Better Tools, Better Judgment?

Hello everyone and welcome to the Human Behavior Podcast. This week, Greg and I are talking about Flock cameras and automated license plate readers, but there's a much bigger conversation here. Because whether it's a camera flagging a vehicle or your GPS telling you where to turn, that information influences what you believe and what you do next. When it shows up on a screen, it's pretty easy to go, okay, that must be right. So we get into what happens when that information is wrong, how relying on these tools changes our skills, and who's accountable for the decisions that follow. There are real benefits to this technology, but we still have to question what it's telling us. Thank you so much for tuning in. We hope you enjoyed the episode. And don't forget to check out our Patreon channel for additional content. Please consider leaving us a review and, more importantly, sharing it with a friend. Thank you for your time. And remember, training changes behavior. All right, and we're live, Greg. Hello everyone, and thanks for tuning in. Excited for this conversation today. We've got a bunch to talk about when it comes to different technologies that people use. Specifically, we're going to be talking about Flock and automatic license plate readers, but there's a larger conversation here. The reason we're using Flock is because it's the first of these technologies that has a household name. What we're getting into is about tools and the human decision-making cycle — how these things influence how we behave, how we think, how we act, and how we talk about these technologies. They're all getting cheaper and rapidly adopted, but there's a whole bunch of issues here too, because not only is this about performance, human behavior, and technology integration — there are civil liberties, safety and security versus privacy, mass surveillance, government tracking, digital footprints. So much in here. A lot of people just throw out wild, bold claims or want to be salacious and get clicks, and it doesn't help the conversation. They're not nuanced at all, they're overly simplified. So I think this is a good topic to dig into, and we've got examples from people we know, our listeners, and folks out there. I also want to mention up front, for everyone who has not been to the new Arcadia Cognerati website — you can find the links in the episode details. Arcadia Cognerati is our sponsor, Greg. Full transparency for those just tuning in for the first time: that's our company. But there's a cool podcast page with a search function, so if you're looking for specific topics, it's all in there because it's got the transcripts it can access. So check out the podcast website and our new brand.

It's a really cool brand. It's a really cool site. It's so easy to navigate, even I can look around on it.

We've gotten some great feedback on that. So for those of you who have been following us for a while, please check it out. You'll enjoy the new branding and logo stuff. Shout out to Clark for nailing that one. But here's the thing, Greg. What happens when machine-generated information enters a human decision-making system and begins influencing what people notice, what they believe, and what they do? Flock is the current example, but this is a larger discussion that applies to all ALPR — automatic license plate readers — as well as AI, GPS, automation, predictive systems, robotics, autonomous vehicles, and medical decision support systems. In general, we would say that technology should increase the capability of the operator without replacing the judgment, skill, or accountability of the operator. Because what starts to happen is it goes from a technology that supports a person to people just supporting the technology. The companies that build this stuff may spend a lot of money on user interface and how people interact with it, but it's not operational in a real sense. It's, here's how the system works, here's where you click, here's where you enter the information — not, how do I operationalize this in real environments? Not enough time is ever spent in that area. So while we're talking about Flock — this technology has existed for years. Greg, why does Flock suddenly seem to be part of the public conversation? Why is it a household name now?

Because of the speed of information, first of all — it wouldn't have made multiple newspaper or magazine articles back in the day. But anytime machine-generated information enters a human decision-making system, it fundamentally alters the cognitive mechanisms that dictate perception, belief, and action. It disrupts how we normally construct baselines, evaluate risk, and process environmental reality, and that scares people. Not too long ago, they came up with a fancy new thing called a laser rangefinder. First it was only for military applications, then every golfer, hunter, and sniper used them, and now it's common — you're not going to see an article about it. Your argument at the beginning was that technology should increase the capability of the operator without replacing judgment. Well, what does the laser rangefinder do? You don't notice you don't have it until you don't have it. For a hunter, it's great because you take a more ethical shot — uphill, downhill, at distance, in rocky terrain, you can get fooled on how far that animal is and injure it. Same thing with the sniper: you'll be more lethal if you know exactly the dope on that scope. And then we come to golfing. Now I'm taking out my judgment and impacting my skill, and it becomes a competition. So the argument became not about ethically killing animals or people at greater distance, but, hey, you're cheating if you use that to determine distances. So what's different now? Machine-generated information — a human has to process that information — and it's cognitively different. It doesn't come at us the same way because there's no problem solving. The sense-making, the decision-making, is inhibited.

Great example, especially with the laser rangefinder, because it's different in all those contexts. The hunter-sniper versus the golfer is a good example because it has a different impact. With Flock camera systems, they didn't invent automated license plate recognition. What's changed is the scale, the accessibility, the effects, the visibility, the public awareness. It's now a recognizable name and technology and makes for a useful case study for a much larger question about technology adoption. People get into the benefits and the negatives, but what should the relationship be between the operator and the technology? We were just at the National Tactical Officers Association conference and met some amazing folks. I spoke with a very highly experienced state trooper who specializes in interdiction. His concern was not that Flock doesn't work or that there aren't benefits — his big concern was that younger troopers were relying on this technology instead of developing the traditional interdiction skills that he spent years learning and teaching. He saw it as something that, across the generational gap, changes their behavior and their level of competence. What do you think he was recognizing there?

Technology alters what people notice. In human behavior and pattern recognition analysis, that's called attentional hijacking. You can only attend to so many things at one time — you can only use 100% of your attention. So when something takes over a big part of my situational awareness, I have to surrender part of my pattern matching and my sense-making skills. We don't need to recognize patterns or make sense of incoming stimuli because this external synthetic media inputs a template straight into our brain, and that takes out a critical step. The first time, not so much of a bother. But by the seventh time, your brain just defers, and now that ability atrophies. If we didn't have it with us, we're not going to be able to create it. In the old days, there used to be a hot sheet on the visor of every patrol car. You read it on duty and did a pass-down log with it. On that hot sheet were important felony warrants and stolen vehicles you might run into — short and to the point. You had to look, Brian. You had to take that, read it, make sense of it, and then actively look for it. And that's what that trooper is talking about. That's what's lacking in those troopers and officers who get to rely on the system.

Let's dive into that a little bit. He was from Colorado, the trooper I was talking about at this conference. What he's describing is: let's say you're doing that interdiction job, you're watching, you're observing, you're reading baselines, you're looking for anomalies, you're seeing incongruent signals. You're developing this picture. Then your brain creates an explanatory storyline — here's what I think it likely is — but you have to do the math first. You have to check things, establish reasonable suspicion, and it has to meet certain criteria. This is all happening in nanoseconds, unconsciously. And then you act on that story. It's not unlike driving down a rural state road at 50 miles an hour and all of a sudden you see a deer poke its head out of the tree line. You might lock up the brakes and slow down. Your brain didn't just say 'deer, hit the brakes.' It said: that deer looks like it's going to cross the road, if it crosses I won't stop in time, it's going to destroy my vehicle, I could get hurt, so I'm going to slow down. That's what it actually did. Now replace that sense-making — which happens in nanoseconds — with a screen giving you the story. My brain goes, okay, that's the story, and it locks in on that as if it has a higher level of authority, more robustness than something I created from my own observations. And that's not always true. When a stolen vehicle pops up on a screen, we go, well, this is the machine they gave me, this is what it says, so I'm acting on it. If I had done that on my own, there would still be some doubt — I'd still need to articulate it before making a decision.

The process you described — the deer, the cognition — that's exactly right. The problem is if a machine is doing that for you, it's giving you a story, and your brain will read that story and start accepting that data even if it's wrong or flawed. That's where the rub is. The guy we were talking to, Brian, great guy, he's retiring in just a couple of months — he does interdiction for Colorado State Patrol. We've had interdiction folks from FDLE, a lot of them come through because these skills help them. I read a great book once — get your hands on a book and read it once in a while. It talks about the salience effect. Our brains inherently respond to the most salient, most attention-getting cue in our environment. Machine-generated feeds hijack our cognitive process and cause us to miss the subtle, critical ground-truth anomalies that make us want to pull over a vehicle, that give us reasonable suspicion — because it's already been done for us. That's why we get fat: instead of hunting and producing our own food, we open something and eat it without thinking about all the sugar and fat in it. Always be suspicious of machine-generated probable cause, because you weren't there. You didn't do the work. You didn't have a robust, fidelity-filled decision-making process that started with sense-making and problem-solving. Continuous electronic comms bombarding us will lead to the erosion of sensory skills over time. Someone's going to argue that may take months or years — but can a cop avoid that? No. So you buy it when you buy that gear. It dampens your micro-observation abilities. It tells you: don't worry about the thing on the left or the right, this is more important. How many times are you going to fight with your brain? Cognition has kept us alive this long. I get using AI to write a better story. But creating the story for you — that's the rub. Your brain wants to be an active part. And conservation of calories doesn't apply to survival tasks. Your brain conserves calories specifically for survival, and it considers a stolen plate or a felony warrant serious enough that it might take your life. That's why we're on this.

Even with large language models and AI systems, they're already showing that people who constantly use them and just say 'I'll use this for everything' show a huge reduction in cognitive development, a huge reduction in cognitive use. It makes you dumber. I remember talking about this on one of the episodes — right when I moved to Minnesota, I was using Google Maps for everything because I didn't know where anything was, brand new area. After a month I realized I didn't know where anything was, didn't know the street names. I'd never had that experience before, because in areas I'd already known, I used it for traffic routing. I literally had to turn it off: look at the directions, find the street name, set the phone down, drive, and go find it myself — even if I missed a turn and had to go back. That builds the knowledge up. That trade-off is never looked at when we start adopting these tools.

Jaeger and I were out near the La Brea Tar Pit area doing an operation. We had radios, cars, all that gear, and I got hopelessly lost trying to find the area. We were building what became Combat Hunter at the time — how do you find certain things interesting, why is this more interesting than that? We were working with cops on the streets. I wasn't going to embarrass myself and radio in asking for directions, so I pulled over at a bus stop and walked up to a woman who had some bags. I asked her how to get there. She said: go down, when you see the In-N-Out take a left, go about halfway down the block, and there's a big car wash on your right — that's the turn. She gave me landmarks. She didn't give me GPS coordinates, streets, and distances. When I got back in the car, Jaeger said: there were 17 people at that bus stop — why did you go straight to her? I said: she's a woman with her groceries, taking the bus. Who else knows the times, the routes, the distances? She doesn't want that ice cream to melt. That's what happens when we reason and think things through. Calculators make us dumber. I see it now at airport restaurants — they calculate the tip for you. That takes your thought process out of it entirely. How do you know that's right? How do you know somebody didn't have their thumb on the scale?

People take data at face value — 'this thing told me, this study said so.' It's like five out of six people enjoy playing Russian roulette; that one guy couldn't give his opinion on it. Can technology improve someone's performance before they've actually developed the underlying competence? Can it act as a bridge or a band-aid, something that helps someone out?

You and I didn't understand the chemical composition of Curlix before we had to use it in combat. We were happy to use it because we saw a difference. The performance, meaning the use case, was strong enough that we didn't have the competence going in. We just knew that this was likely going to work. And for people that don't get that reference, let's talk about optics — thermals, night vision, spotting scopes. We didn't have to be good at them. We just had to hold them up to our eye and go, wow, and you could learn on the fly. Were there mistakes because I had no depth perception? Yes. But the performance was there before I had the competence. And a callback to the anti-lock brakes — when we first got airbags in the scout cars, you didn't have to be very smart because the airbag took over when you made that error. It went off in your face and you went, wow, I don't want to do that again. I just crashed, and it took over right at the moment of the crash, so I learned things afterwards. Those are good things. But most of those, except for thermals and that type of thing, are intermediate to low-end technology rather than the high-tech we're seeing now. My fear — only because I'm an old guy — is that it goes too fast. When processing goes too fast, it outpaces our cognition, and when it does, we fall for things. We'll believe it, just like a magician's trick, until somebody slows it down and we go, holy crap, I never considered that. Well, if you don't consider something, it's not in your reasoning, your sense-making, or your problem solving.

So that kind of brings us back to the flock cameras and stuff. No one's asking what does this technology actually know. Because if something like this generates an alert, what does the machine actually know? It could know certain characteristics about a vehicle with a certain plate at a certain place and time, but another database may associate that information with something else. It doesn't say anything about who is driving, why they're there, whether the information is current, whether the database entry is correct, or when it's been updated. The time and location of stuff is important. Now it becomes, oh yeah, this was absolutely what they were doing here — well, now it's a girls' school. Whoops, we didn't update the database on this before we fired away with this technology. So these conclusions that people take at face value still require some sort of interpretation.

That's essential. One of the things that Brian does in class all the time — Brian's an incredible instructor and he's developed some great strategies to laser focus in on stuff. When we're talking about a dispatch and how information can be mistaken, Brian says, okay, so I call dispatch and I go, hey, I'm in a room and I'm surrounded by people and they all have guns and they're all staring at me. That's a very different thing depending on context. So when officers on the road spot a man with a gun, the recognition of the gun is important, yet it's not as important as the intent — whether that person poses a threat, what the situation actually means in context. It takes a human to determine that. You remember probably 25 years ago they had the new scanner where, as you were driving in the vehicle, it would scan people and show whether they had a concealed weapon. It could also show things in their pockets like wallets and change, but it also showed people as naked because it did away with the clothing. Everybody went absolutely batshit because that technology was rolled out so quickly, and then the accordion effect caught up to it and people went, yeah, but you're seeing that person naked. So we have to balance that, because the machine knows more than you think it knows. If the machine is programmed, can it still tell whether it's a male or female, or the race of the person behind the wheel? Well, if it can read the digits on a plate, it can certainly read things like that — and likely estimate age too. If it does those things, then it can be abused. I understand where people get afraid. We're afraid of things that are faster and smarter than we are, and we always think of the worst-case scenario. We don't sit around and go, oh, that'll be great for saving lives. We go, that son of a bitch is going to know who's driving that car, and that's ripe for abuse. But that abuse has been around forever. Coppers have always used that to stalk people. A flock camera now is no different than guys when I was on the road who followed people home, got their plate, ran all vehicles registered to try to get a name to contact later. That's bullshit, and that's a realistic worry.

We'll get into some of the civil liberties around it — it's a dual-use technology, you can use it for good or for bad. But I want to finish the thought on what this thing actually knows. At what point do we stop talking about what the machine observed and start talking about the explanatory story the human built around it? Because when I get a piece of information, your interpretation of it is going to be different than my interpretation, even if we have similar roles and responsibilities — based on experience, knowledge, training, all of those things. And that is what I mean when I talk about human-technology interaction. It's not just the user interface. It's what you do with the information, because information has inherent value if used correctly, and you don't even necessarily know which information is most valuable or which data points matter most.

That's a much better question. I was still going under the Kamala Harris definition of AI when you first asked that, so let me tighten my shot group. Machine-generated output creates involuntary heuristic manipulation, because heuristics are important to how we've been programmed — in our genetic wiring — to make sense of things. Take the availability heuristic: when external information is repeatedly presented, it's made readily available to the brain, and the brain begins to irrationally weigh that pre-packaged information as more important than the true observed facts right in front of you on the ground. So we start believing the machine, and that's a form of expectation bias. People staring at those screens for a long time begin filtering out conflicting evidence and disproportionately weighing the machine's output. And when you're searching for more information, you only search for what confirms the conclusion you've already been handed — that's confirmation bias as your ally, and that's never a good thing. Anytime we buy into unverified external inputs without checking tangible facts, artifacts, and evidence, mistaken assumptions will occur. They have to.

You brought up some good examples there. When you talk about availability heuristics and the way we see things — you go out and buy a new car, you're excited, you're driving around, and what do you see everywhere? The same car as yours. There were always that many Toyotas on the road; you're just noticing them now because you have one. And that's exactly what happens when you start interacting with this technology and it begins to tell you things — oh wait, I was right, I found those missing kids. Now this system has authority, and you no longer question it and no longer look at conflicting information that you otherwise would have before. That builds over time, and the longer you do that, the more reliant you become. Two or three generations later, we're all walking around with these cell phones and not a single person can tell you how it works — and we don't even use it as a phone anymore. So I think this was a good first part of the episode: talking about the issues with adopting technology and how we actually operationalize it.

Way too early for that much bourbon, that's all I'm saying. So one quick one on that. We don't adopt shit that we don't fully trust down to our core. When the automatic electronic defibrillators came out — the AEDs, the box on the wall that jump-starts the heart — I was a cop for a good long time, and Brian, you've met and worked with a lot of great cops that have double digits already in their careers. Ask them how many times on the road they did CPR and that CPR saved a life. I did CPR probably nine to fifteen times a year. Never had a save. Heimlich saved a bunch, never CPR. Then all of a sudden the AED came out, and guess what? Now you're getting saves. So even if you don't fully understand the technology, you're willing to embrace it because it saves a life. You saw that immediate return. You didn't have to be a cardiologist. Those types of things are a lesser lift, even though it's a high-tech tool — you buy into it because you see an immediate result. We still haven't seen — law enforcement has, but we still haven't seen in our own life — an immediate result. They're out there, but remember, they posed the argument before they gave us potential evidence to go through. They only gave us one side of the argument. It was binary: you're either for them or you're against them. That's not how we should judge this technology.

That's a good segue into some of the arguments for and against these technologies — specifically the mass surveillance systems, as people refer to them. I mean, your phone is a mass surveillance system, your Wi-Fi at your house is. So let's talk about some of the arguments against these systems, specifically Flock and the automatic license plate readers. One thing people immediately bring up is civil liberties — the most common criticism is that this is an invasion of privacy. You and I have talked about that before, and you were saying that's actually one of the weaker forms of the argument. Explain why you think that.

Folks, you can argue with Brian and I anytime you want — it's an open forum, all you have to do is write in. We'll put you on the next podcast if your argument's good enough. But don't send us crappy arguments, and the crappy argument is that this is against my civil liberties. Driving on public roads is highly regulated because it's a privilege, not a right. You have to have a driver's license, register your vehicle, maintain insurance, and obey traffic laws. Those things are in writing in every state, and there are federal guidelines. Driving privileges can be suspended or revoked. License plates and vehicles are openly visible on public roads — there's no assumption of any sort of privacy. Vehicles historically have always had much less protection than a person's own. We can go back to the earliest cases involving cars back when prohibition was running. Vehicles are inherently movable, so there's got to be a different standard. So don't come at me with that argument because it's flawed. Now, if you come at me with other arguments, I'll listen. Like you said, if I'm using this data to judge how many people are taking their kids to a certain restaurant or to a polling place, and that information is unregulated, then we might have a problem. But driving itself is not the issue. When you drive, you give up certain freedoms to use a public road.

If you're on private property and you want to do whatever you want in your vehicle, go for it. But that's just the thing — it's public versus private. When you're in public, you have no reasonable expectation of privacy in this.

I purposely run those stop signs in front of the Walmart in town and a young copper pulls me over — private property. I don't think those apply.

I think the difference with these systems now is it's not just a snapshot in time. It's collecting and aggregating information over time, storing it, and that can be searched and brought up again. That does open up some Fourth and Fourteenth Amendment issues about using that information. If I'm under investigation for something and you trace me back using these cameras to when I wasn't committing a crime — I was just out in public — I get that you could say you'd go watch a hundred hours of CCTV footage to see if someone walked past. But now you've built this historical record and you're going to use that information. It's the use, how it's stored — I think that does change what you initially said.

Folks, do your homework — look up the Yorkshire Ripper. They watched roads for this guy's vehicle for a good long time, managed to make a great record of everybody coming in and leaving the area, and that's how they got the tire impression, the vehicle type and make, and went to his home. They didn't only look at the Yorkshire Ripper's vehicle — they had to come from a larger number to a much smaller number, and that's what you're talking about. So anytime we go after data, we need a form of a warrant that allows us to look for those specific things, absent everybody else. Because if not, you're going to start finding other things you weren't looking for. That's why phone tap laws are what they are — those laws have been around forever. If it's on a phone tap and the persons aren't talking about the case you're after, you have to tune out. So that's what we're talking about here. When in doubt, whip it out on warrants, because they're easy to get. If you have an interest in a specific topic, write it down as an affidavit, go to a judge, and then you can go after those records. Even if the company says no, you can go after them. But what if it's warrantless? Because when you put two people in a room with high tech, they start messing with it, and sooner or later they're going to come up with a way to circumvent logic, reason, and sometimes law, and do something silly with it.

I get what you're saying, and I think that's another argument against — that police can misuse this. We've seen cases of stalking an ex or things like that. But that's always an issue with any technology or any authority. That's almost somewhat of a non-argument on its own, because it's really an argument for governance and how the technology is used and tracked. That's not new, right?

I mean, you even brought that up — regulating AI is not new. AI is new, but regulations aren't.

And misuse of authority for whatever purpose is not new. Like you said — I'm going to use this to look up something about the person who is now dating my ex-wife.

Yesterday there was a case I sent you where a cop's brother was going to be in a lawsuit, so the cop looked up the information and gave it to his brother so he could go confront the person. That's a misuse. That's not the technology's fault — that's the human in the loop's fault. We have to be aware of that anytime humans are paired with technology.

That argument is always there, and it's something that comes with governance and safeguarding. But because of how connected everyone is technologically and how surveilled everyone is — if you have a smartphone, you are — I mean, right now, even completely legally and completely open source, if I get three location data points for you, Greg, I can find you and your phone number just from ad data. There's open source stuff you can do now that isn't hard. So that's a little bit less of an argument to me. To me, the biggest and strongest arguments involve mistakes. What happens when the machine is wrong?

Look, if I write down the wrong plate number and testify and make the stop and do something, the defense attorney's gonna tear that up. But what is the court gonna say? The court's gonna say the cop made a mistake. Because a lot of times you're running and you misstate something about the person that you're chasing on foot, or the vehicle, or anything. Case law, time and time again, says mistakes are made. The tension is high. You thought it was this guy, and then in fact that's the person you apprehended. So the court goes really, really easy on humans that make mistakes in complex environments. Not so with technology. So if an automated system can misread a plate, or confuse a letter or a number, or misidentify the state, or misread an expiration sticker, all of those things are gonna go into a database and they're gonna be compared. And now they associate the wrong database, and you come back as a member of MS-13 — and it's not that the plate's associated with stolen, it's associated with you violating parole. So what's the worst part? That the machine made the mistake, or what's about to happen when those red and blues come on?

That's the biggest thing. Because of all the reasons we talked about at the beginning, now I'm going, oh crap, this is a stolen vehicle — that's what the alert says. And then now I'm going, okay, that's a good stop. And then that turns into a pursuit, that turns into a shootout, that turns into a car crash that kills a bunch of innocent people. Someone's gonna say those are two separate arguments, but if the system got it wrong and now that guy's running for some other reason — this person is freaked out, they're getting pulled over, and everyone pulling them over thinks this is a stolen vehicle. You're not looking for anything to disconfirm your belief right there. So now the guy's reaching for his license or his wallet and you're lighting it up as a gun.

So let's talk about that for just a second in the context of comparison. If you compare the rate of mistakes from expectation to explanatory storyline, the machine doesn't have that. The machine is going on information and gives it to you, and now you think there's a parole violator, it's a gang member, he had a gun before — why wouldn't he have a gun now? So you see how that loop is shortened. Now you're coming up with only negative things, a pseudo-negative feedback loop. That's in your mind. Ask every cop you know — you know thousands of them. Ask every cop you know if they've ever had a pursuit just because the person didn't have their license in their pocket. Did a person flee at high speed and blow intersections just because they were afraid to stop? That happens. So now the person hits the gas pedal and is fleeing from you. With your information loop, it only confirms that this is a dangerous felon, and now you're TVIing and shooting and doing all those other things. So the human expectation-to-explanatory-storyline-to-prediction is flawed, and that's going to create a very, very dangerous outcome at the end of it. Anytime you get a false hit on a name, where a name hits close and they run it on a computer and you go, is that this person? — now you're thinking the person in front of you has got all these felony warrants, and it's not. It's the softball coach for the local team. That's ripe with danger because you're at the precipice of a decision, and that decision and action is shortened because you're afraid. And you're saying, well, I'm not afraid, it's just my job. You're afraid. Your sphincter tightens because you're not sure what this person is going to do to protect their freedom. Anytime you get a warrant, that means the person didn't comply — so why are they going to comply with you? The machine making a tactical error can lead to a catastrophic error of sense-making on the ground, and I think that's exactly what you're describing.

It sort of hijacks this decision-making cycle that all humans do. I see something — hmm, that's interesting. Their vehicle's not registered, or there's a broken taillight. Okay, I can pull them over for that. I have reasonable suspicion. And then I'm talking to the person, this is confirming what I'm seeing. Now I have probable cause, I search the vehicle, I find the dope. That's the normal process of updating your hypothesis, creating an explanatory storyline, looking at most likely and most dangerous course of action. What these systems do — especially these ALPR systems — is go right to decision. It's a stolen vehicle, I need to do this. And this is insane to me: the LAPD went back and studied their data, and they found that 32% of stolen vehicle alerts examined in their review were false alerts. Not 3.2%. 32%. So that's just a recipe for disaster.

How much of a recipe? Let me ask you this from a military context, and I'll give you an answer in a police context because I saw it firsthand over and over. A high-stakes traffic stop — should have been a felony stop. The guy on the passenger side, the copper watching the passengers, draws his revolver. When he draws it, he cranks off a round between his own feet. Thank God he didn't hurt himself. That bullet skipped under the vehicle and hit the driver contact officer. Thank God it was in his vest and didn't kill him. What do you think happened next when there were shots at that traffic stop?

Oh no.

Answer that in a military context. Not that you witnessed it personally, but you've heard stories of negligent or accidental discharges. What does everybody else in that patrol do?

To the point that I've heard, hey, everyone fire a few rounds because I just launched one off on accident.

That's an extreme, but that's hilarious. That's briefly anecdoted.

I just launched a 40-millimeter grenade — everyone fire some rounds. Mine might need some help up here.

And so you got the squad leader firing tracers toward where your round went.

I've heard the comment, hey, where are you guys shooting at? We're suppressing. Suppressing what? The north. Just the general direction of north.

But the idea there is that's a low-tech thing that can happen when a low-tech accident occurs — a mistake, not intentional, no intent. So the machine doesn't establish intent, but we take it that intent is present because it gives us those big declarative words. And that's not fair.

Not only can it not tell us intent, but it's going to affect that decision-making loop — how you process information on the scene — so much that it gives you this false sense of, oh, I arrived at this conclusion, his intent must be that. And it's like, wait a minute, you just jumped three steps ahead in the game based on what you think is accurate information, but there was no comparison. You didn't say, if this is a stolen vehicle, what else should I likely see? What should I expect when I go up there? It hijacks that decision-making process. You're influencing the outcome more than you recognize. That's what these systems do — they influence the outcome of these situations and therefore the second and third order effects, yet they're not responsible for them.

Or the consequences of them right now. Consequence dictates that certain things get better. If you used a false search warrant, the fruit of the poisonous tree — none of the information gathered from that is going to be valid. It spanks you until you learn. The current system doesn't do that. Clearly it didn't do that in LAPD, for a third of the mistakes. There's a term for that, and I hate to bag our own book, but if you do a little studying on anything related to human behavior, entropy is such an important indication. Closed-system entropic failure is a tactical decision loop like Boyd's that relies on an outdated or externally distorted mental model, and then converts an open feedback loop into a closed system. That's why I'm always railing against the OODA loop. I keep hearing it every day on LinkedIn — this will make your cops better, this will make your business better. No, it's a failure. Not for Boyd, not for what he used it for — it's a failure for your misapplication, because you're not a fighter pilot and you weren't in the situation he was. It has to constantly update. A decision-making system that fails to continuously update its mental models against actual ground truth is going to degrade into disorder and make critical errors. Simple entropy. It unintentionally forces and reinforces reactive rather than proactive. So instead of operating left of bang, we're making packaged template matches instead of prototypical matches — and that puts us on the X, right at bang. The perceived crisis is already in progress when you go with an ALPR or a Flock hit. That doesn't give us flash to bang. That doesn't give you and your parents sitting at the dinner table talking about the news — remember those days? We don't have that now because the information cycle is so shortened that we don't put reasoning and decision-making at the end of sense-making and problem-solving. We make it first. Here's your decision, go enforce this. It's not right.

The greater the consequence of being wrong, the more important corroboration becomes. These are potentially life-ending consequences — there's life and death hanging in the balance. I don't think we fully take into account the second- and third-order effects, and how this changes in the moment, over time, generationally. The big argument with the Flock critics is that it's a surveillance problem. I get it from both sides — I understand the argument like, well, if you're not doing anything wrong you don't have to worry about it. I technically agree with that. However, we have a constitution and a Bill of Rights, and that's what makes this so hard. Our whole society is built on the concept that you are free to do whatever you want so long as you are not impeding on the rights of others. That is the fundamental thing. And information is extremely valuable — always has been. Information is power, and who has the knowledge and who knows what and when drives everything, the markets, everything we do. But I find your take on the surveillance criticism interesting, and I want you to share it. You've said your phone knows more about you — so what's your argument? It's almost like that surveillance problem is a non-argument to you.

I don't want anybody to think we're coming away from this choosing a side against ALPR, against Flock, against surveillance, against technology. What we're saying is everything has to be regulated to a point so that there are fewer mistakes and a more pure signal. The less noise, the better the signal — that's my argument. ALPR is uniquely invasive — I hear that, I read that every day. It's overstated. Most people are already carrying devices that continuously generate information about their location, searches, purchases, communications, travel, interests, and behavior. The existence of technological surveillance is not unique, and specifically not to Flock. And I would ask you now, as my dear friend — when I die, please clear my browsing history. That was not a joke, by the way. I was on a flight recently to the East Coast, and leaving Denver we had a total mechanical failure and had to make an emergency landing at an airport that wasn't ready for one. The emergency services people did an incredible job — we had to park far away, take the old-style ramp onto the tarmac, take buses in. They handled that part of the emergency very well. You know who didn't handle it? The passengers. What about my bag? What about my next flight? The idea is we've become so reliant on technology that we can't have it both ways. You can't want your GPS to know your travel reservations and then get indignant when GPS shows you a road and it's a roaring river. There has to be a balance. Is this really a surveillance problem? Gladys Kravitz on Bewitched was looking out her window all the time — that's a surveillance problem. We say if you see something, say something — that's a surveillance problem. So how is this different? My argument is that "uniquely invasive" is overstated — it's happening all the time. The real question is: what am I going to do with that information, and who gets to use it? If I was going to make an argument, that becomes my top argument when it comes to surveillance.

I see your point, and you're caveating that it's about who has access to it and what they're using it for — that's a bigger, broader question. The argument for this technology is, look, if your car gets stolen it's easier to find. So who's going to start pushing for that? Insurance companies, because they recover your car and don't have to pay you out. And then you're going to say, well, kids got kidnapped and we can find them — and of course people want that. I looked at the Amber Alert as almost this use of community with incredible technology. I've had it several times where I'm in the area of something that happened and I get that alert on my phone saying look for this vehicle, and I'm like, let's catch some kidnappers. You could do the same thing with stolen vehicles. How many Uber, Lyft, and Amazon delivery drivers are puffing around every city all the time? You could make it a game — a hundred bucks if you lead to the arrest of this vehicle. Like you said, we're already part of the surveillance state. Why do you think social media works so well? How many people do you know who never post anything but are always online? You're not interacting — you're literally just surveilling everyone. But this is where it comes back to what we brought up earlier: it's an oversight issue, it's a governance issue, it's who gets to use the information and why. That's what concerns me. Due to recent events, Flock had to come out and say they now delete information after a week instead of 30 days. And then you had investigators going, wait a minute, that historical data was really helping me build a timeline to show this person likely killed someone or robbed this place. So there's always a balance in any free society between civil liberties and safety — unless you want to live in a dictatorship, that's how it's always going to be. The important parts of the argument aren't just, oh, this thing does this, let's adopt it. It's how are we implementing this, how are we governing it, who has access, and what are the specific use cases? That's an argument against the business model of just wanting everyone to buy it and start using it. Look what happens when you don't do it correctly — now people are cutting the cameras down, going to town halls saying get this out of here, you spent our tax dollars on it. If you build in that governance and get community adoption where people say yeah, this is a good thing, it does get better over the long run. You don't have to ship it and fix it in version two. This is people's lives and civil liberties. You adopted it, you listened to what some guy from Silicon Valley said, you bought it, and now you've let the genie out.

Now what do we do? If you're listening and you own a corporation and you've got a couple of bucks, get yourself a Flock and an ALPR and put it outside your business and in your parking lot — that's going to save you a lot of damage and trauma. And if you work in the security industry, you should be thinking about having one of those remote speed-reading displays that roll around on tires, but with ALPR capability, at a venue — because those data points are going to help you exclude bad people from an upcoming event. Why do I make that argument? I would interrogate those of you who are listening on the medical industrial complex. Does it want to cure you? Clearly not, because if they cure you, you're not coming back and spending money — they want to give you enough drugs to keep you alive for a longer time. That's the problem with an industry regulating its own ideas. And I'll give you an example of that with the DNA-in-a-box thing — send it in and find out where you're from.

Greg, I think you're just supposed to give blood. I don't think you're supposed to poop in the box.

But the idea is that when the send-in DNA thing came out, it's a moneymaker and people fell for it. However, some savvy cop somewhere figured out that if they had access to that family DNA database, they could trace relatives and work back to a suspect. Every technology has the ability to be abused or used for good or evil. So what scares me? Can another jurisdiction or federal agency read my ALPR or Flock feed? Is a warrant required? Do they need a writ to look at it? You already said it — how long is information stored? Should it be 30 days, one day, 24 hours, or immediate like a TSA camera? Can borrowed recorded information cross state lines without a warrant? What searches are logged, and who routinely audits those logs? What happens when someone abuses it — is it internal discipline or a crime with consequences? Most people don't know this: if you violate a policy you might get fired, but that doesn't mean you violated a law, so you can keep offending elsewhere. All I'm saying is guardrails. I'm not saying be draconian about this. It's going to occur, and in 10 years we're not even going to have this argument anymore — it's going to be automated. Get on a European flight, Brian, as recently as either of us has done it. You don't have to show your ticket anymore; you walk up, it scans you, and you're on. We have to accept certain things. We have to negotiate parts of our freedom to allow for safety, and the U.S. Supreme Court has always backed that — they've always said this situation is different enough for safety and security that we'll allow you to go a little further.

On the private sector side, if I were consulting on security for some of these companies or big campuses, I would absolutely say allow it. It's private property, so you can do what you want — and yes, there's some informed consent from employees, but they either say they're okay with it or they go get a different job. You do have a choice. The use of these tools goes beyond that, and even connects to what you brought up with genealogy testing. That testing is now so good that if someone did their 23andMe, investigators can determine that person is the brother of whoever left that blood sample at the crime scene. It's absolutely insane. I was pissed at my brother because he did that a while back and I was like, why did you do it? He said he just wanted to find out stuff about his ancestry. I said, dude, now if I ever get jammed up for something, you're going to get me caught because you did that. And they actually have caught some serial killers that way — where someone's DNA hit on a crime scene and they traced it through a relative.

Many, many murders.

Where someone was a cousin of someone and their DNA hit on the crime scene, they investigated, and found out the cousin was the one who did it.

Hopefully they're doing their thing in LA, where they only have a one-third chance of catching you anyway.

With any of this, as these technologies increase, the governance has to increase with it. This is a general example of the conversation happening in the United States right now with all of these technologies and with AI. The technology always moves faster than any policy or government can, and that's just how things have always been. Back when they first started putting AM/FM radios in cars, a lot of people were against it, saying it was going to be distracting. Technically those people were correct, but a hundred years later it seems absolutely insane to think a radio was a serious problem. Humans do adapt, and it changes things. But look at cars now — the safety features, the cameras, the alerts, lane assist, self-driving. It's getting us collectively dumber, and we have to accept that reality.

What radio station does your self-driving car listen to? That's all I'm wondering — when you were saying that, I was thinking exactly that.

Just a bunch of machine noises or something, I don't know. There are some other incidents of this stuff that I think would make for a good Patreon segment, where we can go further for those folks. There's the one about the Waymo driverless vehicle where someone had a gun inside and they called the police — and again, you get into, okay, if that was someone about to commit a mass shooting we'd say this is the greatest thing ever, but then what about the Second Amendment, what about privacy and rights inside that vehicle? We'll definitely cover some of these specific cases on Patreon, and anyone who's curious can reach out and we'll cover it on there. This is again why we're constantly beating the drum about the human-first model. What does a human need, what's going to make their job easier, and then what are the second and third order effects? Any technology — what did it actually tell me, what am I building around this, what assumptions am I making, did I take time to corroborate the evidence? How proficient am I? The final question I'd ask you, Greg, and ask our listeners is: after all these different issues we brought up today — and each one could be its own podcast — what should we actually be worried about? Is it the camera, the system, the database, the algorithm, or is it us? Is it people?

Years ago LASIK surgery was something where you might lose your vision, and now robots do it outpatient and you're back on the road in no time. Technology can be immediately and immensely valuable to society, and we have to go there with AI — we're not going back, the genie's out of the bottle. But here's the thing: when someone was getting LASIK surgery, they got assaulted. When they went to the dentist to get a tooth pulled, that person took pictures of them. That's humans. The machine didn't conspire to do that — that was a human, and some humans are deeply flawed. Those are the ones we have to build filters for. We don't have to build filters against you. We have to build filters for the people who would bring devious intent to a technological advantage.

It's always a people issue. Guns don't kill people — people with guns kill people. Drugs don't kill people — people doing drugs kills people. It takes a human to do these things, so people are the problem, but that means people are also the solution. You're not obsolete; you're the main character in the story. Anything can be a dual-use technology, used for good or for evil. A hammer is just a hammer for building things until you start bashing someone's head with it — it goes from tool to weapon because of human interaction with it. If we focus on the human interaction with the technology within context, that's the most important part, because that's the piece we haven't cracked. We're always going to get better tech, and that means someone's always going to come up with a better way to defeat it. Just look at how much the war in Ukraine has changed over the last few years — rapid changes because humans adapt. You can predict likely technological use, positive or negative, based on past human behavior. If you focus on that element, it makes it easier to understand the potential problems and easier to correct them in stride.

We're not oversimplifying this — it's a complex, multifaceted issue. I think I would land on your answer: human problems require human solutions. I'm blanking right now, but it was either Philip K. Dick, Isaac Asimov, or Robert Heinlein who wrote I, Robot — I'm heavy on Asimov. He came up with the Three Laws. Look up when that book was written and how long ago he predicted exactly what we're talking about here: that when you have an artificial intellect, that intellect must always be there to help, never there to hurt.

A lot of the big thinkers, especially the scientists, came back to the idea that with anything — a new AI system, whatever — there are physical limitations because physics is at play, and you can only do so much. If you start to solve those physical limitations and include the human being in that accounting, you can get a better, faster system. Technology lets you see more, but do not let it give you permission to think less — that's what I would take away. Any other final thoughts, Greg?

I'm excited about Patreon. If you haven't checked it out, go to the website, pick up an episode or two, and you'll see why we're so excited about it.

I appreciate everyone for tuning in. You can check out the website and the podcast episode page to search for different topics — if you're ever wondering what episode we covered something, you can find it right there for free. There's a lot more on the Patreon. If you made it this far, thank you so much. Feel free to reach out to us anytime, and don't forget: training changes behavior.