Private Power without Epistemic Preemption

Research question

When does a private institution acquire a duty to keep an algorithmically mediated determination open to informed challenge and consequential reconsideration?

Proposed argument

The paper distinguishes a system’s capacity to support warranted belief from an institution’s practical power to structure a person’s options. Predictive performance does not, by itself, establish the legitimacy of a chosen standard, threshold, or allocation.

The proposed account of Relational Institutional Answerability concerns institutions that intentionally make a purportedly reasons-responsive judgment operative, foreseeably contribute to a nontrivial effect, control grounds needed to investigate a reasonable question, and act within a relationship lacking adequate countervailing routes.

The proposed argument connects the institution’s evaluative use of a judgment with responsibility for the epistemic and correction risks borne by the people subject to it. It asks whether control over both the judgment and the conditions under which it can be challenged, where adequate alternatives are lacking, constitutes a form of arbitrary private power.

The central burden is to explain why this relationship gives the affected person a claim to answerability, and why a proportionate route for informed challenge and consequential reconsideration is an appropriate constraint. Describing the power as arbitrary does not itself establish the duty. The paper must defend both that description and the response it proposes.

The proposed threshold

1. Operative evaluative use. The institution intentionally gives a purportedly reasons-responsive judgment practical effect.
2. Material effects. That use foreseeably contributes nontrivially to an individual or cumulative effect that is not merely negligible.
3. Asymmetric control. The institution or its accountable decision chain has exclusive or materially superior control over grounds needed to investigate a reasonable contestational question.
4. Constrained alternatives. The affected person lacks a reasonably adequate route for avoidance, verification, correction, or representation.

Argument in outline

1. Distinguish epistemic, practical, normative, and legal authority.
2. Reconstruct rational dependence on expertise, including the strongest case for relying on a superior system.
3. Distinguish contributory evidence use from epistemic preemption and from institutional constraints on action.
4. Examine how institutional standards determine which reasons and objections can enter review.
5. Defend the proposed answerability duty, its four-condition threshold, and its proportional limits.
6. Examine institutional routes, protected access, representation, and boundary cases.
7. Develop empirical implications while keeping behavioral evidence separate from normative justification.

What answerability adds

Information can support an objection while the institution remains unable or unwilling to act on it. Answerability connects an informed objection to a competent and sufficiently independent forum capable of giving reasons and producing consequences.

The route need not always involve full disclosure or personal technical mastery. It must make room for relevant grounds and an effective institutional response.

The standards of review

An institution may accept appeals while defining relevant evidence or admissible objections so narrowly that some material questions cannot receive consideration. A person might be allowed to correct the inputs to a score while having no effective route to question whether the score serves an appropriate criterion.

This raises a further question for the proposed account: when must the standards governing the review themselves remain open to challenge? The paper will examine how to distinguish a material objection from disagreement alone without treating the institution’s existing categories as the final authority on that distinction.

What Counts as a Reason supplies the test this paper adopts. An institution’s claim to have decided for reasons holds only where the person affected can raise the excluded objection, including an objection to the criterion of admissibility itself, before someone who must answer for the determination and could change it, and the objection is refuted with reasons rather than ruled out of scope. Paper B uses that test as the content of meaningful contestability: a route counts only if it supplies this participant in the middle, which an explanation, a transparency report, or an appeal form does not by itself. The test demands a channel, not a veto. The institution keeps its determination whenever it can give reasons and loses only the power to keep it by ruling out.

Relationship to the other papers

From Notice to Contestable Grounds develops the informational account where an opportunity to contest is independently required or promised. Paper B extends it and carries the separate normative burden of explaining when a private institution owes that opportunity.

The philosophical paper What Counts as a Reason: Institutional Formation and the Limits of Rational Reconstruction draws on Habermas to establish the test above and a disjunction: an institution that controls both what counts as a reason and what counts as an objection must either supply the channel or stop presenting its decisions as answerable to reasons. That paper stops there. Paper B takes the step it leaves open, when a private institution owes the channel, with the Habermas material supplementing rather than replacing the account of rational dependence on expertise that anchors this paper. Premises shared across the papers are counted once, and the philosophical inquiry may still clarify, qualify, or challenge the proposed account.

Boundaries and open questions

The theory must withstand a strong objection: a validated system may outperform the available reviewer, and poorly designed challenges may add inconsistency or strategic manipulation. The paper therefore asks what proportionate answerability requires even under favorable conditions of system reliability.

A person’s compliance does not prove belief in the determination. A reviewer’s private disagreement does not establish power to depart. The broader research program keeps beliefs, preferred actions, actual capabilities, and observed actions distinct.

Research status

The manuscript is in preparation, with a preprint planned for January 2027. This page is a public outline of the developing argument. It reports no completed empirical findings, peer-reviewed acceptance, or funding award.

Read Paper A’s informational architecture →