Patients Don't Want More Information. They Want Better Interpretation

Anyone with a phone can now access, in an afternoon, more medical information than a GP of the 1990s had in their entire practice library. Every guideline, every drug datasheet, every research abstract, and lately a conversational AI happy to explain any of it at any hour.

By the logic that has driven thirty years of health technology, this should have produced the most reassured, most empowered generation of patients in history. Nobody who has spent time in a consulting room — or read the 2am search history of a worried person — believes that's what happened. People arrive more informed and less certain, carrying printouts and screenshots and contradictory threads, and the question they actually bring has barely changed in a century. It was never 'what is known about this condition?' It is: what does this mean for me? Those are different questions. The entire health information economy is built to answer the first. The second is the one that matters — and it's the one still in shortage.

The abundance paradox

Why did more information make people feel less oriented rather than more?

Because information without weighting is anxiety with references. The reader with a symptom and a search engine doesn't lack facts; they drown in undifferentiated ones. The benign explanation and the catastrophic one arrive on the same page, formatted identically, and the reader has no instrument for assigning probability between them — so attention, which follows threat, assigns it for them. The technical term for reading like this is hypervigilance; the colloquial experience is every parent who has searched a child's rash at midnight.

Information also arrives population-shaped. Almost everything written about a condition describes what happens on average, to people in general — and no reader is people in general. The statistics that genuinely apply to a specific person depend on their age, history, medications, baseline, and a dozen contingencies no published page can know. The reader is asked to perform, on themselves, the exact act of contextualisation that takes clinicians years to learn — with the highest possible personal stakes and no training. The surprise isn't that they struggle. The surprise is that we built an entire information ecosystem assuming they wouldn't.

What interpretation actually is

It's worth being precise about the thing that's scarce, because 'interpretation' can sound like mere translation — turning jargon into plain English. Translation is the trivial part, and the part technology has genuinely solved.

Interpretation, in the clinical sense, is several operations performed together. It is weighting: of everything that could be relevant, what matters here, and how much? It is contextualisation: this finding, in this person, with this history — amplified or discounted accordingly. It is sequencing: what needs deciding now versus watching over time. And it is responsibility: an interpretation worth the name is one somebody stands behind — 'given all of this, here is what I think, and here is what I'd do'. That last operation is what converts information into orientation: the moment when uncertainty stops being the patient's to carry alone.

Notice that every one of those operations is comparative and contextual rather than factual. That's why abundance didn't help — adding facts to a weighting problem is adding haystack.

Why technology keeps shipping the wrong thing

If interpretation is the scarcity, why does the industry keep building information products?

Partly because information scales and interpretation, historically, didn't. A page serves a million readers; weighting-for-this-person was artisanal work. Investors fund the scalable thing; the result is forty companies 'democratising access to health information' for every one that thinks seriously about meaning.

Partly because information is safe to publish and interpretation is accountable. 'Here is what studies show' commits the publisher to nothing. 'Here is what this means for you' is a judgment someone might have to defend — and so the disclaimer economy grew up precisely at the boundary where users most wanted someone to cross it.

And the LLM era complicates this honestly. For the first time, something like interpretation scales: a model will happily weight, contextualise, and advise, fluently, for anyone. Some of that is real value — better questions arrive in consulting rooms than arrived from search engines. But machine interpretation currently fails at the exact load-bearing points: it doesn't reliably know this patient's full context, it isn't calibrated about its own uncertainty, and it stands behind nothing. It produces the form of interpretation without the accountable substance — which, for a worried reader, can be more disorienting than raw information, because it feels finished. The scarcity hasn't been solved. It has been imitated, which is a different thing, and occasionally a more dangerous one.

What this reframes

Take interpretation seriously as the product, and several things shift.

Clinical communication stops being a soft skill. The clinician's irreducible contribution was never possession of facts — that advantage is gone, and good riddance. It is judgment plus accountability, delivered in a form a particular human can absorb: 'I've heard everything you've told me, I know your history, and I think this is what's going on — here's what we'll watch for.' That sentence performs work no document can. Training, and the systems that schedule clinicians' time, mostly still price it as a nicety. It's the product.

Health content gets a better target. The best patient-facing writing already behaves interpretively — it weights ('most cases are X; the ones that aren't usually announce themselves like Y'), it contextualises ('this matters more if you...'), it sequences ('nothing about this needs deciding tonight'). Content that merely informs is a commodity the internet finished producing years ago.

And the AI roadmap clarifies. The valuable system isn't the one that knows the most — knowledge is the solved layer. It's the one that could honestly support the weighting: surfacing what matters for this person, sized correctly, uncertainty declared, handing off to a human precisely where accountability has to take over. That's a harder product than a chatty encyclopaedia. It's also the only version that addresses what people were actually asking for all along.

What this means

The last era of health technology answered 'what is known?' so thoroughly that it revealed the question underneath — what does this mean for me? — in all its difficulty. That question can't be answered by adding facts, because it was never a fact-shaped question. It's a weighting under uncertainty, performed for one person, by something willing to be answerable for the result. Medicine has always known this, in the form of its oldest technology: a person who knows you, looking at all of it, and telling you what they honestly think. Everything now being built should be measured against how close it comes to that — and how clearly it knows when it can't.