A friend starts telling you about a messy fight they had, and by the third sentence, before they've even finished the story, you've picked a side. Not intentionally, sometimes not even consciously, but somehow the umpires whistle is already in your hand.

“He started it.”
“She always does this.”
“I would never have...”

That's the idea this issue is about, and it isn't just a human one anymore. More of us are handing that exact job of ‘judging’ to AI: pasting in our side of a fight and asking it who's right.

And the thing is it always gives biased answers, whether it truly knows or not.

For example a guy I work with, let’s call him Sean, told me about a strange fight he had last week, and as I listened to his tale at an afterwork bar in the city, I had blown the whistle on it before he finished talking.

The fight itself was a few months old.

He'd been out around the traps of Melbourne with someone new. A great night in the city; bars, neon, glitter and laughs. Then back to his flat, more wine. The other guy stayed over and slept on the couch.

The couch was a new Togo sofa. A ground-hugging, crimped foam creation and one of the few things Sean had bought entirely for himself since moving down from interstate…

…into a high-rise flat that was still mostly borrowed furniture and boxes.

You can see where this is going.

His lover woke up sick on the couch and tried to clean it up himself.

It stank anyway.
Sean asked him to get it professionally cleaned.
The guy said no.
Five hundred dollars.

And that was the fight.

Somewhere around "the couch was brand spanking new," I'd already awarded the free kick and stopped listening, the way a crowd decides a decision before the referee has even reached into his pocket.

I’d never met the other guy, never heard one syllable from him, and had no idea whether he'd been genuinely unwell, mortified, or watching five hundred dollars swallow a fortnight's wage in one go.

None of that slowed me down at all.

I did all of that without any AI in the room and two months ago, I found out it does exactly the same thing, just faster.

Two months ago I wrote about a fight with my mum, and about typing my side of it into a chat window hoping for a clever retort. What I landed on was that AI only ever sees the version of the story you’re willing to type.

I called what I'd given it my case for the prosecution.

That was true, though it turned out to be only half the job.

Imagine Sean and his lover had both sat down and typed honest, un-spun accounts into the same AI chat.

Both sides in full.

What would’ve come back would’ve been two distinct stories and still no answer.

The machine would weigh them, soften them, and say something warm about how they are both hurting, but hearing both sides isn't a standard to judge by, it's just twice as much story.

An umpire who has heard both teams shout isn't any closer to a fair decision, he just has more noise booming in his ears. What's missing is a fixed rule, one that doesn't come from either side's story at all.

As it happens, two researchers (Gottman and Levenson) already built four of these rules, decades before anyone thought to paste a fight into a chatbot: not by listening to who was right, but by coding what people did while they argued, then checking years later who was still together.

Hold onto that last word of the quote above.

In 1983, John Gottman and Robert Levenson recruited 73 married couples in Bloomington, Indiana, filmed them arguing about something real, and coded what they saw. Four years later they went back to find out what had happened to them.

They weren't listening to the content of couples fights: no view on the in-laws, money issues, or not washing up or putting the garbage out.

They were watching out for four specific criteria, or markers, they called the four horseman that together cause the apocalyptic collapse of a relationship. Much like in the Bible!

The first marker or horesman was criticism, which is a way to hit the other person in the gut instead of focusing on the problem. “You always.” “You never.” “What’s wrong with you.”

The next one was contempt, which is the eye-roll, the mockery, the sense of standing slightly above the person you’re talking to.

Then it was defensiveness, where every complaint is met with a counter-complaint and nobody ever takes responsibility for anything.

The last one was stonewalling; the shutdown, the silence, the person who has left the room without moving.

The couples whose arguments ran hot on those four were significantly more likely to be in trouble later, including having seriously considered ending the marriage by the time researchers checked back in 1987.

And it isn't only the four criteria, either.

The same researchers put a number on the balance too: stable couples held roughly five positive interactions for every one negative one during a conflict. Outside conflict, on an ordinary day, that climbed to about twenty to one.

The table below shows those ratios.

But here’s the thing, research that is this well known has a way of getting oversold, over time. The more famous a finding gets, the more its accuracy claims need checking before you trust them completely.

For example, you may have heard that Gottman can predict divorce with ninety-something per cent accuracy.

Be careful with that one.

In 2001 Richard Heyman and Amy Smith Slep built a divorce-prediction equation, hit ninety per cent on the data it was built from, then tested it on a fresh sample and watched the accuracy fall hard.

Their conclusion was not to trust a number that hasn't been checked against fresh data, however impressive it looks.

That's the reason "marker" was the word worth remembering earlier.

Gottman and Levenson were explicit that their data couldn't show what caused a marriage to end, only what correlated with it, a marker, not a cause.

A cause is something you argue about, but a marker is something you can just point at and check off, which is exactly why it's usable as a rule and a verdict never is.

So remember that the four horseman, or markers, and the ratio around them, are not a crystal ball. Think of them more as a rule book, and a rule book with tested data is exactly what nobody ever plugs into AI when trying to solve their own fights with others.

Stop asking AI who was right. That question doesn't really have an answer, because AI has nothing to check your story against except your own telling of it; which will always be biased to yours truly.

Instead, give it the rule book first: the four markers you just read about, plus the ratio around them.

Only after that, tell it what really happened, and ask what the replay shows, the same way a referee checks the tape against the actual rules, not just an impression of who looked more upset.

That's the switch: rules before the story, every single time. Get the order right, and AI has something real to measure you and your fight against. Get it backwards, and all it's doing is agreeing with whoever spoke first.

Here's the switch itself, spelled out.

You’re not plugging your story into AI so it can pick a winner, you’re getting AI to check your story against a fixed set of rules, the four markers and the ratio you just read about, rules that came from neither of you.

Don’t worry, I’ll show you how in a minute.

The order matters because whatever AI hears first becomes the frame for everything that comes after.

Tell your story first, and by the time you ask "who was right," it's already built its synthetic sympathy towards you around your version, because there’s nothing else in the chat to measure it against.

Give it the rules first and that can't happen. The standard is fixed before you type in your side of the story.

That's exactly what the prompt below is built to enforce.

It won't let you describe what happened until after you've pasted in the four markers and their definitions, not because it’s a clever trick, but because it's the same discipline as a referee reading the rulebook before kickoff, not after a coach starts shouting from the sidelines.

Once it has that standard, what comes back reads like a match review, rather than a concrete verdict.

For example, say you told AI you'd opened the fight you’d had with someone with, "You always leave this for me to sort out." It would flag that straight away: that's criticism, not a complaint, and "you always" twice in one conversation.

It can then give you back the same message but with the markers taken out, something you can say out loud; but without the critical tone and inflections.

Same four moves, different strip.

Last issue was about finding the smallest dose that makes the biggest difference, this week the dose is smaller still. Four things to look for, not forty.

Here's what the difference looks like in practice, story first versus rules first, side by side.

Before you use this, one boundary.

This is for ordinary fights between two consenting adults who are both in a safe space. It’s not for situations involving fear, threats, coercion, control, physical harm, or a relationship you’re frightened of.

If that’s where you are, a rule book is the wrong tool completely. A real person like a GP, counsellor, trusted professional or appropriate support service is the better place to take it.

For everything else, here’s where the umpire finally puts the whistle down.

The prompt below is not designed to tell you who won the fight. It can’t know that from one person’s recollection, and the four markers weren’t built to answer that question anyway.

Instead, it does something smaller and far more useful.

It fixes the standard before you tell the story, then asks AI to look at the part of the replay you can examine honestly: the words and behaviours you brought into the conversation.

Not:
“Was I right?”

But:
“What did I do when I was angry, hurt or defensive — and is there a cleaner way to say the same thing?”

That shift matters because you can’t control the other person’s half of the argument; all you can do is inspect your own.

Rule book first.
Replay second.

Your side first.
Repair, not verdict.

Here’s the copy-and-paste prompt.

[ Start of Prompt ]

[ End of Prompt ]

The useful thing you leave with is not a score.

It’s something much smaller: one moment you can see more clearly, one move you can own, and perhaps one sentence you can say differently next time.

That’s enough.

Because the point of reviewing a fight is not to build a better case against the other person, it’s to catch your own ‘judgy’ hand reaching for the whistle.

And, just for a moment, put it down.

Sean and his lover broke up not long after, badly, and Sean bought a new designer couch, which is either the saddest detail in this whole story or the only part of the narrative he ever truly controlled.

Think about it and sit with the scoreline for a second.

The argument was about five hundred dollars. It cost him a couch and the guy.

I’m not going to tell you a rule book would have saved the relationship. It might not have. Some things are only meant to last a few weeks, and a stain is just what turns up at the right moment to end them.

But neither of them ever found out which of the four horseman they were embodying knowledge that might help them in future relationships, because they had nothing to check against.

Two certainties in one small flat and a smell that would not shift.

But here's the part of the story that's mine.
And likely yours too…

Nobody made me, or you, the umpire. Sean was just telling me a story. I picked up the whistle anyway, because picking it up is free, easy and it feels a lot like listening.

It isn’t!

Maybe your version isn't a couch. Maybe it's your sister-in-law's account of her ‘horrendous’ husband, the redundancy as told by the person at the desk next to yours, or the fight between your two kids where you already know who started it.

The whistle goes off before you've seen the replay.

That's not a character flaw, it's just what your brain does when handed one side along with no rules.

A monk learns to notice these moments when the whistle comes up, without taking on board the judgement, just seeing it arrive, and asking what it was measured against, which is usually nothing at all.

So this week, when a fight is still hot and heavy within you, do it in the wrong order on purpose.

Rules first.
Story second.

Hit reply and tell me what it was about and I’ll take the other side, badly, and we’ll both learn something.

So this week, borrow the rule book before you borrow the whistle. Nothing dramatic or a grand apology, just one line, said the way you would want it said to you. And if you know someone currently carrying an argument around like a stone in their shoe, forward this to them. Gently.

Ps. I think out loud between issues over on @cedricchenefront or @wellwireddaily, where we talk everything AI + wellbeing and self growth.

Talk soon,
Cedric the AI Monk; stay well, stay wired! 🧠⚡

I write a private, twice-monthly brief specifically for professionals like you. Inside you’ll find ways to harness AI to eliminate, delegate or automate your admin (notes, intake, referrals, etc), grow your wellness business and elevate your wellbeing, one working prompt or workflow per issue.

Your AI stack should fit your life, not someone else's workflow.

30 minutes. Three specific AI actions mapped to your actual situation. Voice note delivered same day. CDI Award-winning AI design + Zen monk + 4 years in health AI. Full refund if it doesn't deliver.

I want an AI Clarity Micro Session →

_{Disclaimer: None of the content in this newsletter is medical or mental health advice. The content of this newsletter is strictly for information purposes only. The information and eLearning courses provided by Well Wired are not designed as a treatment for individuals experiencing a medical or mental health condition. Nothing in this newsletter should be viewed as a substitute for professional advice (including, without limitation, medical or mental health advice). Well Wired has to the best of its knowledge and belief provided information that it considers accurate, but makes no representation and takes no responsibility as to the accuracy or completeness of any information in this newsletter. Well Wired disclaims to the maximum extent permissible by law any liability for any loss or damage however caused, arising as a result of any user relying on the information in this newsletter.}