Nora’s place changes one morning although she has made no new application. The previous week she was among people likely to receive a call to view temporary housing; now she is forty-seven places lower. At the office she is shown a number, 61, and a sentence: “priority recalculated.” Nobody can point to the data that changed the result, the rule that carried most weight, or whether the calculation noticed that Nora left her former address to avoid living with someone who was harassing her. The number looks exact, yet it gives her no way to prepare an answer. In a housing service, waiting is not abstract when someone is sleeping in a borrowed room or living from one notification to the next.
A score can help organise a large number of cases. It is reasonable for a public service to need criteria when time and housing are scarce. The trouble starts when an ordering is treated as though it were a complete explanation. A list can place one application ahead of another, but it does not automatically say which facts were counted, which did not fit the form, what information is uncertain, or who can correct a mistake. Giving someone only a number shifts the work onto the affected person: they must guess why they are in that position and gather evidence without knowing what issue they are meant to challenge. A useful explanation is not the disclosure of technical secrets; it makes a conversation about a basic need possible.
Separate the outcome from its factors#
The first part of a clear explanation is a readable account of the factors involved. It might say that the service considers application date, household composition, a verified eviction notice, documented health conditions, and the suitability of available homes. Next to each factor, it should show the data used, its source, and when it was updated. “Vulnerability: 12 points” tells Nora very little if she cannot tell whether it comes from a social-work report, a box ticked in error, or evidence that is out of date. A short table in ordinary language lets a person locate the disagreement before a question becomes a race between service desks.
The first part of a clear explanation is a readable account of the factors involved.
The explanation should also state what the score does not do. If a circumstance can only be assessed by a human team, that should be explicit. If some information is deliberately not collected because it would be excessive, that protection belongs in the explanation too. If the calculation does not itself decide an allocation, the screen should not imply otherwise. Such limits prevent two opposite mistakes: giving a number authority it does not have and demanding that a tool know the whole life of a person asking for help. A respectful system is not one that gathers data until it looks all-knowing. It defines its reach and makes visible what still calls for judgement, listening, or additional evidence.
Let people see uncertainty as well#
Priority lists often communicate more certainty than the situation warrants. Data may arrive late, an address may be incorrectly standardised, a report may still need validation, or two households may be close enough that a small change reverses their order. Hiding that fragility turns an operational estimate into a verdict. Nora needs to know whether her place is stable, provisional, or conditional on a check. She does not need a page of formulas. A note such as “temporary position; change of address awaiting confirmation” explains why the figure may move and which step remains open.
Uncertainty cannot be an excuse for postponing an answer forever. It needs a deadline, a responsible team, and a stated consequence. If the office is waiting for a document from another authority, it can say when the case will be checked again and what happens if the document does not arrive. If there is time to provide clarification, the service must explain how to do so without requiring digital tools that not everyone has. Turning uncertainty into visible work offers more protection than hiding it behind a score that looks final.
Correction is not a message sent into empty space#
A correction route works when it turns a specific objection into a process that can be followed. Nora’s starting point is not simply “I disagree.” It is “the start date of my unstable accommodation is wrong,” or “my safety-related move is absent from the data considered.” The interface and in-person service should let her identify the factor, provide only necessary evidence, describe a circumstance that does not fit a field, and receive a receipt. That receipt needs a date, a way to ask questions, and confirmation that her application will not lose its place while the error is examined.
Correction must be accessible in more than one sense. Some people do not speak administrative language confidently, lack a stable internet connection, cannot print documents, or cannot disclose certain details in a shared waiting room. Offering appointments, telephone contact, advocacy support, and understandable formats is not decorative customer service. It prevents the right to contest a score from depending on skill at navigating a platform. The office can still minimise data: ask for what is needed to assess the disputed point, explain who will see it, and keep sensitive attachments no longer than necessary.
A second look that does not depend on the first calculation#
Correcting a fact will not settle every disagreement. Sometimes the question is whether a rule has a disproportionate effect in one particular case. That is why an independent review is needed before a final allocation or exclusion decision. Independent does not mean hostile, nor does it automatically mean slow. It means that the reviewer is not required to defend the score produced by their own team or provider. They should be able to read the factors, hear the person’s account, check the case file, and give written reasons for retaining or changing the priority.
Independence needs practical support. A reviewer must have authority to request verification, amend records, and recommend a reasoned exception where the rules allow one. The service needs separate timescales for urgent situations, so asking for review does not leave someone without an accommodation option. The result must explain its reasoning as plainly as the original calculation was expected to. “Review completed” does not answer Nora. A note distinguishing the corrected data, the rule applied, the reason another point was not accepted, and the date for a later review if circumstances change does.
Test the route with real cases, not only screens#
Before deploying an automated ordering, a service can test it with anonymised files and people who know the route through a housing application. The purpose is not to anticipate every possible case. It is to expose simple questions: can someone understand why their place moved? Can they spot a wrong date? Do they know which evidence is relevant? Does the explanation reveal another person’s information? Does the review arrive in time? Social workers, front-desk staff, tenants’ groups, and people who have applied for housing can identify barriers that never appear in a technical meeting.
Testing should lead to public, bounded changes. If many people misunderstand a label, rewrite it. If a data source creates recurring delays, mark it provisional or stop using it until it is repaired. If a group is consistently ranked lower because a data point poorly reflects its circumstances, examine the rule rather than only individual cases. Counting appeals is not enough: a low number may mean the route is clear, but it may also mean it is inaccessible. The service needs to ask who cannot use it and which specific barrier stops them.
Explanation should have a practical effect#
Explaining a score is not translating an algorithm into kinder words. It redistributes some power over a decision. Nora should be able to see her file, understand what supports her position, correct a verifiable error, and ask a person with no stake in the first outcome to examine the case before a door is closed. That does not guarantee immediate housing or remove scarcity. It does prevent scarce resources from becoming an absence of reasons as well.
A useful first step for any service is to read its most recent priority decision as though it were being received for the first time. The reader should find the factors, data, and dates; identify what is provisional; have a correction channel that leaves a record; and know who will review a disagreement and by when. When those answers are present, a score returns to what it ought to be: a limited aid for organising work, not an opaque wall between a person and the attention they need.
