Strive for Just Enough Information

On the GOV.UK page headed “Report a pothole”, there is a box asking for the postcode of the road. Beneath it sits a button: “Find your local council”. It's pretty intuitive. The page has not asked me to understand how responsibility for road maintenance is divided before allowing me to begin. There are separate instructions for major roads, and for Scotland and Northern Ireland, so the complications have not been wished away. They simply have somewhere to go. For the moment, I can concentrate on identifying the road rather than identifying the correct branch of government. It is not an exceptional piece of writing, which is partially why I like it.

Too often, when I look at official information, I find myself mentally moving the useful paragraph to the top. The explanation of why a programme exists comes before the explanation of whether I can use it. A description of the responsible organisation occupies the space where I would put the deadline. I don’t object to that background existing; I object to having to walk through it on every visit. This is hardly an argument against civil servants. GOV.UK’s own content guidance tells writers to start with what people need to do or find out, and to consider reducing, splitting or removing content accordingly. I would take that advice seriously enough to begin with a blank page rather than another round of repairs to the old one.

Before rewriting anything, I would ask the people answering the phones which questions keep coming back. Suppose applicants repeatedly ask which documents they need. That is a reason to give those documents a clear heading, near the application instructions, rather than add another sentence to an already sprawling introduction. I would arrange the page around the order in which someone actually tackles the job: checking whether they qualify before gathering papers and applying. Links should say where they lead. “Check which documents you need” gives me a reason to click that “Further information” does not. And a page whose only purpose is to send me to another page that sends me somewhere else should have to make a fairly convincing case for its continued existence.

Then comes the editing, which need not be mysterious. Imagine a notice for an invented scheme: “Individuals who are resident within the district and have attained the age of 18 years are permitted to submit an application online no later than 30 June.” I would write: “District residents aged 18 or over can apply online by 30 June.” The age limit, place, application method and deadline remain. What has gone is the long way round. “Apply online by 30 June” would be shorter still, but would lose the eligibility rules. That is a different notice, not a better edit. Below the shorter, complete version, I would put the required documents, naming them rather than assuming everyone knows what “supporting evidence” means.

I find the research on legal language encouraging here. In a 2022 study published in Cognition, Eric Martínez, Francis Mollica and Edward Gibson reported two experiments involving 184 participants: contract passages containing difficult linguistic features were understood and recalled less well than versions without those features. The changes included wording and sentence structure, so this was not proof that fewer words alone produced better understanding. It does, however, give me reason to resist the suggestion that difficult subject matter must be expressed in difficult prose. A condition need not disappear because it interrupts a sentence. It can have a sentence of its own, where I can read it without holding the unfinished beginning of another thought in my head.

The idea I keep returning to has the rather unhelpful name “minimum description length”. In statistics, it asks for a model that makes the complete description of the data as short as possible, counting the description of the model too. The whole description counts, not just its neatest part. I borrow that idea for communication rather than pretend it supplies a mathematical word limit. An unexplained abbreviation saves the writer space while leaving the reader to find out what it means. Add the definition, the searching and the rereading, and the apparent saving may disappear. I think of the alternative as “just enough information”: cutting what does no useful work while allowing the explanation someone needs. A short example may earn its place more honestly than three paragraphs of formal reassurance.

I do have to distrust my own pleasure in a tidy page. In research published in 2021, Martin Kerwer and colleagues found that adding structure to complex summaries of psychological research improved readers’ ratings of how easy they were to read and understand, without improving their scores on content knowledge. That distinction matters to anyone, including me, who likes a well-placed heading. Looking manageable is welcome, but it is not the same as making sense. I would still use headings; I would stop treating my approval of the layout as evidence that the job was done. A reader might find “Proof of address” immediately and still need an example of an acceptable document before being able to act.

I become less comfortable with “just enough” when the government gets to define enough entirely on my behalf. Someone applying for help may also need to understand why they have been refused, or how to challenge a decision. Those are not distractions from using a service. They are part of dealing with an institution that has power over you. A page designed only to move people towards its preferred outcome could be beautifully clear and still serve them badly. This changes what I would remove, and what I would test. The underlying rules need not sit above every application button, but the route to them must remain visible, with a usable explanation of any right to challenge a decision. Some of the information I would impatiently skip is exactly what another reader has come to find.

The practical test must therefore be more demanding than asking whether people like the new wording. The government’s service manual already recommends watching people attempt tasks to find problems with language and layout. I would compare drafts that way, including readers unfamiliar with the process and people using assistive technology, while having an expert check that the conditions remain accurate. Before choosing a winner, I would set limits on misunderstandings and completion time, with especially strict checks on mistakes that could cost someone the chance to apply. Each reader group would have to meet those limits; a good average should not conceal one group’s difficulties. Only then would I select the shortest version that passes. This proposed approach needs testing itself, but it gives an editor something firmer than taste to work with. When a draft fails, I would look at where readers went wrong, restore whatever they needed and try again.

Looking again at the pothole page, my eye settles on a detail beneath the postcode field: the example, SW1A 2AA. Those extra characters are unnecessary for me, and an editor concentrating on my experience alone might happily remove them. Someone less familiar with the format might have a use for them. The page is also more trustworthy about the next step. The example costs me almost nothing to pass over, though I might not have thought to put it there myself. I would leave it beneath the box for whoever comes along after me.