Before You Call Someone a Liar

Part 8 of 8 in a series on lies and deception at work.

Every way of catching liars at work accuses some honest people. What the law, philosophy and research say about catching people out fairly.

In 1987 the Court of Appeal heard the case of Carl Mason, who had been convicted of arson. The police suspected him, but they had nothing that tied him to the fire. So they told him they had.

Officers said his fingerprint had been found on a fragment of a bottle that had held flammable liquid. When his solicitor asked one of the detectives to confirm it, the detective did. It was a deliberate lie. On the strength of it, the solicitor advised Mason to give his side of the story, and Mason admitted filling two bottles, one with petrol and one with paint thinners, and asking a friend to do the rest.

In court, the detective explained his thinking. If Mason had nothing to do with the bottle, there was no way he’d confess. If he had handled it, he’d start to doubt himself. It’s the logic of the photocopier lie detector in the third essay: the trick detects nothing, and the guilty give themselves away.

The trial judge admitted the confession and Mason was convicted. The Court of Appeal quashed the conviction. The police had deceived not only Mason but the solicitor whose job was to advise him, which the court called “a most reprehensible thing to do”. The confession should have been kept from the jury.

The court didn’t find that Mason was innocent. It found that the way the police had got his confession made the trial unfair.

That’s the subject of this last essay in the series. The first seven were about lies at work and the people who tell them. This one is about what you’re entitled to do about them.

Two ways to be wrong

Any way of deciding who’s lying can fail in two directions. It can call an honest person a liar, or it can believe a liar. It doesn’t matter whether the method is a hunch or a polygraph.

You can’t reduce one mistake without making more of the other. Lower the bar for suspicion and you catch more liars, but you accuse more honest people on the way. Raise it and you protect the honest, and more liars get through. One well-studied method of analysing statements, Criteria-Based Content Analysis from the second essay, shows the problem neatly. Set to catch 95% of lies, it wrongly flags about three-quarters of true statements. Set to raise almost no false alarms, it catches fewer than one lie in ten.

Numbers can tell you what a method will do. They can’t tell you which mistake is worse. That’s a moral choice, and English law made it a long time ago.

In the 1760s the jurist William Blackstone wrote in his Commentaries on the Laws of England that it is “better that ten guilty persons escape, than that one innocent suffer”. He wasn’t the first to put a number on it. Sir John Fortescue, writing around 1470, would rather twenty guilty men escaped execution than one innocent man was condemned. Benjamin Franklin, in a letter in 1785, went for a hundred. In 1997 the legal scholar Alexander Volokh collected dozens of versions, with numbers ranging from one to a thousand. One of his sources was a former state executioner in Mississippi, who settled on five.

The number has always been a matter of taste. The direction hasn’t. In a criminal court, punishing an innocent person counts as the worse mistake, which is why a jury has to be sure of guilt before it convicts.

Work doesn’t run on Blackstone’s ratio, and it isn’t meant to. An employer who dismisses someone for dishonesty doesn’t have to prove it to a criminal standard. The test comes from a 1978 case, British Home Stores v Burchell, which began with a pair of sunglasses. The employer has to show it genuinely believed the employee was guilty. It needs reasonable grounds for that belief, and it must have investigated as far as was reasonable. The tribunal’s phrase was “a reasonable suspicion amounting to a belief”.

There are fair reasons for a lower bar. An employer isn’t a court. It can’t make witnesses attend or put anyone under oath. But a lower bar will catch more honest people, so the way an employer reaches its belief matters more, not less.

The two mistakes don’t land on the same people either. When a liar gets away with it, the organisation usually pays, in money or in time spent checking, which the seventh essay tried to count. When an honest person is wrongly accused, they pay personally, and they’re usually the one with less power in the room. The Post Office spent years treating its computer system as right and its sub-postmasters as liars. I covered that in the sixth essay.

It happened to me once, in a small way. When someone told the person in charge that I hadn’t done a piece of work I had, the document itself and a colleague who’d been in the room settled it, as I described in the fourth essay. Without them it would have been one word against another.

Most lies aren’t worth catching

Most lies at work don’t need catching at all. As the first essay showed, the great majority of everyday lies are white lies. I’ve told my share, from admiring haircuts I thought were too short to telling a presenter their talk was good when it was full of holes.

At the far end of the argument sits Immanuel Kant. In 1797 he argued that you may not lie even to a murderer who comes to your door asking where your friend is hiding. He was answering the political writer Benjamin Constant, who had said that nobody has a right to a truth that would harm others. Kant’s position is admirably consistent and very hard to live by.

The philosopher Sissela Bok took a more practical line in her 1978 book Lying. Telling the truth needs no justification, she argued, but a lie does, and the burden falls on the liar. Her check on any excuse was what she called the test of publicity. Would your justification hold up if you had to explain it openly to reasonable people, ideally including the people you’d deceived? A kind word about a haircut passes easily. A doctored report doesn’t.

Bok wrote her test for liars. It works just as well for the people trying to catch them.

If every white lie counted, I’d have been hauled in over a haircut years ago. Even a kind lie isn’t free, as the seventh essay showed. But a lie that only spares someone’s feelings isn’t worth chasing, and I’d leave it alone. The lies worth chasing are the ones that land on someone else: the ones that move the blame or the credit onto another person, or end up in a report that someone will act on. At work, those usually come with a paper trail.

When the catcher lies

The detective in Mason’s case thought a lie was a fair price for a confession. In the United States, the law largely agrees. In 1969, in Frazier v Cupp, the Supreme Court upheld a murder conviction resting on a confession the police got after falsely telling the suspect that his cousin had already confessed. American police can still lie to suspects about the evidence, although in 2021 Illinois became the first state to ban the practice when questioning children.

England and Wales went the other way. After a run of miscarriages of justice, several built on confessions that turned out to be false, the Home Office set out new principles for police interviews in 1992. They became the basis of the PEACE model that officers are still trained in.

The first principle says the point of an interview is to get accurate, reliable information and find out the truth. Another says anything a person says should be tested against what the officer already knows or can reasonably find out. Officers must act fairly. They don’t have to accept the first answer, though, and questioning isn’t unfair just because it’s persistent.

That’s close to everything this series has found, set down as police policy more than thirty years ago. Check what people tell you against the evidence, and keep asking. None of it needs a trick.

Tricks aren’t only unfair. They produce false confessions. In 1996 the psychologists Saul Kassin and Katherine Kiechel asked students to type letters as they were read out, and warned them that pressing the Alt key would crash the computer and lose the data. A minute or so in, the computer crashed, exactly as it had been set up to. Nobody had pressed Alt, and everyone said so at first. In some sessions a woman working for the researchers said she’d seen them do it.

Across the study, 69% signed a confession. Where the task was fast and the “witness” backed the accusation, every one of them signed. Some came to believe they’d done it, and a few described how, recalling details of something that never happened. When researchers in the Netherlands repeated the study in 2003 and made confessing cost the students money, 82% still signed.

The detective in Mason’s case assumed an innocent person wouldn’t confess to something they hadn’t done. That’s the assumption this research tested, and it didn’t hold.

Organisations set traps of their own. A mystery shopper, like the ones who exposed the gap between a sales process and reality in the sixth essay, is a small deception: someone pretending to be a customer. So is the fake phishing email that IT departments send to see who clicks.

In April 2021 West Midlands Trains emailed staff to thank them for their hard work during the pandemic and promise them a one-off payment. Anyone who clicked the link to claim it was told it had been a phishing test, and that there was no bonus. The Transport Salaried Staffs’ Association called it “crass and reprehensible”. GoDaddy had done much the same to its own staff a few months earlier, with a fake $650 holiday bonus.

Bok’s test sorts these quickly. A mystery shopper can pass, as long as staff know the company uses them and the shopper is testing the service rather than anyone’s honesty. A phishing test announced in policy, with only the timing kept secret, can pass too. A fake bonus sent to people who’d worked through a pandemic fails the moment you imagine explaining it to them face to face.

It’s also worth asking whether the trap does any good. In 2025 researchers published the largest trial of phishing training yet, run over eight months with more than 19,500 staff at UC San Diego Health. Staff knew simulated phishing emails were sent, but not when. The training people got after clicking made only about a 2% difference to how likely they were to click on later ones. What mattered most was the bait. An email about updating an Outlook password fooled under 2% of people. One about a change to the holiday policy fooled nearly 31%. A test like that tells you more about the bait than about the people.

I’ve been on the other side of this. In IT, I ran phishing tests myself. Only the IT office and the company directors knew one was happening.

I set up an outside email address that looked almost exactly like a senior member of staff’s. The difference was subtle enough that you wouldn’t notice it at a glance. The email asked the recipient to contact someone else in the company on a popular messaging app, and to send them some sensitive information about someone on the company’s system. It gave a mobile number to message. We’d set up a phone with a new number just for the test.

Company policy banned that app, because its security fell short of the cyber security regulations the company was required to meet. It wasn’t on company phones, and staff couldn’t install it, because only IT could add apps to them. So anyone who replied would have to use their personal phone, which was also against the rules, for data protection reasons.

The person whose details the email asked for didn’t exist. We’d created them as a phantom on the system for the exercise, so no real data was ever at risk, only the information we’d set up for the test. Everything else was built to behave the way a criminal would when trying to get hold of sensitive information.

We sent it to staff at random, not long after they’d all completed the mandatory IT training we had to deliver. That training covered company policy and legal requirements such as data protection, and a few sections covered exactly what this test was looking for.

Staff hadn’t been told a test was coming, but we believed it was fair. It was run so that anyone who’d taken in the training should have spotted it. The company directors had signed it off. And during the training I delivered, I told people to pay attention, because there might be a test at the end. It sounded as though I meant a quiz at the end of the session, and no quiz ever came. Perhaps that was a little misdirection on my part. It wasn’t untrue. It just let people believe something else.

Quite a few people still fell for it. They messaged the number on the banned app from their personal phones, without checking it against the company phone book, which all of them could see, and handed over the phantom’s details. To fall for it, they had to break two rules on the way.

The people who fell for it were warned to be on the lookout for phishing scams, and given more training to make sure they understood. They blamed IT. That told us more than the test did. To certain members of staff, we were a room full of geeks flexing our muscles, telling them what they could and couldn’t do, even though the rules we were enforcing were a legal requirement. Either that, or they just didn’t care about IT and what IT actually did. It was the attitude I described in the sixth essay.

Looking back, I don’t think the warning was the right response, or a fair one. Perhaps it was something they simply didn’t understand from the training they’d received. A better solution might have been to review the training materials, and perhaps to have an actual test at the end of the training, to see what people really understood and what needed going over again. It would also have given us feedback on the training itself.

The test itself is a different matter. It wasn’t announced, so it doesn’t pass Bok’s test as easily as one written into policy. But I thought it was fair at the time, and I still do. We weren’t looking to catch people out. We were looking for vulnerabilities and training opportunities.

Wrong isn’t the same as lying

Before anyone is accused, it’s worth separating a lie from a mistake, because they can look identical. In the second essay I described a photograph I’d have told you was taken in Soho. The metadata said Shoreditch. I wasn’t lying. I was wrong.

Contradictions aren’t proof either. Even with the Strategic Use of Evidence technique, the most promising interviewing method in the second essay, innocent people sometimes contradict the evidence too, through a poor memory, for instance. An inconsistency is a reason to ask another question, not a verdict. And the fear of not being believed looks almost exactly like the fear of being caught, which is Ekman’s Othello error from the first essay.

I saw both halves of this in the criminal courts. As I described in the fourth essay, most of the repeat offenders with a history of lying came across as convincing, even when the evidence plainly said otherwise. The defence witnesses in the second essay were the other half: people sticking to a story the CCTV had already contradicted, without knowing it had collapsed. Some may simply have been wrong.

Looking honest didn’t mean someone was honest, and contradicting the evidence didn’t always mean they were lying. What settled things in court wasn’t who seemed truthful. It was whether what they said fitted the evidence. A workplace should hold itself to the same standard before anyone gets accused of anything.

What fair looks like

British employment law doesn’t only set the bar. Acas’s code of practice on disciplinary procedures, which employment tribunals must take into account, sets out how to reach it. Establish the facts without unreasonable delay. Tell the person what the concern is and let them put their side before anything is decided. Let them bring a companion to formal meetings, and let them appeal. In misconduct cases, where practicable, the person who investigates shouldn’t be the person who decides.

None of this is red tape. Each step gives the accused a chance to show you where you’ve gone wrong, while it can still make a difference. Most people spot lies at not much better than chance, as the first essay showed. A fair process is how you find out you were wrong in time.

The last rule guards against something this series keeps finding. People see what they expect to see, and nobody expects more firmly than the person who built the case.

Records and watching

I spent years in IT, which is about as close as most workplaces get to a room that can see everything. In the third essay I described being asked whether we could measure staff productivity, and up to a point we could. In the fifth essay I described colleagues asking me what I knew when someone’s access was removed or someone left suddenly. I kept quiet.

There’s a difference between a record and surveillance, and it matters. A record is something the work produces anyway: a sent email, a file history, an audit log. Nobody made it to catch anyone, which is why it’s so persuasive when it does. The logs that showed who had altered the audit reports in the second essay existed because the systems kept them, not because anyone was watching the senior team.

Surveillance is collecting information about people in case they turn out to be lying. British law treats the two differently. The Information Commissioner’s Office says monitoring has to be necessary and proportionate, and workers should be told what’s monitored and why. Covert monitoring is only for exceptional cases, such as suspected crime.

In February 2024 the ICO ordered Serco Leisure and the leisure trusts it worked with to stop using facial recognition and fingerprint scanning to clock more than 2,000 staff in and out at 38 leisure centres. The ICO said less intrusive options, such as ID cards or key fobs, would have done the job. Staff had been told the scans were needed to get paid, and the ICO said the imbalance of power made it unlikely they’d feel able to refuse.

The EU’s ban on AI that reads emotions at work, which I covered in the third essay, rests on the same reasoning. People being watched at work are rarely in a position to say no. And the seventh essay covered research suggesting that close monitoring can make people more likely to break rules, not less.

I was a systems administrator, and I had access to everything. I do mean everything. The job needed top-level permissions, because to set people up and support them properly, I had to have access myself before I could grant it to them. If I’d wanted to, I could have looked at HR files, payroll, who was on what salary and the company’s finances. I had access to people’s accounts and their email.

Having access didn’t mean I was going to go snooping through things that were none of my business or had nothing to do with my job.

There were rules for everyone else’s access too. If someone needed to get into a colleague’s account to cover them while they were on leave or off sick, the request had to come to IT by email from a department head. The director that department head reported to had to sign it off. So did HR, in case anything confidential had been discussed with them that could compromise the person whose account it was. Now and then people asked us for information they might have needed but didn’t have access to. We never disclosed anything without sign-off from the director and HR, and that sign-off had to come by email, so there was an audit trail.

The rules made records of their own. If anyone ever asked who had been given access to what, and on whose say-so, the answer was in writing.

The thing that isn’t a lie

In 1986 the philosopher Harry Frankfurt published a short essay called On Bullshit. Reissued as a small book in 2005, it spent 27 weeks on the New York Times bestseller list.

Frankfurt’s point was that a liar and a bullshitter are doing different things. A liar knows the truth and steers you away from it. A bullshitter doesn’t much care what’s true, only about the impression they’re making. That’s why, he argued, “bullshit is a greater enemy of the truth than lies are”.

Some of what went wrong in this series may sit closer to Frankfurt’s category than to lying. A figure rounded up in a meeting, or a department head telling directors that everyone’s happy, may come less from knowing the truth and hiding it than from not minding very much what it is.

No lie detector can catch that, because there’s no hidden truth to find. The question that does catch it is “How do you know?” It works on lies and bullshit alike. It asks about the claim rather than the person, so it accuses nobody. And you can ask it of people more senior than you, which is more than can be said for a polygraph.

What I believe

I’d rather believe people. Believing people by default is how anything gets done, and most of the time it’s the right call. When something doesn’t add up, check the claim, not the person, and give them the chance to show you where you’ve got it wrong.

After everything in this series, I don’t think the few prolific liars are the biggest problem at work. The bigger one is a workplace where people are afraid to tell the truth. A company that can hear the hard truths it doesn’t want, but desperately needs, will get far more of them. The most useful thing it can do isn’t to catch more liars. It’s to become somewhere the truth can be told.

Honest people pay for lies far more than liars do. They shouldn’t pay a second time for the way we try to catch them.

This is the last of eight essays on lies and deception at work. The series starts with Why We’re Bad at Spotting Lies.

Sources and further reading

Acas (2015). Code of Practice on Disciplinary and Grievance Procedures.

Blackstone, W. (1769). Commentaries on the Laws of England, Book 4.

Bok, S. (1978). Lying: Moral Choice in Public and Private Life. Pantheon.

British Home Stores Ltd v Burchell [1978] IRLR 379 (EAT).

Fortescue, J. (c. 1470). De Laudibus Legum Angliae.

Frankfurt, H. G. (2005). On Bullshit. Princeton University Press. First published in Raritan, 1986.

Franklin, B. (1785). Letter to Benjamin Vaughan, 14 March.

Frazier v. Cupp, 394 U.S. 731 (1969).

Ho, G. et al. (2025). Understanding the efficacy of phishing training in practice. IEEE Symposium on Security and Privacy.

Home Office (1992). Circular 22/1992: Principles of investigative interviewing.

Horselenberg, R., Merckelbach, H. & Josephs, S. (2003). Individual differences and false confessions: a conceptual replication of Kassin and Kiechel (1996). Psychology, Crime & Law, 9, 1–8.

Information Commissioner’s Office (2023). Employment practices and data protection: monitoring workers.

Information Commissioner’s Office (2024). ICO orders Serco Leisure to stop using facial recognition technology to monitor attendance of leisure centre employees. 23 February.

Kant, I. (1797). On a supposed right to lie from philanthropy.

Kassin, S. M. & Kiechel, K. L. (1996). The social psychology of false confessions: compliance, internalization, and confabulation. Psychological Science, 7, 125–128.

Oberlader, V. A. et al. (2016). Validity of content-based techniques to distinguish true and fabricated statements: a meta-analysis. Law and Human Behavior, 40, 440–457.

Personnel Today (2021). Rail staff falsely promised bonus in cyber security exercise. 11 May.

R v Mason [1988] 1 WLR 139 (CA).

Regulation (EU) 2024/1689 (AI Act), Article 5(1)(f).

Volokh, A. (1997). n Guilty Men. University of Pennsylvania Law Review, 146, 173–216.