The phrase "the reasonable person" has been sitting in my head since Daniel sent this one over.
Reasonable by whose lights, though. That's the whole question.
Right, and here's what he's getting at. Reasonableness is the load-bearing wall of the legal system. Negligence, standards of care, consumer protection, contractual obligations, what counts as acceptable business practice — courts reach for the hypothetical reasonable person constantly. But reasonableness isn't a measurable quality. What one person calls reasonable is shaped by upbringing, class, professional experience, culture. So a test presented as neutral may in practice be encoding the expectations of one particular slice of society. Daniel wants to know how deep that goes. When judges apply supposedly objective standards, how much subjective judgment is actually happening? And then the bigger question underneath it: is that subjectivity something you can constrain, or something legal systems should just admit to instead of hiding behind the language of objectivity.
And he's pointing at real scholarship, not just a shower thought. Mayo Moran's book on the reasonable person, Victoria Nourse's article on getting past the subjectivity-objectivity split. Both of them think the standard debate is misframed.
Then let's misframe it properly.
Moran calls the reasonable person the common law's most enduring fiction. That's her phrase from a keynote she gave at Lewis and Clark, and it's earned, because this thing turns up everywhere. Tort law, obviously, the standard of care in negligence. Criminal law — provocation, self-defence, duress. Contract law, the duty to mitigate your losses. And then the newer stuff, which is where it gets interesting. Administrative law, constitutional law, discrimination law. American sexual harassment law runs on it. Canadian constitutional equality law runs on it.
A fiction with a very good agent.
Here's what makes it clever. It's a hybrid. Courts reach for the reasonable person when the standard needs two things at once — some attentiveness to the individual qualities of the litigant, and the objective content of the legal norm. Moran says that unique blend of subjective and objective is the concept's foundation and the source of its utility. It's not a bug they never got around to fixing.
So it's a dial, not a switch.
And that's what produces Moran's central paradox, which she states about as sharply as it can be stated. The reasonable person is apparently vital to securing the law's commitment to interpersonal equality, at the very same time that he also appears to fatally undermine it.
So the thing that's supposed to make us equal is also the thing importing the bias.
The egalitarian critics say it all too often serves as a vehicle for importing discriminatory views into the heart of the legal standard. Same construct, both charges, and neither one is wrong.
Where does it come from?
Eighteen thirty-seven. Vaughan v. Menlove. Menlove owned a hay rick, the hay rick caught fire, the fire took out his neighbour's buildings. His defence was that he should only be judged by his own honest best judgment — bona fide, to the best of his judgment. Chief Justice Tindal wasn't having it. He said liability co-extensive with the judgment of each individual would be as variable as the length of the foot of each individual.
That's a good line for a man ruling on hay.
It's a great line. And the substance matters more than the line. Moran reads Vaughan as insisting that the law, not the individual in question, determines the values that demand respect. The reasonable person is central to negligence precisely because he does that in a way that ensures interpersonal equality.
Which is the honest defence of the whole construct.
It is. If everyone gets measured by their own private yardstick, you don't get a legal system, you get a negotiation.
So the fix creates the next problem.
The calibration. The standard is set by what's ordinary or normal, not by moral culpability. Moran quotes the tort treatise Harper, James and Gray — the reasonable person represents, and does not excel, the general average of the community.
Average. Not good. Average.
Average. And average of whom is where it comes apart. Courts will adjust the standard for children. They'll adjust it for physical disability. They'll adjust it for sudden incapacity — a driver who has a stroke at the wheel. Then a defendant walks in with an intellectual disability or a mental illness and the court says no, we don't adjust for that. Prosser's formulation is the one Moran uses: the law holds the mentally deranged or insane defendant accountable for his negligence as if the person were a normal, prudent person.
So a child gets the benefit of the doubt and a man with schizophrenia doesn't.
And the reasoning, when courts give it, is usually about drawing lines. Where would it stop, how would you measure it, who decides. Which is a real administrative problem, but notice what it does. It means the law is willing to bend its average toward some departures from normal and not others.
Children, physical. Not mental. That's a choice about whose ordinariness counts.
And Moran's case study on gender is McHale v Watson, Australia, decided in the mid sixties. Twelve-year-old boy throws a metal object, head height, and blinds a playmate.
Throws it at head height and the court doesn't see a problem.
The court exonerated him, invoking essentially that boys will be boys. Justice Menzies dissented. Moran argues that a gender-inclusive standard would likely have found negligence, the way Menzies did, because the baseline being applied was a boy's baseline dressed up as an average.
So the "ordinary person" was ordinary in a specific way.
And here's the twist that makes Moran's paradox bite. The reasonable person gets imported into equality-sensitive public law — sexual harassment, constitutional equality — precisely for egalitarian reasons. To correct the untroubled judicial point of view. And it's imported at the exact moment it's being discredited on equality grounds in private and criminal law.
Out the front door, back in through the side.
Moran calls it paradoxical and it is. The same tool, sold as the cure in one courthouse and indicted as the disease in another.
Let's put the anecdote down and ask whether this is measurable. Because "judges are people" is not news.
Nineteen thousand French child-custody rulings.
That's a number.
Ten thousand three hundred cases, eighteen thousand nine hundred and thirty-seven rulings. Train two models. One is judge-agnostic, trained on twenty to a hundred times more data. The other is trained on the individual judge's own past rulings. The judge-specific model hit F1 of ninety-two point eight five. The generalist got eighty-two point six three.
So knowing the judge beats knowing more law.
The author's own framing is that it's empirical support for legal realism — judicial identity plays a measurable role in outcomes. And a separate study, roughly six hundred and forty thousand circuit court decisions from nineteen eighty-five to twenty twenty, found panel political composition associated with outcomes across about ninety percent of all cases.
Ninety percent of the docket tracks who's sitting.
Associated with, I'd underline that. But the association is broad enough that you can't call it noise.
Here's my problem with all of it. Every one of these findings is compatible with two completely different stories. One is that judges are injecting their private values into the law. The other is that reasonableness has a range, the range is wide, and different judges land in different places inside a legitimate range. Same data, no way to tell them apart from outside.
That's the honest limit of the empirics, and I don't think you can resolve it from the numbers.
And it's not just individual judges. If the standard were crisp, two reasonable jurists would converge. They don't. So either the law is being bent or the standard is doing less work than its rhetoric.
Which is roughly where Nourse comes in, and she comes in swinging at the framing itself. She says scholars have made a category mistake in believing the reasonable man is a person.
What is he, then?
An institutional heuristic. Her line is that the reasonable man is the site for debate over majoritarian norms. He's not a character with a personality you're meant to imagine. He's the place where the argument about what the majority expects gets held.
That's a lot less romantic than the phrase sounds.
Her point is the subjective-objective debate obscures the commonsense necessity of having a hybrid standard. You were never going to have a pure version of either. And then she gives the equality argument, which is my favourite thing in the whole literature. Equality takes two, it is inherently comparative. It cannot be found in the minds or hearts of individuals, male or female.
Say the O.K. Corral one.
She imagines a standard that would have eliminated prior threats for battered women but allowed them for men at the O.K. Corral. She says that's not a matter of psychology or a particular woman's perception. That's basic inequality.
That one I'll give her. Because it's not about what the woman felt. It's about the rule being different for two people in the same position.
And her model is structured. She calls it release and restraint. Step one, the law releases majoritarian application norms — it lets the community's standards in. Step two, it restrains their excesses, asking whether what comes through yields oppression or bias. Two steps, in that order, and the second one is the check.
So the reasonable person becomes a two-stage mechanism, not a person.
Which is much closer to how it actually operates than the fiction is.
And Moran isn't willing to give up on the objective standard at all, is she? Because some feminists argued the only egalitarian response was to reject objectivity entirely and go to subjectivity where the disadvantaged are concerned.
She doesn't take that route, and Nourse explicitly disavows it as inadequate. Moran wants an egalitarian objective standard — keep the objective test, but use it to correct structural deficiencies in the judicial point of view. Correct bias with the tool rather than encode bias with it.
That's optimistic. I'd want to know who's doing the correcting.
So would I. But it's a serious answer, and it's better than pretending the standard is neutral as it stands.
Which takes us to whether this is an American and Commonwealth affliction or a common law one.
The textbook position is that the traditions differ. Civil law is abstract, systematizing, code-focused. Common law is concrete, casuistic, fact-based. The civil law judge is subordinated to the legislator, the common law judge makes law case by case. Legrand called it an irreducible epistemological chasm.
A chasm.
That's the word. Then somebody went and tested it. Judges in the Lab. Two hundred and ninety-nine real judges, seven jurisdictions — Argentina, Brazil, China, France, Germany, India, the United States. Each judge got a fully briefed appeal from the International Criminal Tribunal for the former Yugoslavia. Up to fifty-five minutes on the case. Real deliberation, not a survey question.
And the chasm?
No common law versus civil law difference. Country differences, yes — small but there, on the order of four percent of the variance explained. But the family difference came in below one percent and wasn't statistically significant.
There is no chasm.
The biggest divergences were between the two common law countries in the study. India and the United States, both common law, further apart than France and Germany.
That kills a very tidy story.
It gets worse for the tidy story. The same study tried to measure precedent, which is the thing common lawyers treat as the mechanism that keeps decisions consistent. The effect was barely detectable. Their own summary is that the affirmance rate under a reversed precedent came in only about six percentage points lower than under an affirmed one.
Six points.
Against a defendant sympathy effect of about nine. Their conclusion is that law is less determinative than previously thought, and they immediately add that legal nihilism is equally inapposite, which I appreciate, because it's the obvious misreading.
So precedent is doing less work than a sympathetic defendant.
That's their bound, yes.
I want to push on this, though. The judges decided anonymously. Real judges sign their names.
That's the limitation the authors themselves flag, and they call it the most pressing agenda for future research. An anonymous judge has nothing riding on the outcome. A signed judgment goes into a career, a reputation, an appeal.
Which could cut both ways. It could make them more careful, or more cowardly.
And they couldn't get English judges into the study, so the common law sample is India and the United States. They caution that common law countries at a similar level of development, or within the Commonwealth, might share more than their data shows.
Fair. But the India versus USA finding is still sitting there.
It's the most interesting result in the paper to me. Two common law systems, and the widest gap in the room.
Let me try to hold the whole thing in my head. The standard exists to stop each person being their own yardstick. To do that it has to define an ordinary person. Defining ordinary smuggles in a class, a gender, a set of manners. Then the empirical work says judges vary a lot anyway. Then the other empirical work says the traditions you'd blame for that variation don't actually differ — it's the individual and the country that drive it.
And both challenges, Nourse's from the theory side and the lab studies from the data side, arrive at the same place. The objective standard is doing less work than its rhetoric claims.
So Daniel's third option looks strongest to me. Not constrain it, not eliminate it. Acknowledge it.
The AppealCase benchmark is built on exactly that premise, by the way. A dataset for civil appeal scenarios that explicitly states its goal as improving consistency in judicial decision-making. If subjectivity is a real variable, you can at least measure whether a tool narrows it.
And does it?
No. Every current model scores under fifty percent F1 on predicting whether a case gets reversed.
The machines are less consistent than the humans they're auditing.
On that narrow task, yes. Which is worth saying out loud before anybody proposes a robot bench.
I'd take the weaker version of the claim. Not that AI would judge better. That measuring and publishing these patterns might shame the system into being more explicit about what it's already doing.
That's the appeal, and the counter-argument is that you can measure a bias without being able to remove it, and the measurement becomes a fig leaf.
Then at least the fig leaf is honest.
A Hundred and Nineteen Dollars.
Nineteen eighty-one. I was doing deliveries, and there was a neighbour, Mrs. Kovacs, whose dog got into my van and ate the seat. Not a bite out of it. Ate it. Upholstery, foam, the lot.
Which seat?
Passenger side, front. The point is she wouldn't pay, so I took it to small claims. Costs me a hundred and nineteen dollars to file, on a van I'd paid four hundred for. The magistrate asks me do I have a witness. I said the dog's the witness. He didn't laugh.
Did you win?
He said I should have known the van was parked where a dog could get at it. Contributory. I got forty back.
Which lines up with everything we've been saying. The judge decided what a reasonable man would have done, and the reasonable man owns a better van than yours.
He wasn't wrong about the van.
Hilbert, hold on.
Give me a second. Sorry. Level was riding hot on this mic for the last ten minutes, I've pulled it down. Where was I. Mrs. Kovacs. She had a claim in the other direction, by the way. Said the dog hurt its teeth on the springs and I owed for the vet. Seven dollars and fifty.
Did she get it?
The magistrate said the dog's enjoyment of the seat was not my responsibility. So no.
That is the most judicial sentence I've ever heard.
I still see her at the market. We don't speak.
I've got one more thing, and it's the smallest thing on the list, which is why I like it. The Judges in the Lab authors admit their own study can't tell you the thing everybody actually wants to know, because their judges ruled anonymously. Real judges sign their names, and a signature carries consequences. They call that the most pressing question left open, and I think they're right, because every consistency mechanism anyone has proposed runs through reputation in some form. The French custody model works because a judge has a record. The circuit court pattern works because a panel has a known composition. The reasonable person works, when it works, because we imagine a person and then argue about which person.
Hilbert's magistrate wouldn't have said anything different with his name on the ruling. He'd just have said it more carefully.
Which is the entire business. A hundred and nineteen dollars to file, forty back, and the law decided my brother's van was unreasonable.
That's the show. Thanks as ever to Hilbert Flumingtop, who did in fact fix the level while we were talking. If you want more of this, try episode thirty-one fifty-one, When Courts Need a Conscience; episode nine, Benchmarking Custom ASR Tools - Beyond The WER; and episode twenty-one, Is Your AI Secretly American. This has been My Weird Prompts.
If you've got a question of your own, send it to us on Telegram at t dot me slash MWP listener bot. We'll be back soon.
See you then.
