates you on the skills and pitfalls of determining whether someone is being truthful or lying to you, with opportunities to apply your new knowledge straight away and reflect on the outcomes as part of your studies.”
End quote.
Note that Navarro’s wording here is purposefully vague. He doesn’t straight up say he’ll teach you to detect deception. He knows that would be a bad thing to do. But still, he communicates that his teachings will help you in that area. Also note he uses the phrase ‘determine credibility’ here. The term ‘credibility assessment’ is a common term these days, and I think it’s a good one, as it encompasses many more concepts than the term deception detection. But also note that another benefit of talking about assessing ‘credibility’ is that it makes a similar claim to detecting deception without using the word ‘deception.’
One foundational myth that Navarro and other alleged “behavior experts” spread is that our true motives are often revealed by our face and body; that we are powerless to help revealing our inner thoughts and emotions. For example, you can find the following on Navarro’s website: “Our emotions and intentions are expressed through the body, and they are very hard to fake. This means that body language signals are highly reliable and full of useful information. Body language can give you information about what the other person is thinking, feeling or intend to do.” end quote. This oft-repeated myth, that our bodies and faces are constantly revealing our true feelings against our will, can be seen as a way to imply that you’ll get deception detection skills from such training. They are selling the idea that you will be able to see beyond the curtain, see what makes people tick, see what they’re really thinking and feeling; and even if people like Joe Navarro avoid saying it directly, the message is still the same as if they’d directly said ‘we’ll help you read when people are not being truthful.’
It’s worth mentioning that some of these kinds of claims are spread and promoted by some academics who are selling similar products. I recently bought a course from David Matsumoto’s Humintell company; it was called “Detecting Deception Advanced Training Package” and it costs $299 right now on their site, discounted from $387, and it was just very very bad. It was straight up garbage, in my view; just some micro-expression training, coupled with a surprisingly low-effort presentation that vaguely tied micro-expressions to deception, even though no research shows this to be valuable. Even the research Matsumoto would reference to try to defend this stuff shows such a small effect size as to make it practically useless, and that’s not even getting into whether such studies would even replicate. I want to make an episode about just how bad this Humintell course was, but my point here is that, whether they’re doing it directly or indirectly, “behavior experts” are selling people deception detection; that is a significant reason why people buy their content and courses. And I think “behavior experts” want to have their cake and eat it, too; by mentioning the science they can seem sophisticated and intelligent, while at the same time they can sell you the myth of being a wizard at seeing through people and therefore being a better deception detector.
Anyway, sorry for the long notes. Part of this is just me working through these ideas myself for this book project i’m working on.
Back to the talk.
Zach Elwood: You would think most people would, would realize, like logically it doesn’t hold much water. But yeah, it is kind of amazing how many people still put so much stock in those little things, but,
Andrew Thompson: um- But it’s, it’s easier, right? I think at the end of the day, I think a r- a lot of why this continues to exist and even continues to exist in people that theoretically should know better, professionals who do this all the time-
Zach Elwood: Mm-hmm
Andrew Thompson: is it’s a lot easier to aggressively question someone than it is to think about their behaviors and how they present themselves and come up with a profile based on that and make your pitch attuned to that profile, and then adjust it as you start interacting with it. Like that is… Well, I’m, I can tell you from personal experience, that is exhausting work.
It is exha- it is mentally exhausting to do that, but it gets you better results and it, and especially in the private sector, you know, PI work is the ultimate meritocracy, right? There’s, there’s a lot of PIs out there, if you don’t get results, you will not be hired again. Hmm. Or if you work, uh, if you’re a W-2 employee, like you, you are employed directly by, uh, a law firm, like you work in-house at a law firm or, uh, wherever, an ana- like a hedge fund or something, they’re not gonna tolerate you being wrong because being wrong equals money, and if you’re costing the company money, they’ll just churn to the next guy or girl.
Zach Elwood: Hmm. Yeah, much more pressure to, uh, get results and in the m- uh, uh, in the short term. Right. Yeah. On the, uh, topic of, uh, people being, uh, interrogators being more aggressive than they should be, it’s hard for me to get a sense of it because I just, you know, I’m pretty, obviously pretty distant from it. But you know, I, I, I watch a good amount of interrogation footage on YouTube and such, and it does seem…
I mean, I see a good amount of police, uh, taking more You know, very, uh, rapport building and, and gentle approaches. Would, would you agree that it has improved in, in, in, in that, in that area a lot, even if you also think you see too many people taking hard and, uh, aggressive approaches more, more than they should?
But I, I’m curious, do you, have you seen a big… Do you think there’s a, been a big up- upswing in, like, the quality of taking more rapport building approaches?
Andrew Thompson: Yeah, I think so. I can only speak to, like, the, the areas I’ve operated in, which in, you know, in Canada and then here in New York, more progressive, larger agencies that probably are faster, right, than maybe some places in smaller communities.
Right. Um, so I’m sure there’s some variability there. Um, but absolutely, you know, you, you can see that in the training regimes, you can see that in, in just watching, like you said, watching interrogation tapes. There are certainly… And, you know, and also with groups like Aletheia who are out there trying to sort of preach and, and train people in better ways of doing things, and bringing researchers and practitioners together so it’s a more, uh, united front to come out and sort of do this.
Um- Mm-hmm … but yeah, I, I certainly think it’s better than it has been, and continues to get better. Um, again, I think that it’s honestly, it’s just, it’s the timing, right? I think that the more pressure police officers face, the more they’re gonna go back to the well of these more aggressive techniques and more fear-based techniques, whether that’s deception detection or, like, just questioning in general.
Because again, they’re, they’re faster and, and they are effective. Um, sometimes someone would argue they’re too effective, right, is the problem.
Zach Elwood: Right. Hence, yeah, the, the, the false confession concern. Yeah. Uh, right. Uh, I was curious, um, it do- you know, in this area, we, we briefly touched on it, but it does, it does seem like There can be some polarized views about, you know, the use of behavior in general, in the sense that it’s understandable why people are concerned about using behavior because we know, A, there’s a bunch of bad ideas that are out there, and then B, they’re concerned about people, uh, being way too certain about these things.
For example, the scan thing you, you mentioned, uh- Right … the guy who did that, does that, uh, scan stuff acts way too confident about like this means that, this means that. Uh, I also think it’s possible to be like too biased against the use of any behavioral things because I do think, you know, we– there are things that we pick up and I think to me it’s like as long as you’re ensuring that you’re not being overly confident and you’re thinking carefully about like, “If I’m wrong about this, what’s the downside of me being wrong?
And what are these situations where I can, you know, where I would actually get some use from like making a read about something,” right? I, I, I think it’s possible to be both way… Value that kind of stuff way too highly and be too certain about it and also like completely devalue it and say like, “Don’t use it at all.”
Would you agree with that? And, you know, uh, uh, ’cause I do think, you know, there’s, there’s all sort- like you, you said studying the verbal kind of statement, uh, uh, uh, patterns in, uh, people. I mean, there’s all sorts of ways that we get information about people, whether it’s verbal or like, you know, someone’s overall demeanor in a, uh, you know, how they’re, how they’re acting.
Uh, and it might be that we, m- a lot of the times those things aren’t actionable, but occasionally they can be a bit actionable. Uh, like a few examples we, we talked about, but I’m curious if you’d agree that, you know, there can be kind of this polarization of like thinking about behavior.
Andrew Thompson: Yeah, absolutely, right.
I think kinda it goes back to that art versus science th- th- feeling that I touched on, right? Um, you know, I think you have practitioners, a lot of practitioners at least traditionally would say like, “No, listen, it’s an art form. You have to be able to… You know, you can tell idiosyncratic, you know, movements might be, might be relevant.”
And then you have researchers say, “No, it’s a science and it has to be empirically valid.” I, I think it is in the middle, like you said, right? I don’t think you should be– I think you should be trying to focus on behaviors that have some sort of empirical basis, right? If you’re focused on something that’s just totally out of no- like out of left field Right?
The, the, like, you know, the eye movement stuff. Mm-hmm, mm-hmm. Right? Well, there’s, there’s absolutely zero base in that. So, like, relying on that in- for anything is just, it’s gonna get you bad outcomes. Right. Regardless of how focused you are- Yeah … on it or not.
Zach Elwood: Yeah.
Andrew Thompson: Um.
Zach Elwood: Yeah.
Andrew Thompson: But yeah, some of these other behaviors, the, like, the, the cognitive load type behaviors, and there’s certainly non-verbal behaviors that go along with that.
Um, some of the, the paralinguistic stuff, even some idiosyncratic stuff, right? You know, I think it’s all about context. If I’m first meeting someone, is, you know, whether they cross their arms or push themselves away or how they, like, align their, their body to mine, is any of that gonna be relevant if I’ve known you for two hours?
Zach Elwood: Right.
Andrew Thompson: Absolutely not.
Zach Elwood: Mm-hmm, mm-hmm.
Andrew Thompson: You know, I, I don’t know about you, I’ve been married a long time. Uh, I pretty reliably can tell when my wife is lying, right? And a lot of those behaviors probably are idiosyncratic to her. It’s all gonna be relevant to context. Um, but even anxiety, right? I think that everyone, you know, I think especially on the scientific side of things, people say, “Well, anxiety has nothing to do with deception.”
Well, again, it’s context dependent. Right. Right? If the FBI knocks on my door after we’re done this interview and says, “Give me your phone, I wanna look through it,” I’m probably gonna be pretty anxious. Right? I think that’s a reasonable response. Mm-hmm. If, you know, my wife later tonight says, “Oh, hey, I left my phone upstairs.
Can I, like, I have to call, you know, my patient. Can I use your phone?” And I am suddenly, like, shaking and fumbling for my phone.
Zach Elwood: Giving excuses, yeah.
Andrew Thompson: Right. Yeah. But again, the, the danger of that is, like, well, but why? Right. Right? You know, maybe I, you know, my wife’s birthday’s coming up soon. Maybe it’s the, that, like I know that I have an open tab that has like her gift on it or whatever, right?
Um, so it’s, it’s about balancing it, but I don’t think it’s useless. I don’t think– I think- Right … all data is useful.
Zach Elwood: Yeah, it’s data.
Andrew Thompson: It’s just you have to weight it appropriately.
Zach Elwood: Yeah. That’s what, you know, I, when I think about these things, it’s like when we talk about quick reads of people and whatever it is, they’re deductions, and like some deductions can be very bad.
Like if you’re making a dedu- deduction based on some really small and highly variable piece of data, that- that’s gonna be a bad deduction. But like the more logic you can put behind the deduction, the stronger it’s gonna be. But as, and, and you also have to, you know, recognize that your own fallibility and, and recognize that this, this might be a deduction that I, I have like, you know, 55 to 60% confidence in versus like, you know, 90.
Like any deductive re- logical reasoning, it’s like you’re, you’re chaining together ideas and tho- those ideas can either be very bad or very good or have limited, you know, reliability and all that.
Andrew Thompson: Yeah. And I think, and, and it’s important to sort of keep taking data points. And like I, like I was sort of saying, right, on the approach, right?
I mean, listen, it happens all the time to me, right? I make an approach to recruit a source, and this is the theme I’m developing ’cause I think based on their background or their, their language that this is gonna be the, this is the narrative that is gonna most resonate with their story, and that’s how I’m gonna get them on my side.
And then it turns out like, oh no, that is not working correctly. Or like, oh, their, their body language is changing and their language is not matching what I would expect it to. Mm. That’s all good data for me to like, oh, I need to change my approach now because I’m just digging myself deeper in a hole.
And listen, there’s, there’s, I have lots of examples of where I’ve failed at that, right? You get locked into this story and it ends up screwing you, so you, you have to be open-minded.
Zach Elwood: It’s really good… It was a really good point you brought up too about how reading anxiety or irritation in someone i- in your, you know, in, in the private investigative line of work, it’s almost like you might approach it theoretically the opposite of a cop would, where you might wanna back off and be like, “I don’t wanna blow,” you know, “I, I don’t wanna blow this or get them to stop talking to me, so I might back off and return to it later.”
You know, you have to, uh… That, that’s a, that’s a really interesting point about using those indicators of like, “I’m bothering… I’m getting some signs that I’m bothering this person in some way or another.” Yeah.
Andrew Thompson: Yeah, or you need to approach a different way, right? There’s a really, um, there’s a set of techniques that are been researched recently called the Scharff techniques, and they’re actually very old techniques.
It’s, they’re based on a… I don’t know if you’ve, you’ve ever heard of Han Scharff.
Zach Elwood: Yeah.
Andrew Thompson: And that’s a big part of the Scharff technique, right? Is like you don’t, you purposely don’t react to very important information that the person’s revealed. You don’t let them know that it’s important to you.
Zach Elwood: Right.
Right? Because the goal, the goal is to make them think that you know everything kind of a thing, right?
Andrew Thompson: That you either you already know the information, and so why would they not tell you? And also, like, it encourages them to continue to talk to you more, right? So, like, for example, I rarely, to the best of my ability, I rarely try to take notes when I’m- Hmm
interviewing someone face to face.
Zach Elwood: Yeah.
Andrew Thompson: Because that just subconsciously the person will tell if I’m really interested.
Zach Elwood: Yeah, and it kind of ups the stakes, too, where it’s like… Yeah, it’s the sa- it’s the same way, like re- reason that many people don’t wear sunglasses in poker ’cause it makes other people, puts them on guard more.
Zach Elwood: A quick note: I realized a much better analogy in poker is much more direct; many serious players avoid taking obvious notes when playing, as they don’t want other players to know how serious they are. One nice thing about mobile phones, in the poker realm, is that you can take notes without anyone ever knowing what you’re doing. When I would take notes on my phone when playing poker, I would go out of my way to make sure people didn’t know I was doing that. Back to the talk.
Andrew Thompson: Yeah, and, and, you know, and that’s important for both, like get- eliciting information in terms of, like people engaging in countermeasures. Even, even for an extent when, you know, you’re interrogating someone and you catch them in a lie. Well, why would I let that person know that I have them, right? It’s, I mean, that kind of goes part of SHARF, it’s also part of the, the strategic use of evidence technique that Dr.
Hartwig developed. Hmm. Lead them down the path, let them sort of lock themselves in, don’t reveal anything, and then start working it back. Right. Um- Right … I think that those are probably the, the more effective ways to go, right? Um, and then terms, in terms of elicitation, I think people, like to your point, people can start feeling really defensive or really sort of scared once they, if they start realizing like, “Oh, oh shit, I’ve, like I have the, the keys to unlock this multi-billion dollar securities fraud.
I don’t really wanna be involved in this. I don’t really wanna be, you know, dragged into this like crazy, you know, titan versus titan lawsuit or, or investigation.” So all of a sudden I’m very keen on something you said, maybe you start backing away from it, whereas if I just ignore it- Right … or ignore that you’re anxious about it and attack it a different way later on, you think you’ve sort of, you either, you don’t realize that it’s that important, right?
Mm-hmm. Mm-hmm. And certainly some of the best witnesses I’ve ever had and best sources I’ve ever developed never really knew that, that what they gave me was of any value whatsoever until it came time to sort of, you know, confront them about it at trial or, or all of a sudden that information appears in a, in some sort of pleading, right?
Zach Elwood: Right. No, yeah. No, that’s a very, yeah, very good points about that. I, I was kind of curious how AI had maybe changed your work in the last few years.
Andrew Thompson: So AI is huge right now. Uh, I’ve, I’ve seen pitches, like I’ve personally been pitched on stuff. I’ve also sort of heard through, you know, just through the industry about a lot of pitches, and it’s, it’s being pitched for everything, right?
Um, there’s pitches on, “Our AI can listen to earnings call transcripts and figure out when CEOs are lying about whatever res- financial results.”
Zach Elwood: Like using, uh, is that like pitch anal- uh, pitch analysis or is that like statement analysis type stuff?
Andrew Thompson: It’s usually state- it’s like content analysis.
Zach Elwood: A quick note: for a discussion of the advances in AI and machine learning in the deception detection realm, and why machines are better at doing that than humans, check out a recent talk I had with Bennett Kleinberg. I think it’s clear that there can be major recognizable patterns in what people say, and what they don’t say. One of the biggest most practically useful things in the area of listening to what people say is just noticing when people actually answer questions, and if they’re being specific or if they’re being vague or misdirecting to something else.
Andrew Thompson: You know, so I’ve seen those pitches. I’ve seen pitches, uh, I think what’s I think is actively being used in places for insider threat analysis. So Fortune 500 company has this system that sort of scrapes everyone’s emails and sees how their language has changed, and based on different characteristics and changes in that language, it detects if someone might be an insider threat, someone might be willing to sort of give up information or leave the company and potentially sell secrets.
Zach Elwood: It’s like a Minority Report for, uh, disgruntled employees, huh?
Andrew Thompson: K- kind of, right? Um, you know, in my opinion from the pitches I’ve seen and just from my knowledge of, like, what AI is capable of, I don’t think it’s there.
Zach Elwood: Mm.
Andrew Thompson: I also don’t necessarily think it’s ever going to get there. The thing, the problem with especially lie detection in particular, ’cause that’s more my area special- uh, specialty, but is that the behaviors, like, those passive cues are so not detectable in many ways.
They’re not, they’re not strongly correlated.
Zach Elwood: Mm-hmm.
Andrew Thompson: That I don’t know if AI will ever be able to, like, with any sort of sense of, of actionable confidence make those determinations.
Zach Elwood: Yeah. It’s the actionable thing ’cause it’s like h- h- how, how reliable does something have to be before it is actionable? In a lot of cases it has to be quite high to be actionable.
So yeah, if it never crosses that threshold, it’s not gonna be very useful. Totally. Yeah.
Andrew Thompson: That being said, what I personally have used AI for and what I think it could become very effective at as the AI develops is sort of taking all this volumous amounts of information that you develop during the course of an investigation and comparing that to, say, individual statements, right?
Hmm. So I interrogate, I question the defendant of a case about something. AI compare that to all the other witness statements, all the physical evidence, all of this. Tell me what inconsistencies you found, tell me what’s consistent. Point, you know, develop, show that. Yeah. That seems more what it’s capable of.
Zach Elwood: Yeah. No, that’s, that, that’s what… I mean, that’s hugely valuable i- in the work, i- in my career in, uh, technical documentation for, for software. So for example, you know, one application is, like, we’re putting out a new document and we wanna make sure, you know, parse the whole documentation site and make sure that there’s nothing contradictory in this, you know, with the whole doc site.
So it’s, like, really great for, uh, analyzing and parsing large volumous, uh, content or, or data or whatever it is and saying, “Find discrepancies or find contradictions or find conflicts.” Uh, yeah. It, it, it’s great for that, and I can imagine, like, in a big legal case where you have a whole bunch of information and that, like, as a s- as opposed to, you know, getting paralegals or, or, you know, uh, whoever to, or, or the investigator to go through all that to compare things.
It’s like, yeah, AI is, is great for that kind of thing. That makes sense.
Andrew Thompson: Yeah. Yeah. And, and that is, that is exactly a use case that is actively being used right now in cases I work, right? Which is it will alert me if from, you know- a public co- public companies have to file these massive, you know, hundreds of pages financial reports, and a lot of the language, like the, especially like the risk language, right?
They give warnings about what could happen in the future, is boilerplate. It, it doesn’t change quarter to quarter, and one of the things that AI will do for me is instantly tell me, “Oh, hey, company ABC changed its, its language, its risk warning about its, you know, potential exposure to criminal investigation.”
Well, why did it do that, right? If you f- if a lawyer has suddenly decided to change that language, it’s for a reason. Mm-hmm. And if that, if that change is negative, if it implies more risk, then, you know, maybe there’s some, there’s some smoke there, right? That’s something to pursue.
Zach Elwood: Right.
Andrew Thompson: And you’re right, that would take, you know, hours of incredibly tedious work to manually do, and AI can just do that instantaneously.
Zach Elwood: Mm-hmm.
Andrew Thompson: And like I said, and same with, you know, take the statement and show me how it compares to all the other witness statements about this same murder, right? About this same-
Zach Elwood: Right …
Andrew Thompson: fight that happened.
Zach Elwood: Right. No, that’s-
Andrew Thompson: Point out the inco- inconsistencies.
Zach Elwood: Yeah. No, that’s, that, that is, that is the coolest thing about AI, yeah, what it, what it, what it can do with that, that kind of thing.
Um, this has been great, Andrew. Do you wanna mention anything else before we, we end it?
Andrew Thompson: No, no. Thank you for having me. This, this is always, it’s always fun to talk about. I could talk about this stuff forever. Um, you know, the, yeah, the, the only thing I would maybe, I would maybe want to just point out and point out to any, like, current private investigators or, or people that are thinking about getting into the field is I would just really, really stress this idea that anytime you recruit a source, whether it’s just through persuasion, right?
Telling someone who you, who you are and why you want the information and convincing them to give it to you, or through manipulation, approaching them under some sort of cover identity and manipulating them into giving you stuff. Anytime you approach a, a source it’s a risk to them, right? It’s, you know, in the m- in the most dire cases it can be a risk to their safety.
In the case of, like, the security guard scenario I was, I was talking about earlier, it could be a risk to their job. Mm. Right? Mm-hmm. Um, but it is a real risk, and I think that PIs and cops tend to downplay that just ’cause of how, how common it is to use those techniques. Mm. Um, but you really have to think about what you’re doing and make sure that the potential payoff is gonna be worth it.
Mm. Because that is, you know, when you’re applying all these techniques, you can be very, become very effective and get people to do things they wouldn’t other normal i- normally tell you.
Zach Elwood: Right.
Andrew Thompson: And there is ethical concerns with that, that you have to think about.
Zach Elwood: Ah, yeah. Yeah. Thinking about the other person, yeah, being, yeah, being ethical in your, in, in, in the pursuit of your objectives, yeah, yeah.
Andrew Thompson: Yeah, and at the end of the day, like, you… It’s your job. You have to, right? Like, at the end of the day if I have to burn a source because they have information about a multi-billion dollar securities fraud, well, I’m gonna burn the source because that’s my job. That’s what I’m getting paid to do.
Zach Elwood: Mm-hmm.
Mm-hmm.
Andrew Thompson: But you at least have to own it and recognize that that is-
Zach Elwood: Right …
Andrew Thompson: that’s where you’re living, right? Yeah, recognize. You’re living in that gray area.
Zach Elwood: Yeah, you have to at least see the impacts and weigh the- Right … weigh the outcomes and such, yeah.
Andrew Thompson: Exactly.
Zach Elwood: Yeah. No, this has been great. Thanks a lot, Andrew.
Andrew Thompson: No, thanks so much, Zach. I appreciate it.