A Public Light, Not a Digital Master

Could public institutions build a new kind of Civic AI - one that helps citizens and public servants see more clearly without acquiring authority to govern?

My new article considers that possibility through a Franklinian civic lens: when people recognize a shared need, they can build a public institution that helps others meet it.

I would welcome the perspectives of public servants, policymakers, librarians, archivists, civic technologists, cybersecurity professionals, and public administration colleagues on a question that reaches beyond the automation of government tasks.

What if Civic AI were designed not primarily to make decisions for people, but to strengthen their capacity to understand public institutions, inspect evidence, find assistance, ask informed questions, and participate in democratic life?

Imagine a resident trying to discover which office is responsible for an unresolved problem ...

Or a public servant helping someone navigate a service pathway whose records, requirements, and explanations are scattered across several systems ...

Or a neighborhood group seeking the documents needed to understand a public decision ...

Or a citizen trying to distinguish what a source establishes from what remains uncertain.

The proposed Civic AI would help make those civic pathways more visible. It would not decide whose claim should prevail, determine eligibility, adjudicate disputes, manufacture consensus, or replace the human institutions responsible for public judgment.

I explore this distinction in my new article, A Franklinian Way Forward for U.S. Civic AI: Civic Principles and Methods to Build Public Capacity - Guiding the Public Design, Ownership, and Operation of Non-Agentic Civic AI Infrastructure for Our Shared Future.

The work is not a claim that Benjamin Franklin anticipated artificial intelligence. Nor does it present an existing government program, a finished technical blueprint, or a machine solution to democratic difficulties.

Instead, it draws from Franklin’s civic method: recognize a shared need, gather useful knowledge, bring people together, build practical capacity, and maintain and improve what has been built.

Its central image is a lighthouse. A lighthouse does not choose a ship’s destination, take command from its captain, or make the sea safe. It provides light through which human beings may better recognize their surroundings and exercise their own judgment.

Could publicly governed Civic AI serve a similarly bounded purpose in the digital conditions of democratic life?

The article calls this approach Non-Agentic Civic AI: civic information infrastructure intended to illuminate public conditions while leaving governing authority, consequential decisions, and democratic responsibility with people and their accountable institutions.

For public servants, the proposed distinction is practical as well as constitutional.

Such infrastructure could help make authorized public records discoverable, explain institutional responsibilities in accessible language, trace explanations to their sources, reveal gaps in available information, and show lawful routes for participation, correction, appeal, or human assistance.

For citizens, it could make public life more intelligible without requiring them to accept a machine’s judgment about what they should believe or do.

For both, the purpose would be greater civic capacity, not greater machine authority.

The article develops seven principles for that proposed infrastructure:

• No masters, only aids: AI should assist understanding without becoming a civic authority.

• Useful knowledge must circulate: public information should be discoverable, understandable, source-traceable, and usable.

• Inquiry must precede judgment: the system should help people ask and examine questions rather than present generated conclusions as settled public truth.

• Association creates durable capacity: digital assistance should help people find human institutions and one another, not substitute synthetic interaction for civic life.

• Visibility must not become surveillance: public conditions should become more legible without turning citizens into objects of covert profiling, political targeting, or person-level scoring.

• Public utility requires public stewardship: the essential civic core should remain subject to enforceable democratic authority, with safeguards against private enclosure and partisan capture.

• Improvement requires humility, maintenance, and revision: the infrastructure should remain inspectable, contestable, correctable, auditable, and replaceable.

These principles lead to a broader institutional proposal: a U.S. Non-Agentic Civic AI Public Infrastructure System, coordinated through a Franklin–Dewey–Citizenry Civic Data Observatory based in Washington, DC.

The proposed observatory, and its nationally distributed local public subscription model, the Civic Digital Archivist AI, would help citizens and public institutions locate records, examine sources, understand responsibilities, recognize access barriers, and see uncertainty. This system would not become a national political brain, administer benefits, decide cases, score citizens, or replace public servants and democratic institutions.

The article proposes the archival, non-partisan Library of Congress as a possible institutional home. (This is the author’s proposal, not a Library of Congress initiative or endorsement.)

Public stewardship requires meaningful public control over the civic mission, governing rules, data protections, accessibility, audit rights, correction procedures, and continuity of essential services.

But public ownership alone is not a sufficient safeguard. The proposal calls for protections against capture by an administration, political faction, contractor, or other institution that might redirect the system away from its public purpose.

The central distinction is therefore not simply between public and private AI.

It is between Non-Agentic Civic AI Public Infrastructure that strengthens people’s capacity for self-government, and Agentic AI with mission creep that gradually acquires more authority over them.

A non-agentic system could be active in organizing, explaining, preserving, and illuminating information while remaining bounded against governing, coercing, or replacing public judgment. It could help make the civic field visible, but citizens and public institutions would still have to deliberate, decide, act, and answer for the consequences.

I would welcome the practical experience and concerns of those working in government and public policy:

• Where could source-traceable civic information most help residents and public servants navigate fragmented records, responsibilities, or service pathways?

• What would distinguish a genuinely non-agentic public resource from an automated decision system with a human nominally supervising it?

• How could civic visibility reveal institutional gaps without becoming surveillance of the people affected by them?

• What governance arrangements would protect such infrastructure from partisan capture, vendor dependence, or gradual expansion of its authority?

• How should correction, appeal, independent review, and non-digital assistance work in practice?

• Could libraries, archives, public agencies, and civic associations help build and maintain this capacity together?

The complete article can be quickly accessed gratis, without any website registration, by copying this Medium link to your browser:

medium.com/@crob2011/a-franklinian-way-forward-for-u-s-1cb2e2bb5d7d

You can also access it through a browser search of my full name.

Let us imagine a better path for our shared future.