Who Gets a Say—and Who Decides?

What having a say can change

Greater governing capacity raises a question about power. Who may exercise it, whose interests should shape its use, and who can challenge the result? An institution might become better at analyzing consequences while leaving the people affected with little influence over the choices it makes. We need to examine the relationship between those who know, those who experience the consequences, and those authorized to decide.

Participation gives that relationship practical form. It can bring a missed experience into view, question a proposed purpose, or reveal that a burden considered minor by one person is substantial for another. It should also give people a way to influence consequential decisions without requiring them to become specialists in AI. The governing question is how participation, expertise, and legitimate authority fit together.

An explanation delivered after a decision can help people understand it. That matters, especially when they need to know how to act or seek review. But understanding a decision and influencing it are different achievements. A consultation can likewise receive thoughtful contributions while leaving the relevant choice closed. We should ask what remains open to influence and what can happen because someone contributes.

Four relationships help clarify the question. They can overlap: an expert may also be affected, and a decision-maker may need to hear from others before exercising authority. They describe what someone can contribute or do, rather than ranks of importance.

Having a say need not mean controlling the outcome. A contribution can change the question being considered, prompt an examination of missing evidence, or alter the conditions of a proposal. Decision-makers should give examinable reasons for treating material objections as they do, with usable routes to challenge. The important point is that relevant contributions can matter to the reasoning while consequential choices remain open, and that later challenge can lead to reconsideration where warranted.

Institutions should be clear about the scope of that influence. Asking people to help choose how an authorized service operates is different from asking whether it should operate at all. Either question may be appropriate. Presenting the first as though it includes the second would obscure a limit people need to understand. Where a decision is constrained by a binding duty, those constraints should be explained rather than disguised as options participants can freely choose.

We will first consider what participation contributes, then how expertise and authority differ. That will leave two practical questions: whether participation is usable, and whether a challenge can reach someone able to respond.

What the institutional account leaves out

Return to the hypothetical public institution in Chapter 2, using AI to change how people obtain a service. Imagine that, while a further change is being considered, officials report that people are completing the process more quickly. An affected person asks whether that account includes those unable to begin or finish it. The contribution changes the inquiry: whose access is represented by the reported improvement?

The person need not explain the model or propose a replacement system to raise a consequential question. Experience can reveal something the institution's chosen account does not describe. That does not establish that the reported improvement is false, or prove why anyone was excluded. It gives the institution a reason to examine the missing perspective before relying on the account to justify its choice.

Suppose inquiry finds that the faster route helps many people while imposing a substantial burden on a smaller group. Participation still matters after that fact is established. Whether the burden is acceptable, whether another route should remain available, and who should bear the cost of maintaining it are governing choices. More accurate measurement would inform those choices without deciding them.

This gives participation two connected purposes. People can contribute knowledge that improves a decision's grounds. They also have interests that deserve consideration, whether or not their contribution reveals a new fact. Someone's difficulty matters because of what the decision means for them, not only because the experience might improve the institution's analysis.

That distinction protects people from being valued solely as information sources. If an institution already knows that a proposed arrangement will make access harder for a group, it still needs to consider that burden. Inviting the group to describe it again is not a substitute for deciding whether it is justified. Conversely, listening seriously does not mean treating every claim as established or every requested remedy as appropriate.

Affected people can disagree with one another. Faster service may be a pressing benefit for some, while others need an alternative that makes administration more expensive. The institution must consider those differences, its obligations, and the feasible choices. Participation makes the conflict visible and gives people practical influence over its treatment. It cannot promise that everyone will receive the arrangement they prefer.

The useful possibility in this hypothetical is modest but consequential. A question about speed becomes a question about access, including people the original account may have omitted. No successful remedy has been demonstrated. Yet the proposed decision can now be examined on more relevant grounds. Competent listening begins by allowing such a question to change what the institution considers.

Expertise informs an authorized choice

Expertise is indispensable where a decision depends on understanding a system's capabilities, limitations, or likely effects. A qualified evaluator may identify a weakness others would miss. A public-service professional may understand an obligation or practical constraint that a technical designer has overlooked. People responsible for implementation may know which proposed safeguards could work and which would fail under actual conditions.

Their knowledge should carry weight appropriate to the question. Participation should not turn a factual matter into a popularity contest or ask non-specialists to supply judgments they cannot reasonably make. Experts should explain the grounds and limits of their conclusions well enough for others to understand the consequential choice. Making expertise accessible preserves its value while opening its use to scrutiny.

Some judgments require expertise and legitimate discretion together. An authorized professional may properly decide a particular matter within a defined remit. The distinction is not between experts who merely advise and everyone else who decides. It is between the knowledge a person possesses and the authority they may exercise. Where both belong to the same role, each still needs a basis and limits.

Technical competence alone does not confer a mandate to determine whose interests should prevail. An evaluator might estimate that one option would produce greater aggregate benefit. That estimate leaves open whether the distribution is justified, whether a protected interest would be violated, and which authority may make the choice. Disagreement about those matters cannot always be resolved by asking for a more accurate forecast.

Reforming a governing method can redistribute whose knowledge has influence. A new evidence requirement might bring excluded experience into view while diminishing the role of an established reviewer. Existing roles may preserve a necessary protection or leave a consequential gap; objections need examination on their merits. Expertise and concern about lost safeguards cannot simply be dismissed as resistance to progress. People affected should help determine which protections reform must preserve and whether the proposed arrangement gives their experience a usable place in judgment.

An institution also needs to explain the source of its authority. A firm's ownership of a system supports some decisions about its work, within applicable obligations. It does not confer unrestricted authority over people affected outside the firm. Public bodies exercise different powers under different mandates. A community can organize collective activity without possessing every power of a public authority. Knowledge shared between these actors does not make their decision rights interchangeable.

Nor does consultation create authority by itself. Gathering views may improve the grounds for a choice, but it cannot enlarge a mandate merely because participants support the proposal. Rights and binding obligations remain relevant when a majority favors a particular outcome. Legitimate participation should help identify and examine the limits of power, as well as influence its exercise.

AI assistance belongs within this relationship. A tool might help compare submissions, explain a technical issue, or examine possible consequences. Its value would depend on whether the assistance is reliable and useful for that task. Even then, a recommendation does not acquire authority because it is computationally sophisticated. Choosing what the tool should optimize, or which contributions it should treat as significant, may already shape the governing decision.

Those choices require identifiable responsibility. A decision-maker should be able to examine the grounds of an assisted recommendation and respond when someone questions a material omission. Placing a person at the end of the process would not establish that capacity. The arrangement must preserve the ability to exercise judgment within a legitimate remit, including the ability to seek further inquiry or reject the recommendation.

Participation, then, should improve the basis and answerability of a decision without pretending that knowledge eliminates political judgment. An institution can hear an objection, understand it, and still choose a different course for reasons it can defend. The obligation is to consider the objection appropriately and remain open to challenge, rather than manufacture agreement.

Disagreement can remain

Participation can improve a decision without producing consensus. A material objection may change the evidence examined or the reasons given even when the final choice remains contested. The useful question is how the objection affected the decision and whether the answer can be challenged. Agreement alone would not establish that the choice respected its obligations.

Participation must be usable

An opportunity to contribute has little practical value if the people most affected cannot use it. Time, language, technical knowledge, money, and institutional access can shape who is able to respond. The terms of participation should therefore be examined alongside the decision: what must a person understand, prepare, disclose, or endure to make a consequential concern heard?

An institution might offer an accessible explanation of the choice, receive contributions in different forms, or provide assistance where it is needed. Representation can help bring a concern into deliberation without requiring every affected person to attend. Each arrangement has limits. A representative cannot be assumed to express everyone's experience, and an invitation cannot establish that people understood or could use it.

The aim is practical influence proportionate to the decision. A narrow operational adjustment may warrant a different opportunity to contribute from a change in the terms under which consequential power is exercised. Existing channels may already provide relevant participation; a new consultation should earn the effort it requires. People should not have to repeat a concern through several processes simply because responsibility is divided inside the institution.

Protected reporting can matter here. A worker or user may have relevant information but reasonable grounds to fear the consequences of identifying themselves. Receiving confidential or anonymous evidence can support answerability rather than weaken it. The factual account still needs appropriate examination. Protection for the person reporting and responsibility for the authority responding serve different purposes and can coexist.

Participation also needs boundaries around information. Meaningful influence does not require making private records or sensitive security details public. Some reviewers may need access that others do not. The institution should provide enough intelligible information for the relevant contribution and scrutiny, while explaining justified limits. Confidentiality must not become a convenient reason to make every consequential claim impossible to question.

Tools may make some contributions easier to understand or compare, but their operation needs care. Imagine an AI-assisted summary that groups a small number of objections as minor because they are infrequent. A burden can be consequential even when few people report it. The method used to summarize should preserve material differences and provide a way to question omissions. A clearer document would not compensate for an important concern disappearing from the account.

Accessible participation takes resources, and a longer process can delay help or exhaust those it invites. Where needs are urgent, those costs matter. An unusably slow route to review can undermine accountability as well as efficiency.

We value public accountability over administrative efficiency.

Administrative efficiency is a genuine good. It can conserve resources for useful work and reduce the effort people spend obtaining help. The priority asks efficiency to serve an answerable institution. Saving staff time by making a challenge difficult may transfer a larger burden to someone less able to bear it.

The responsible choice is therefore not to maximize consultation. It is to preserve a usable opportunity to contribute and challenge, with scrutiny appropriate to the stakes. Where urgency limits participation before a decision, the institution should explain why, retain applicable protections, and provide appropriate reconsideration. Later review cannot undo every consequence, so urgency needs examination rather than automatic acceptance as a reason to exclude people.

Silence deserves care too. Someone may lack time, support, or confidence that contributing will matter. Others may simply choose not to participate. Their absence does not establish agreement or remove the interests the decision must consider. A governing arrangement should be capable of protecting people without making their protection conditional on successful participation.

Australia's automated debt administration illustrates the burdens of obtaining explanations and challenging decisions. In 2017, the Commonwealth Ombudsman reported that Centrelink's initial messages omitted the compliance team's phone number. Complainants struggled to obtain explanations and collect old employment records. The report found that more thorough manual intervention could sometimes have saved people time and effort. These findings concern an earlier automated administrative system, not a test of contemporary AI.

The 2023 Royal Commission into the Robodebt Scheme subsequently recommended clear routes to review automated decisions, plain-language explanations of the process, and independent scrutiny of rules and algorithms. These were recommendations, not evidence that the proposed safeguards had been implemented or worked. The lesson for the present argument is narrower: access to a system and a formal opportunity to respond can still leave the person carrying work the institution is better equipped to perform.¹

Competent assistance should help a concern reach the right place. People should be able to question a decision without first becoming expert navigators of the institution's internal divisions.

Decisions people can influence and challenge

The remaining question is what happens when a consequential choice is challenged. An institution may explain its decision clearly and receive objections courteously while lacking a route to reconsider it. Answerability requires more: a material concern must be able to reach someone with legitimate authority and practical means to obtain an appropriate response.

Chapter 3 distinguished a reviewer who establishes a problem from an authority able to correct it. Here that gap becomes a question about recourse. If one body can examine the evidence and another can alter the decision, the connection must work. Naming one person responsible cannot supply powers they lack or resolve a division of authority by itself. The institution should take responsibility for that connection rather than leave the person affected to discover it unaided.

Reasons help make the route usable. They should identify the consequential grounds of the choice, the relevant constraints, and how material objections were considered. A person can then question a mistaken premise, an omitted interest, or an exercise of power beyond its limits. A generic assurance that all relevant factors were considered offers little purchase for such a challenge.

Appropriate review should be able to lead to correction or remedy where warranted. Receiving a complaint, restating the decision, and reconsidering its grounds are different acts. The appropriate response depends on the matter and the authority involved; it may confirm the choice as well as change it. But a route that can never affect anything important provides only nominal recourse. Where consequences are time-sensitive, the response must also arrive in time to matter.

This does not give every objection a veto or promise an unlimited sequence of reviews. The process needs defined responsibilities and justified limits. Those limits should preserve meaningful examination of a material concern, including unwelcome evidence. A refusal to change a decision can remain answerable when its reasons are intelligible and appropriately reviewable. Refusing to hear the concern removes the basis for that examination.

Confidence may increase when people obtain competent help or correction. It may decrease when participation exposes a serious deficiency. Both are possible results of better scrutiny. The purpose is justified authority and practical agency, rather than a predetermined level of trust. An institution should not measure successful participation solely by how many people accept its preferred decision.

We should expect consequential choices to have an identifiable authority, grounds informed by appropriate knowledge and affected interests, and usable ways to influence or challenge their exercise. These conditions allow disagreement to be treated seriously while decisions remain possible. They also give governing capacity a public purpose: people can obtain help and question power without having to perform all the governing work themselves.

A legitimate decision needs credible grounds for judging accounts from experts, operators, workers, and affected people, whose access and limits differ. What must relevant actors know, and how can they know enough to act?