The arrivals hall at Dulles, eleven at night, and the whole room funnels toward a row of glass booths. And the strange thing is that two people in the same family, walking the same line, can be preparing for two completely different events.
One of them is hoping to get through. The other one is going home.
That's the setup Daniel sent us. He was born in Ireland, he's married to an American, they fly over to see her family. And when they get to that booth, he locks down. Phone away, back straight, eye contact, courteous, short answers. Asks him where he's staying, he answers the question and stops talking. His wife, meanwhile, is picking up the little social nudges from the officer, chatting about the flight, mentioning the state she's visiting. And Daniel says he's sat there at least twice feeling uneasy watching her laugh during what he experiences as a vetting.
Two people, one marriage, two entirely different rooms.
And he's careful about it. He's not saying she's wrong or he's right. He's asking whether the two of them are archetypes. Whether the officer behind the glass sees these two modes all day long, and how many people actually banter versus how many go quiet and serious like him. And then the part he really wants, because it's the part I want: what does the fringe look like. What actually strikes an officer as atypical enough to matter.
He also said he doesn't think there's any data on it.
He's right about that. There isn't. Base rates of demeanour at a passport booth, nobody has counted. But there is very, very good data on the next question down, which is what officers do with demeanour once they see it.
The best study on this is from 1980 and it has never really been improved on.
Then start there. Because Daniel thinks the asymmetry is about personality. He's a serious man and she's an outgoing woman and that's the story. It isn't.
It's about what happens if the conversation goes badly. For a foreign national, refusing to answer questions at a port of entry isn't a mood, it's a decision with a price attached. You can be denied entry. Your visa stamp can be revoked. And you can be put into expedited removal, which carries a five year bar on coming back.
Five years. Daniel flies over to see his in-laws. That's the thing sitting under the polite conversation about where he's staying.
For his wife, the same refusal means a delay in a side room and then she goes home, because a US citizen has a right of entry. Harvard's general counsel office lays it out about as plainly as it can be laid out. Citizens and green card holders ultimately have the right to come in. Foreign nationals are subject to a review of their admissibility, and the officer has considerable discretion. A citizen who declines to answer additional questions beyond identity and citizenship may be inconvenienced, but cannot be refused entry on that basis. A foreign national can be.
So one of them is negotiating over a delay. The other is negotiating over whether the trip happens at all.
That's the whole case, and Daniel half knows it, which is why he's standing up straight with his phone in his pocket while she's telling a story about a rental car.
The other half is the part he's shy about, which is that her behaviour might actually be the riskier one and he might be the one doing it by the book. Because the archetypes are real and observable. The guarded mode, short answers, no volunteering, no eye contact past what's required. And the relaxed mode, banter, details offered up unprompted, reading the officer for a cue to drop the official register.
Both of those modes are documented. One of them is photographed, named and taught.
And the study that measures them found something strange, which is that the officers doing the reading were not good at it.
Daniel's question about what truly strikes an officer as atypical runs straight into that study, so let's go through it properly.
Syracuse, 1980, Hancock Airport. The researchers were Kraut and Poe, and they published it in the Journal of Personality and Social Psychology. The setup was this. A hundred and ten volunteer airline passengers, real people waiting for real flights. Half of them were randomly assigned to smuggle contraband through customs. Not told they'd do it, assigned to it, and then interrogated at the airport by an actual uniformed US Customs inspector.
A real inspector, real uniform, and half these people are carrying something.
The inspectors did not know the study existed in the sense that mattered. They were doing their jobs, talking to travellers, deciding who to send to secondary. The whole thing was videotaped. Then the tapes went to thirty nine actual Customs inspectors, people averaging almost nine years on the job, and forty nine laypeople who weren't in customs at all.
Both groups watched the same tapes and picked who they'd search.
And here's the headline. All of them, professionals and laypeople, were inaccurate. The correlation between actually carrying contraband and being selected for a search was negative point two two overall. Negative point one four for the professional Customs inspectors.
Negative. So the people who were actually smuggling were less likely to get pulled.
Slightly less likely, yes. The correlation runs the wrong way. The inspectors were somewhat worse than a coin flip at identifying who had the goods. And the researchers checked whether the smugglers had some tell they were giving off, some demographic or behavioural difference that made them legible. Quote, the mock smugglers did not differ from other travelers on any characteristic we measured.
So they're not giving anything away.
They're not giving anything away that a camera can catch. And yet.
And yet they agreed with each other.
Enormously. That's the second finding and it's the one I find unnerving. The judges did not agree with reality at all, and they agreed with each other almost perfectly. Twenty one variables, comportment and background together, accounted for seventy two percent of the variance in who they decided to search. Six variables on their own accounted for sixty percent. Nervousness. Length of stay. Trip purpose. Average answer length. Whether the traveller volunteered information. And age.
Six things doing nearly two thirds of the work.
Three variables did forty nine percent on their own. Nervousness, postural shifting, and answer length. That's the search decision, mostly. How long your answers are and whether you fidget while you give them.
So the officer's decision is not being driven by what he knows. It's being driven by how you carry yourself in a forty five second conversation.
And the specific behaviours that triggered the search were, in order of how much they mattered: appearing nervous, hesitating before you answer, giving short answers, avoiding eye contact, shifting your posture, and volunteering extra information the officer didn't ask for.
Hold on.
I know.
Volunteering extra information is on the list of things that get you searched.
It is. That's from the same study. Travellers who offered up details beyond the question were more likely to be sent to secondary, not less.
That's the wife's whole mode. She's volunteering which state she's visiting.
Which is the exact opposite of what Daniel assumes is happening. He assumes she's safer because she's relaxed and forthcoming. On paper, in the only study that measures this, she's doing one of the things that correlates with being searched.
And he's doing one of the others. Short answers is right there on the list. He deliberately gives short answers.
He does. Because he's being careful. And it's on the list of cues that get you searched.
Which staggers me a bit. The man is doing everything he thinks will get him through cleanly, and he's doing three of the six.
The study isn't telling you the short answer caused the search. It's telling you the inspectors weighted short answers heavily when deciding. Same thing, functionally, from the traveller's chair.
Read me the third part of it, because I remember there's a third part and it's the one that matters most for what Daniel actually asked.
The curvilinear question.
In English.
Whether being too calm was as suspicious as being too nervous. Because the inspectors told the researchers they watched for that. They said it in interviews. A traveller who's too relaxed, too composed, too easy about it, that's a flag for them too. They believe they're watching both tails of the distribution.
And they aren't.
They aren't. Kraut and Poe tested for it directly. The line in the paper is, no cue was curvilinearly related to judges' decisions to search travelers.
None.
None. Not nervousness, not calm, not answer length, not eye contact. The relationship was monotonic. More nervous, more likely to be searched. Calmer, less likely. There was no point at which calm tipped over into suspicious.
So the too calm thing is a belief officers hold about their own reading, and it isn't showing up in their behaviour.
That's the honest state of the evidence. One rigorous study, a strong belief among practitioners, and the two don't line up.
Which means Daniel's wife sits there laughing at a story about an overhead bin, and the officer's own self-reported model says that should be a concern, and the only data anyone has says it isn't.
Right up until you read the training manual.
Say the bit about the liars first, because it's the part that explains why the reading doesn't work.
Kraut and Poe's own explanation. Through a process of feedback and adaptation, liars have become able to control many of the external signs of stress, guilt, fear or concentration that are easily readable cues to deception, while not being able to control the covert signs as easily.
Meaning the signs the officers are trained to read are the signs anyone can suppress if they've had reason to suppress them.
And then the line that should be printed on a card above every behavioural detection program, which is that researchers seeking universal and easily observable cues to deception are engaged in a futile search.
Futile. That's their word, not yours.
Their word. Published. Seventeen years before the TSA existed.
And yet the training continued, because a system has to decide something at the booth.
It does. Let's go to what it decided.
There's a photograph problem here, which is that most of what I think I know about customs officers comes from television. The sunglasses, the slow walk, the sixth sense.
The training is more interesting than the television. CBP runs its Field Operations Academy at Brunswick, Georgia, and they use role players. Officers in training get put in front of actors playing travellers, and the actors get scripts. Some of them are told to smile and flirt with the trainee.
Flirt.
Flirt. And other role players are told to avoid eye contact, bite their nails, and appear to be nervous. Those are the two ends of the training script. The accommodating traveller and the visibly anxious one.
So the academy is deliberately training on both tails. The friendly-flirty one and the nervous one.
Both. Which puts the wife's mode into the curriculum as a case the officer is meant to be alert to.
Does anyone on the record say what the officer does with it?
The branch chief at the academy, Nick Sunderhaus. He was asked about border search authority, meaning what's left to the officer's own judgment beyond the databases. And he said, the things you're wearing, how your mannerisms are, all of those things can still come into play. How often that happens, there's no quantitative number.
No quantitative number. Because nobody's counting.
He was more specific than that. It's not kept track of, he said, because sometimes I might get that intuitive speak and nothing comes out of it. Only if there's an apprehension, a seizure, would that data be collected.
So the intuition that was wrong gets filed as nothing. The intuition that was right gets filed as a seizure. The ledger only has one column.
That's the structural problem exactly. You can't score a prediction when you only record the hits. Carol Rose of the ACLU of Massachusetts put it about as sharply as it can be put. The problem here is that officers bring their own subjective bias to the situation, and if they're not actually measuring their success rate, there's no way to actually judge or to know whether their training is working or not.
I'll take the point without taking the politics. If you never record the misses you can't tell a good reader from a lucky one. That's true of a customs academy and it's true of a fund manager.
Which is where the World Customs Organization comes in, because they are the ones who wrote the list you'd want. The Risk Management Compendium, volume two, the passenger risk indicators. They were released in 2011, developed the January after, and they're the closest thing to a published answer to Daniel's question about the fringe.
This is the list of what actually strikes a border officer as atypical.
Under the interview heading. Extremely nervous. Heavy perspiration. Overly confident or overly passive, that's one indicator with both ends on it. Hands fidgety or trembling. Carotid artery pulsing. A bobbing Adam's apple. Gulping for air. Avoiding eye contact. Restless. And, this is the one, overly friendly or cooperative.
Overly friendly or cooperative is on the same list as a pulsing carotid.
Same list. And the verbal ones. Reluctant or unable to answer routine questions. Evasive, vague or incoherent answers. Changing the story. Answering a question with a question. Talkative or unusually accommodating with authorities. Overemphasis on innocence, honesty or truthfulness. And, my favourite, because it's the most sophisticated item on the whole page, response conflicts with non verbal cues.
That last one is different in kind from the others.
It is. That's not a behaviour. That's a relationship between two behaviours. And the WCO is explicit that no single indicator means anything on its own. The line in the compendium is that only one indicator is not necessarily indicative of the presence of high risk. It's the combination that matters.
So the fringe isn't nervousness and it isn't friendliness. It's a mismatch.
The story that shifts halfway through. The calm face with the shaking hands. The answer that's a little too polished for the question. What reads as atypical is being off key relative to the situation, not being at any particular point on the nervous-calm scale.
Which reframes the couple one more time.
It does.
Because if the fringe is mismatch, then the question for Hannah isn't how friendly she's being. It's whether the friendliness matches everything else about her. And by the sound of it, it does. She is an outwardly friendly person who is at ease and behaving as she behaves.
And Daniel isn't off key either. He's a foreign national in a vetting situation being sober in a vetting situation.
Two people being exactly who they are, one behind a legal asymmetry that neither of them can fix by being nicer.
And the apparatus around that booth has grown since the study was done. CBP searched more than fifty five thousand electronic devices in fiscal year 2025. That's devices, not people. In January this year there was a new directive, 3340-049B, updating device search policy, and the searches require no warrant, no probable cause, no individualized suspicion.
Fifty five thousand phones and laptops.
And DHS pre screens something like two hundred and sixty three thousand passengers on international flights into, out of, or over the United States. And the EU's Entry/Exit System now cross references the paper stamp in your passport against a digital ledger, so the stamp stops being the record.
So the officer at the glass has more information than the inspector at Syracuse ever had, and is still reading your face on top of it.
The chief at Brunswick was candid about that too. The databases and the demeanour both come into play. He just couldn't tell you the hit rate on the demeanour half, because it isn't kept.
Which is where I'd land on Daniel's suspicion. He said he doubts there's any specific data on how many people banter and how many go quiet. He's right. Nobody has counted that, and I don't think anybody can. The one piece of rigorous work we have tells you what officers do with the behaviour once they see it, and it tells you they're worse at the reading than they think they are.
It also tells you something odd about their self image, which is that they believe they're alive to the too calm traveller and the data says they aren't.
And the WCO list, written thirty years later by the people who think hardest about this, includes overly friendly as an indicator, which suggests the belief didn't die, it got promoted into a manual.
The practitioner belief and the published indicator point one way. The one rigorous study points the other.
And nobody has run the experiment again at scale, which I find remarkable.
So does everyone who reads it. Thirty nine inspectors and forty nine laypeople is a small study. It's just the only one where anybody bothered to find out whether the officers were right.
Forty five years, Herman.
Forty five years, and the only honest answer is that we don't know.
Alright. Before we close this out, I want to come at something from the side.
Nothing to do with the border officers.
You keep saying the reading doesn't work on the traveller. Fine. I want to know if it works on me walking the other way.
Explain the direction.
The officers at arrivals are reading people coming in. What about the man on the desk in departures who decides whether what's in your bag is what you say is in your bag. There was one at the airport here in the early nineties, name was Farouk. Loaded the trucks for the caterer. My cousin Moshe did the crates and I did three nights on the belts when his back went, which the agency said wasn't their problem, so he asked me and I said yes because there's a version of money that isn't the hourly rate. Farouk would walk the line before the belt started. Didn't look at the bags. Looked at the men holding the manifests. And he'd pull one out of every forty or so and say, open your pockets. And he was right nearly every time.
Right about what?
Right that something wasn't declared. The agency kept a sheet. Moshe showed me once. Three months of it. Forty one times he pulls someone, thirty six of them had something. Small stuff. Cigarettes over the limit. A watch still in the box. The agency stopped keeping the sheet after a while because they said it made the men uncomfortable. Farouk didn't care one way or the other. Nobody told him about the sheet. I asked him once how he knew. He said he didn't look at their faces, he looked at their feet. And the ones who moved their feet to stand the way they thought an honest man stands.
That's a trained gait read. He was reading the performance, not the person.
And about twelve years later, over in the departures hall, there was a poster. It had a man on it, arms folded, and the writing said we know the honest ones by the way they walk. Nobody could remember who printed it. There were four of them in the hall and they stayed up for years.
Did the agency ever get the sheet back out?
The agency got taken over by a bigger agency in ninety four, and the new one didn't keep sheets. They said the sheet was unprofessional. Farouk retired the year after that and the line got slower for a while. That's all I've got on it.
Thirty six out of forty one is the number I'll take with me. That's a hit rate nobody in this episode has been able to quote for an actual border officer.
Because no one measures the ones they don't catch. Which is exactly the thing you just told us twice.
And the way they don't catch them.
What do you mean?
The men who moved their feet to look honest are the ones who got pulled. If you're standing the way you think an honest man stands, you've told the whole line you spend your evenings thinking about how an honest man stands. Farouk could see it. The man was doing a job on his own legs.
Which means the honest ones were going through untouched and nobody wrote their names down.
Nobody writes down the ones who go through. That's all of it.
So the man with nothing to declare who looks like a caterer loading crates walks through, and the man with a watch still in the box who has read one article about body language gets pulled.
Every night.
And that's the same shape as the study, only read forwards instead of backwards. The inspectors were reading comportment and finding the ones who were managing it. Kraut and Poe found the same thing. Volunteering extra information got you searched, because the officer read it as performance. Standing the way you think an honest man stands got you pulled, because Farouk read it the same way.
The men who were doing nothing were the ones he never saw.
That's the mismatch again, Hilbert. The performance was the tell, not the smuggling.
Which is why I bring it up. Because the officers at arrivals are looking for people doing a job on themselves. The woman in the story isn't doing one. That's all I'm saying.
So Daniel doesn't need to worry about the laughing. The laughing is the one thing nobody in this episode has a reason to flag.
Unless it's mismatched. Which is what the WCO was saying.
Unless it's mismatched. Which, by his own account, it isn't.
Which leaves the legal half standing exactly where it was. She can laugh because the stakes of the conversation are a delay. He can't, because the stakes of the conversation are the next five years.
That's the entire business.
Where does that leave Daniel's question? Because he asked it in good faith and I don't want to leave him with a shrug.
The base rates. He wanted to know how many people banter and how many go quiet. There is no count of that and I don't think there ever will be. Nobody is asking travellers at the booth what their demeanour strategy was, and if they did they'd get answers shaped by having been asked.
The closest thing to an empirical answer is the study, and the study measures what officers do with the behaviour, not who arrives at the booth in which mode.
The unresolved thing is the one that'll stay with me, which is that the officers told researchers they watch for too calm and too friendly, the WCO wrote overly friendly into the indicator list, and the one time anybody tested the curvilinear hypothesis they found nothing. No point where calm flipped over into suspicious.
Practitioner belief against the data, and the data hasn't been updated in four and a half decades.
The honest answer to Daniel is that his instinct is the regulatory one. The textbook traveller is the serious, self censored one, and he more or less is the textbook. And his wife's mode isn't wrong either, it's just that the friendly-volunteering pattern is written down in two places as something to watch for, and both places are run by people who have spent a long time thinking about behaviour at borders.
Which is an uncomfortable place to land. She can't adjust to it deliberately. If she tried to be less friendly, that's a job on herself. That's the mismatch.
He can't stop the serious presentation without lying to his own face. So you have two people doing the correct thing for their own situation, and the officer reading both of them through a model that the only rigorous study says doesn't work.
Fifty five thousand device searches in a year, two hundred and sixty three thousand passengers pre screened, a directive in January refining the device search policy, and the human reading of a face at the booth hasn't been measured since 1980.
If you take one thing from this, it's the gap. The behaviour officers say they weigh and the behaviour the data says they actually respond to are not the same list, and no one has bothered to close it. If the wife sits down on the couch beside Daniel when this drops, I'd nudge her to have a listen and tell him he's overthinking it. She can tell from the sofa what no academic has put a number on.
The asymmetry Daniel might not always feel is real. She's laughing her way toward a five minute delay if it goes sideways. He's standing straight toward a five year bar. Both are reading the room correctly, and only one of them can afford to read it wrong.
That's My Weird Prompts. Thanks to our producer, Hilbert Flumingtop, who is back at the desk and apparently has a sheet from 1993 that nobody else kept.
We're the human-AI collaboration podcast. Send us your own prompt on Telegram at t dot me slash MWP listener bot, or find the archive at my weird prompts dot com.
We'll be back soon.
See you then.
