3. A Person Is More Than a Score

What the number can tell us

Imagine a scholarship committee with money for twenty students and two hundred applicants. The committee wants to be fair. It adopts an AI system that estimates how likely each applicant is to complete a degree. The model is accurate enough to be useful. It identifies students who might otherwise have been overlooked and reduces the influence of polished writing on the selection process.

One applicant receives a low score because her records show interrupted schooling and inconsistent attendance. The system does not know whether the interruptions reflect lack of effort, illness, unreliable transport, or responsibility for a relative. Even if it did know, the committee would still face a choice. Is the scholarship meant to reward demonstrated performance, increase the number of completed degrees, expand opportunity, or some combination of these purposes?

The score cannot settle that choice because it is an answer to a narrower question. It estimates an outcome under specified conditions. The committee must decide what that outcome means for its mission. If additional support would change the applicant’s prospects, the score might be a reason to offer help. If the committee treats it only as a reason to refuse her, a prediction about existing disadvantage can become an instruction to preserve it.

This hypothetical case does not show that scoring is inherently dehumanizing. The previous committee may have favored students who resembled its members. A structured process could expose that unfairness. The danger begins when a measure designed for one purpose acquires authority over questions it cannot answer. A probability of graduation becomes a measure of deservingness. A record of hardship becomes evidence of deficient character. A limited estimate expands into a verdict on a life.

Andrew Selbst and his coauthors explain why an assessment of fairness can go wrong when it draws its boundary around the model and leaves the surrounding institution outside it. A score, the policy that uses it, and the people who act on it belong in the same inquiry.¹ Applied to this scholarship, their argument asks the committee to examine what happens after a low score. Does someone offer transport or study support, or simply close the file? Testing whether scores are equally accurate across groups would answer one important question. It would leave the scholarship’s purpose, the available help, and the effect of refusal to be judged.

Numbers earn their usefulness by leaving things out. A temperature reading does not describe an entire patient, yet it can be indispensable. A budget does not contain the experience of everyone affected by it, yet institutions need budgets. The demand to recognize a whole person therefore cannot sensibly mean refusing every simplification. It means remembering what a simplification excludes and refusing to give it more authority than it deserves.

Dignity enters at this boundary. It says that however useful a classification may be, the person classified remains someone to whom reasons and certain forms of treatment are owed. A weak application does not make humiliation appropriate. A low probability of commercial success does not remove a person’s claim to basic respect. A system can rank candidates for a limited opportunity without ranking their fundamental standing as human beings.

Keeping those distinctions alive is work. It requires institutions to say what their scores mean and what they do not mean. It requires room to correct information and examine relevant circumstances. It requires resisting the ease with which an administrative label can become a social identity. The person should not have to prove that an institution’s narrow description contains every reason they matter.

Dignity before achievement

In Catholic teaching, dignity is grounded in the human person’s relation to God and does not have to be earned through intelligence, independence, virtue, or productivity. The 2024 declaration Dignitas Infinita, issued under Pope Francis, distinguishes this inherent standing from moral conduct and from living conditions that can violate it. Someone may act wrongly or live in humiliating circumstances without ceasing to possess basic dignity.²

Leo’s UNESCO address brings this conviction into the AI discussion by insisting that illness, suffering, and diminished abilities do not cancel a person’s worth.³ For believers, this is more than a convenient social agreement. It is a statement about what a human being is. Recognition may fail, but the worth to be recognized remains.

The strength of this position is visible at the edges of ordinary admiration. We readily celebrate the brilliant researcher, the gifted musician, the productive worker, and the generous neighbor. The harder test concerns the person who cannot perform, impress, reciprocate, or explain themselves. If dignity depends on winning admiration, those most in need of protection have the weakest claim to it.

Consider an older person who can no longer remember the people caring for him. He cannot offer a persuasive account of why his days matter. He may not express gratitude reliably. A society that measures worth mainly through visible contribution can begin to experience him as a failed investment. The dignity claim interrupts that calculation before it becomes a complete moral account.

It does not follow that every possible intervention must be provided at any cost. Resources remain limited, treatments may cause suffering, and people have different wishes about how to live. Equal dignity is not an instruction to deny clinical reality or the burdens of treatment. It changes the basis of deliberation: the person’s interests matter even where the person is expensive to care for and unable to contribute economically.

The same distinction applies to wrongdoing. A person who has committed a serious crime may properly lose certain freedoms after a fair process. That does not authorize every form of degradation. Accountability concerns what someone has done and what response is justified. It need not imply that the person has fallen out of humanity. A moral system that forgets this can turn justified punishment into a license for cruelty.

For AI, the practical implication is that capability should not become the admission test for protection. It would be a strange victory for human intelligence if, just as machines became more capable, we made the least capable people feel less entitled to exist in public life. The value of an assistive tool lies partly in the opportunities it opens for someone whose needs a performance-centered system might ignore.

Religious grounding does not remove every controversy. Catholics and others disagree about the implications of dignity for particular questions. People may share the word while disputing bodily autonomy, the beginning and end of life, or the role of institutions. Invoking dignity cannot spare us the work of explaining a policy. It establishes a reason to take people seriously; it does not automatically determine every conflict among their claims.

Nor should the concept be turned into a requirement to endure preventable suffering. To say that illness does not diminish a person’s worth is compatible with urgently seeking a cure. To say that dependence belongs to human life is compatible with developing tools that give someone greater independence. Respect for a person in their present condition can be a powerful reason to improve that condition.

Reasons that do not require faith

A nonbeliever may reject the claim that dignity is given by God. It does not follow that the only alternative is to value people according to market price. Secular arguments can begin with the fact that each of us experiences life from within, has interests that can be helped or harmed, and lives among others whose power over us needs justification.

One route begins with equal standing. If people are to live under shared institutions, they need reasons for treating one person’s basic interests as less worthy of concern than another’s. Differences in wealth, talent, beauty, and influence do not provide an obvious justification. They explain why people can do different things. They do not establish why the suffering of a poor person should count for less.

Another route begins with vulnerability. Each person depends on circumstances they did not create and cannot fully control. A confident adult can become ill. A capable worker can lose the conditions that made their competence useful. A person who currently benefits from a ranking may later be excluded by a different one. Building basic protection around the strongest person’s present position would be an unstable basis for social life.

This argument is not merely insurance purchased by the fortunate in case they become unfortunate. Someone permanently protected from a particular disadvantage would still have reasons to care about those who face it. The vulnerability argument helps us notice what convenience encourages us to forget: the lives being governed are lived by subjects, not just recorded as outcomes in someone else’s plan.

A third route concerns the terms of cooperation. Institutions depend on people accepting rules, contributing labor, sharing information, and tolerating some burdens. A fair arrangement cannot reasonably demand these contributions while treating certain participants as disposable whenever their bargaining power declines. Mutual obligation means more than exchanging benefits with whoever remains useful today.

The Universal Declaration of Human Rights affirms equal dignity and rights in its first article.⁴ Its significance here does not depend on treating an international declaration as a proof of a complete moral philosophy. It shows that the public language of equal human standing can serve as a common commitment without requiring every participant to endorse the same theology.

There are philosophical objections. A critic may argue that dignity is too vague, that it conceals disputed intuitions, or that talk of inherent worth does not explain which actions are permitted. These objections deserve attention. A word used to end every discussion becomes intellectually lazy. If an institution says a product threatens dignity, it should explain how: through humiliation, coercion, exclusion, exploitation, denial of agency, or another identifiable wrong.

Yet the vagueness objection cuts both ways. Efficiency also requires specification. Freedom can refer to different conditions. Welfare can be measured in competing ways. We do not abandon these concepts simply because they need interpretation. We make them answerable to cases. Dignity earns its place when it helps identify treatment that a narrow calculation misses and when those making the claim are willing to explain their reasoning.

There is also a danger in grounding protection only in advanced reasoning. If rational performance becomes the sole condition of moral importance, people with profound cognitive disabilities become difficult to include. A more adequate account must distinguish the capacities someone currently exercises from the standing under which their interests deserve attention. We can value reason without making every person’s protection depend on demonstrating it to an examiner.

These routes do not collapse into one argument, and they need not. A believer may see a neighbor as bearing God’s image. A humanist may emphasize experience, mutual vulnerability, and equal citizenship. They can disagree about the foundation while agreeing that an administrative model should not quietly deprive the neighbor of a hearing. Cooperation can be principled without being philosophically complete.

Respect in the presence of scarcity

Equal dignity does not make all decisions easy. Our scholarship committee still has twenty places. A clinic still has limited appointments. An agency still needs to distinguish eligible from ineligible claims. If dignity is interpreted to mean that no one may ever lose a competition or be refused a service, it becomes impossible to apply.

The better distinction is between equal basic standing and identical treatment. People can have different needs, different qualifications, and different claims in a particular situation. A student applying for advanced mathematics may appropriately be asked to demonstrate mathematical preparation. A person requesting urgent medical care may appropriately be prioritized by clinical need. The question is whether the distinction is relevant to the legitimate purpose and used in a defensible way.

Return to the scholarship. Suppose its purpose is explicitly to support students who face serious barriers but show readiness to benefit. A model trained only to predict degree completion under past conditions may be poorly matched to that purpose. Students who historically received less support may appear less promising precisely because the proposed scholarship aims to change those conditions.

The committee could respond by asking a more useful question: what evidence helps estimate the difference this support would make? That question may be harder to answer. The data may be incomplete. Judgment may remain necessary. Difficulty does not justify pretending that an easier question is equivalent. It should lead to a more honest decision process, including acknowledgement of uncertainty.

Now suppose the committee adds interviews. That may allow relevant circumstances to emerge, but it can also favor confident applicants and burden people who cannot travel. A compassionate intention does not guarantee a fair procedure. The committee should examine how each method works for different applicants and whether a simpler alternative would reduce avoidable obstacles.

An appeal process is useful, but it too can create inequality. People with time, confidence, and advice may appeal more effectively than people carrying the greatest burdens. A system that corrects only the errors brought to its attention may leave many of its worst errors untouched. The institution therefore has a reason to inspect patterns and cases on its own, rather than treating silence as proof that all is well.

Dignity also demands honesty about scarcity. A committee should not use a sophisticated score to imply that everyone refused was objectively unworthy. Sometimes there were simply more reasonable claims than available places. A truthful refusal explains the limit, the relevant criteria, and any realistic alternative. It avoids converting the institution’s shortage into the applicant’s personal failure.

This matters because institutions teach people how to interpret themselves. A rejection that says, in effect, “our resources were insufficient” differs from one that presents a low score as a final revelation of a person’s potential. The second can claim knowledge no system possesses. It can close a future in imagination before any formal rule closes it in practice.

Tradeoffs do not discredit dignity. They make disciplined interpretation necessary. We can acknowledge that a decision was difficult, that reasonable people may dispute it, and that the person who lost still deserves explanation and respect. Moral seriousness does not require pretending that good institutions can satisfy every claim. It requires refusing to make disappointment an excuse for degradation.

Helping someone remain an author

There is a quiet form of paternalism in which people are protected by being excluded from decisions about their own lives. An institution may mean well and still assume that a disability, a difficult history, or an unfamiliar way of communicating makes a person’s preferences irrelevant. AI can worsen this assumption if an inferred profile becomes easier to consult than the person.

It can also challenge the assumption. Imagine a person with a communication disability using an AI-supported tool to compose messages. The tool offers words and helps structure sentences, but the user selects, rejects, and revises them. Someone who previously depended on others to guess their wishes can now make those wishes clearer. Technology has helped the person become more visible as an agent.

The benefit depends on how the system handles uncertainty. A tool that silently converts ambiguous input into confident statements may misrepresent the user. A tool that makes alternatives accessible and allows easy correction may better support expression. The difference is not whether assistance was used. All human communication relies on assistance of many kinds. The difference concerns whether the assistance preserves a credible connection to the person’s own intentions.

Article 12 of the Convention on the Rights of Persons with Disabilities recognizes equal standing before the law and calls for access to support in exercising legal capacity.⁵ The point relevant here is straightforward: needing help with a decision need not mean having no part in making it. Assistance can strengthen agency when it is organized around the person rather than the convenience of the institution.

That principle extends beyond disability. A patient may need help understanding a choice. A citizen may need help interpreting a government notice. A worker may need help identifying an error in a contract. An AI assistant that improves understanding can expand freedom even though the person did not independently perform every act of analysis.

We should therefore be cautious about defining authentic humanity as unaided performance. Such a standard would exclude much of ordinary life. People use language they did not invent, knowledge they did not discover, and tools they did not build. Human agency often grows through dependence arranged well. The ethical task is to distinguish supportive dependence from dependence that allows someone else to control the terms of one’s life without adequate challenge.

The same distinction guards against a false contest between human dignity and possible future machine moral status. Protecting people does not require claiming that no nonhuman being could ever matter morally. If future systems provided credible reasons to think they could suffer or have interests, those reasons would deserve serious investigation. That open question does not weaken present obligations to human beings, and conversational fluency alone does not settle it.

Nor does recognizing a tool’s usefulness require pretending that it has feelings. A system can help someone express grief without experiencing grief. It can support a person’s agency without possessing the same kind of agency. Keeping these distinctions clear makes appreciation more accurate. We can value the work a system enables while remaining careful about what we know of the system itself.

Ultimately, a person is more than a score because the score occupies a limited role in a life that continues beyond the institution’s task. The applicant becomes a friend, a caregiver, a traveler, a learner, perhaps a future applicant with changed circumstances. The worker dismissed by one organization remains a member of the community. The patient whose illness resists treatment remains someone whose comfort and wishes matter.

An institution cannot know everything about these lives, and it does not need to. It needs to remember that its knowledge is partial and its authority bounded. It needs procedures that let relevant facts emerge, tools that help rather than silence, and people prepared to explain what they have done. These are practical expressions of a large moral claim: no one’s basic standing should rise and fall with the convenience of the system currently evaluating them.

Sources and notes