Human by choice · Commentary · A Policy Window advocacy programme
A human check does not promise a conversation
Prepared Published Edition 3
In this reading
Human by choice · Policy Window’s advocacy programme
A fictional misuse
Imagine a community advice service citing a health-care survey as public backing for replacing conversations with staff-checked automated replies. That is a fictional misuse. The published comparisons report 70% favouring human checking of AI outputs over quicker test results; 55% favouring always receiving health advice from a person over quicker advice.
Those answers address different tasks and speed alternatives. They do not rank the human roles or reveal who chose both. A check need not provide contact, and health-care preferences cannot stand for preferences about this service. The table supplies neither item-level bases nor the full instrument; it measures no specified wait or cost.
The Health Foundation reports an overall sample of 8,000 UK public participants aged 16+, recruited through Censuswide’s online access panel with a telephone booster, surveyed 30 July–1 October 2025. The publisher already describes the trade-offs as illustrative.
Compare workable arrangements
Our proposal is to make the service choice concrete. One option could provide prompt automated help where it can safely answer the question. Another could offer a conversation, with more staff time and a longer queue. These are hypothetical conditions, not measured outcomes. Specify what a conversation could change and how someone would obtain one.
For each option, show the expected wait, staff cost and who does the checking. Distinguish a delay borne by everyone from extra help available to some. Consider people waiting, those whose circumstances do not fit a standard reply, and workers whose workload may move elsewhere. Any estimates need evidence from the actual service.
A serious comparison gives safe automation and quick access a fair hearing. Requiring a conversation for every enquiry could use time needed by someone else. Preserving every job is not the proposed decision rule. Retaining a human role needs a clear account of the contribution it provides and the burdens of providing it.
Make room to challenge the choice
Service users and workers should be able to challenge those arrangements before the decision. Let them identify a missing alternative, question the costs and disagree about priorities. Record who took part, whose experience is missing and what participants could influence. That input must be organised; it has not happened in this example.
A limited discussion cannot establish everyone’s agreement. The institution with actual authority must decide and explain why it accepts particular costs and objections. Its justification should name where automated help can suffice and where people must be able to reach someone. Set a review point and record what would change the plan—for example, if the promised conversation is inaccessible or checking work removes the expected speed advantage. This is our proposed application, not a new engagement method or an empirically validated result.
Source
Survey analysis and methods — Table 1 and Box 1.