Oversight
The 10 Civic Oversight Questions
Ten questions every resident, reporter, and council member should be asking. Designed to force a paper trail. Copy any question and send it — by email, in public comment, or as a public records request.
1. The AI Inventory Question
Please provide a complete list of every AI, algorithmic, predictive, automated, or machine-learning system currently used by this agency, city, county, school district, or state government, including the vendor, purpose, cost, and oversight process.
What it does
Forces the agency to produce — in writing — a complete inventory of every automated system it operates.
Why it works
Most agencies have never compiled this list. The act of producing it surfaces shadow tools, unsigned pilots, and quiet vendor deployments.
2. The Human Override Question
For every AI system currently in use, what decisions can a human overturn, and what decisions are effectively final once the system has made a recommendation?
What it does
Separates advisory tools from systems that are de facto decision-makers.
Why it works
Any decision that is effectively final triggers due-process obligations the agency cannot legally avoid.
3. The Bias Audit Question
What independent bias audits have been conducted on these AI systems, and will the full results be made available to the public?
What it does
Demands documented proof of fairness testing, not vendor assurances.
Why it works
The absence of an audit is itself a public record — and a headline. Agencies cannot defend systems they have never tested.
4. The Civil Rights Question
How has this agency evaluated whether AI systems disproportionately impact Black residents, low-income residents, immigrants, people with disabilities, or other historically underserved communities?
What it does
Invokes disparate-impact analysis under Title VI and state civil rights law.
Why it works
Any agency receiving federal funds is legally obligated to answer this. A non-answer is itself evidence.
5. The Surveillance Question
What technologies currently monitor, track, score, identify, predict, or analyze the behavior of residents, and what safeguards exist to prevent misuse?
What it does
Maps the full monitoring stack: cameras, license-plate readers, social-media scrapers, predictive policing, biometric ID.
Why it works
Most residents — and most council members — do not know what is already deployed in their own city. Disclosure changes the politics overnight.
6. The Vendor Influence Question
Which private companies currently provide AI systems to this agency, and what contractual protections exist to ensure public interests are prioritized over vendor interests?
What it does
Exposes contract clauses that let vendors shield their models as 'trade secrets' from the public.
Why it works
Procurement records are public. Vendor marketing rarely survives contact with the actual contract language.
7. The Data Ownership Question
When residents interact with government AI systems, who owns the data, how long is it stored, who can access it, and can residents request its deletion?
What it does
Forces clarity on retention, third-party sharing, and resident deletion rights.
Why it works
Vague or evasive answers create a documented record that supports later legal challenges and legislative reform.
8. The Accountability Question
If an AI system causes harm, denies a benefit, makes a mistake, or contributes to a wrongful decision, who is personally accountable for correcting that harm?
What it does
Demands a name — a specific human being accountable for the system's outcomes.
Why it works
'The algorithm decided' is not a legal defense. Officials behave differently when their name is publicly attached to the harm.
9. The Future-Proofing Question
What safeguards are being implemented today to prevent future AI systems from making decisions that undermine privacy, civil liberties, due process, or equal opportunity?
What it does
Shifts the conversation from reactive damage control to preventive policy.
Why it works
It puts officials on the record committing to safeguards they must later honor — or be confronted with.
10. The Democracy Question
Before deploying AI systems that affect residents, what opportunities are provided for public review, public comment, community input, and democratic oversight?
What it does
Establishes that secret deployment of AI is itself a violation of democratic norms.
Why it works
It creates grounds — political and legal — to challenge any system rolled out without public process.
Prompts
Civic Engagement Prompts
Copy any prompt into your AI of choice. Replace the bracketed fields with your local details.
Public Records Request (AI Inventory)
Draft a public records request to [CITY/COUNTY] asking for: (1) a complete inventory of all AI, machine learning, and automated decision-making systems currently in use by any department; (2) the vendor name, contract value, and term for each; (3) any bias audits, impact assessments, or civil rights reviews performed; (4) the name and title of the official responsible for oversight of each system. Cite the relevant state public records statute.
City Council Public Comment (2 minutes)
Write a 2-minute public comment for a [CITY] council meeting urging the council to pause adoption of [SYSTEM, e.g., predictive policing / Flock cameras / automated benefits screening] until an independent bias audit is published and a civilian oversight mechanism with binding authority is in place. Center the impact on Black residents. Include one specific local data point if I provide it: [DATA].
Letter to a Council Member — The Three Questions
Compose a respectful but firm letter from a constituent to [COUNCIL MEMBER NAME] asking three questions: (1) What AI systems does the city currently use? (2) What decisions are they involved in making? (3) Who is responsible for oversight and accountability? Request a written response within 30 days.
School Board Comment — Student Data & AI
Draft a school board comment opposing the deployment of AI-driven student monitoring, predictive discipline, or proctoring software in [DISTRICT] without parent opt-in, a public bias audit, and an appeal process. Reference disparate impact on Black students in discipline data.
Police Oversight Board Testimony
Prepare testimony for the [CITY] civilian review board challenging the use of [TOOL: COMPAS / ShotSpotter / Flock cameras / facial recognition]. Cite the documented disparate impact on Black communities, ask for the contract, training data source, and a moratorium pending independent audit.
Hospital Board — Medical Algorithm Equity
Write a 3-minute statement for the [HOSPITAL] board urging removal of race-based adjustments in clinical algorithms (eGFR, VBAC, pulmonary function) and asking what algorithmic tools are used in triage, risk scoring, and Black maternal care. Demand a public equity audit.
Vendor Contract Review Questions
I am reviewing a proposed contract between [AGENCY] and [VENDOR] for [SYSTEM]. Generate a list of accountability questions covering: bias audit clauses, data ownership, model retraining cadence, sunset/termination rights, indemnification for civil rights harm, transparency reports, appeal rights for affected residents, and a human-override requirement.
Op-Ed Draft — Algorithmic Monoculture
Help me draft an 800-word op-ed for a local paper explaining how a single biased Applicant Tracking System (e.g., Workday-class tools) can silently lock Black applicants out of hundreds of jobs at once, and call on [STATE/CITY] to require employer disclosure of automated hiring tools and a right to human review.