Structured Complementarity Civic AI Alignment Theory - Introduced in U.S. Civic AI Field Context (Alignment Traditions and the Missing Civic Layer)
Year: 2026
Type: article
Abstract: Contemporary AI alignment research has developed along several productive lines: long-termist concern with existential risk and technical control, justice-first documentation of automated harm in public administration, political-economy critique of behavioral data extraction, and design ethics centered on attention and individual agency. Each illuminates an essential facet of the alignment problem. This paper argues that one operating layer has not been sufficiently treated as a primary design target: the civic infrastructure through which democratic societies perceive, interpret, contest, and respond to emerging harms. The paper introduces Structured Complementarity Civic AI Alignment Theory (SCCAI-A): the intentional structuring of complementarity among different kinds of civic capacity. Non-Agentic Civic AI may extend the human ability to observe, preserve, organize, compare, translate, and inquire. Human beings retain the capacity to determine what those observations mean, what obligations they disclose, and what responsibilities follow. The alignment sought is therefore not an alignment of machine values with civic values, nor an attempt to model, predict, or represent collective democratic conscience. It is a constitutional division of labor between illumination and judgment. The paper's practical proposal is Civic Visibility Public Infrastructure: rights-bounded public institutions, procedures, data practices, and non-agentic instruments that preserve public records, display aggregate patterns, identify system-level pressures, and help institutions and communities recognize emerging conditions earlier - while categorically refusing to score individuals, recommend policy outcomes, trigger enforcement, or exercise delegated governmental judgment. The governing concept is moral visibility: the practical capacity of a free society to perceive consequential patterns, emerging harms, uncertainty, accountability pathways, and system-level drift in time to deliberate. The framework joins a Benjamin Franklin-inspired public-works ethic to a John Dewey-inspired democratic ethic of inquiry. Franklin supplies the imperative to build durable civic instruments; Dewey supplies the requirement that those instruments remain communicative, publicly intelligible, corrigible, and answerable to the people whose common life they illuminate. James C. Scott's warning about administrative legibility operates as a constitutional constraint rather than an incidental caution: such infrastructure must not make citizens legible to power, but must make power, risk, uncertainty, and institutional responsibility legible to citizens, communities, and accountable public servants alike. The paper does not claim to solve alignment. It identifies a prior condition under which technical, regulatory, and philosophical alignment work can become democratically usable, and offers that condition for open testing by any alignment agenda that takes democratic self-government seriously. Presented in the e-book series, Non-Agentic Civic AI — Illuminating America's Civic Digital Field. LITERATURE CONTRIBUTIONS: (1) Identifies an insufficiently addressed operating layer in alignment scholarship. Existing traditions address models, incentives, institutions of harm, and individual agency. This paper argues that the civic perceptual infrastructure surrounding those objects is itself an alignment variable, and that its degradation can render otherwise sound alignment work democratically unusable. (2) Introduces Structured Complementarity Civic AI Alignment Theory (SCCAI-A) as a distinct conceptual category. Rather than value alignment between machine and human objectives, the paper proposes alignment of AI's civic capabilities with the limited role democratic government may legitimately permit them to perform - a constitutional rather than a technical criterion. (3) Advances non-agentic design as a constitutional allocation of function. The paper reframes non-agency not as a capability limitation or interim safety measure, but as a durable separation of powers between illumination and judgment, with explicit categorical prohibitions: no individual scoring, no policy recommendation, no enforcement triggering, no delegated governmental judgment. (4) Defines moral visibility as civic infrastructure. The paper offers an operational definition and treats visibility as a public work to be built, funded, audited, and maintained, rather than as an epistemic virtue or media-literacy outcome. (5) Synthesizes the perspectives of Benjamin Franklin and John Dewey into a working design discipline. Public-works pragmatism supplies the imperative to build; democratic inquiry supplies constraints of communicability, corrigibility, and public authorization. The pairing yields testable design requirements rather than philosophical framing alone. (6) Converts James C. Scott's legibility critique into an affirmative design constraint. The paper specifies a directional test: visibility must flow toward citizens, communities, and accountable public servants rather than toward administrative power seeking to render populations legible. This provides a criterion by which proposed civic AI systems can be evaluated and, where warranted, rejected. (7) Extends structural-harm analysis from diagnosis to institutional response. Building on documentation of automated harm in public services, the paper proposes system-monitoring rather than person-monitoring instruments, designed to surface hidden harms, uncertainty, and overlooked stakeholders without judging or scoring anyone. (8) Offers a concrete institutional sketch for critique and empirical examination. The proposed U.S. Franklin–Dewey–Citizenry Civic Data Observatory System is presented as a testable public-works proposal whose components can be piloted in limited settings, publicly explained, independently audited, challenged by affected communities, revised in light of experience, and where warranted, stopped. (9) Supplies a compact governing formulation for interdisciplinary use, portable across policy, engineering, and democratic-theory audiences: AI illuminates. Humans judge. Lawful institutions decide.
Author: Robertson, Christopher Hunt
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7204447132
FWCI:
Cites:
Cited by:
Related to: