The New Imitation Game Part 2: The Judgement Problem

Most people who work with writers, ghosts, story consultants or, as I heard recently, "narrative architects," come to the party thinking they have a writing problem.

That's not quite the case. Most have a judgement problem.

As I argued in Part 1, the writing is almost the easy part. Tough to say given that's the bit I invoice for, or at least the bit people think I invoice for.

What usually happens is this.

Someone who has lived an interesting life drops me a line, having decided that it probably ought to be written down before it's too late. They have lots of notes, old photos, and in fairness, a good story - usually. There is, as they often remind me, a book in all of us, and they want theirs written down properly by someone more adept than themselves at the writing game.

It's a reasonable brief but it's only about half the problem. The other half is best summed up by something Marc Andreessen said recently.

"When one thing becomes abundant and cheap, another thing becomes scarce and valuable".

Competent generative AI narrative is now the abundant thing. Give a large language model enough material and it will produce something structured, clear and depending on your pov, usually non embarrassing. No accident there as they are built on probability and predict likely sequences, which makes them pretty effective at that exact job.

There is, however, a more fundamental problem.

Good memoir runs on what doesn't fit the pattern, the awkward detail, the contradiction that never quite resolves. It's called tension, and human lives are defined by it.

Which means the thing that has become scarce, the thing that now actually matters, is not better prose, though there’s a strong argument here about AI vanilla mediocrity leading all of us to 'boresville' USA, it’s judgement, which is even harder to express as a salable item on an invoice.

By judgement, I don’t mean judgemental, I mean deciding what to include in the story and what to leave out. Also it’s about how to express sometimes difficult and sensitive subject matter without prejudice or harm. 

So, what arrives as I just need someone to help me get this down is usually something closer to: I need someone to tell me what this story actually is, which parts of it make it worth reading and mean something, and how do I express it.

Because memoir is not just a writing problem, it’s a disclosure problem.

People are being asked to talk about things that are not always comfortable. Decisions that affected others or moments they got things badly wrong and screwed up. Relationships that broke down in ways that still don't reflect especially well on anyone involved. That requires a specific kind of environment, a trusting one.

There’s a fundamental difference between entering information into a system and telling someone something. The latter involves discretion, nuance and contextual understanding.

It allows for things to be said, reconsidered, then re-expressed slightly differently to get to the core of any particular point and it allows for silence, which is often more telling than what’s said. 

AI can simulate conversation and maintain a thread, but it’s not the same thing as sitting with someone who understands the terrain (life), who knows when to push, when to leave well alone, and who carries actual responsibility for how those words are eventually handled on the page.

Another human is a far more skilled collaborator than the machine, and far more likely to provide a duty of care to the subject. 

Not least because memory, is not a filing system. It's more like a creative director, whose main brief is to smooth over the parts that don't fit the hero's journey and pass over anything that complicates the comfortable narrative. 

We are all, to some degree, the unreliable narrators of our own lives and AI will not find the gaps in that myth. It will take our prompted version, believe every word of it, and help us faithfully reproduce whatever version of events it is given, whether complete or not.

The writer's job is to notice these tells. To find the places where the account is working too hard. Where reach overextends grasp and where that Soho House anecdote now sounds a bit tin-eared - all with a human touch. 

This is also why memoir is unusually resistant to automation. A memoir is not the arrangement of facts but the recomposition of memory into meaning, and is selectively shaped by joy, regret, shame, pride, hindsight, grief and all other very human emotions.

People don’t simply retrieve it, they discover things in its discovery and telling. They hear themselves say something and realise, sometimes for the first time, that it was more important than they had always realised.

In that sense, memoir is not the reporting of a settled truth. It’s often part of the process by which truth is found. AI can process material about a life but cannot remember one. More importantly it can’t carry responsibility for how a memory is framed, tested and placed on the page.

Readers feel this difference, even when they can't name it.

I should say here, none of this is about doing a hatchet job on a client's carefully curated history. Stones and glasshouses and all that. Plus many, I'd say most of us, have far more to leave behind on the credit side of the balance sheet than in the debit column.

What I am saying is that many people actually benefit from a more rigorous process, one where the truth is more important than the simplified versions of ourselves that time naturally creates.

You might get away with that if your goal is framed around PR, but if you're writing for a book-buying audience, that isn't optional. Bland doesn't sell, readers can spot inauthenticity a mile off. 

Many people want to talk not because they can't use AI, but because being properly listened to is part of the process. It helps them think, remember and, in some cases, understand their own story clearly for the first time. That dynamic is not a warm-up to the work. It is the work.

So when people ask me whether AI will eventually take over my job, my answer is probably yes, for a certain kind of ghostwriting. The kind where the material is ‘clean’, straightforward and the output just needs to be competent. For that, the machines will have the upper hand in the not too distant future (probably).

But the work I find genuinely interesting, figuring out what a life actually meant and how to express it with pace, authenticity, humanity and readability, has not become easier or cheaper.

It has become more valuable.

Because the writing is no longer the scarce thing. The truth is. And that’s where great stories hide.