Civic Impact Analysis

Analysis is not the same as criticism. A civic impact analysis of an AI system asks a structured set of questions — about what the system does, who it affects, what benefits it produces, what harms it creates, how it interacts with democratic institutions and values — and tries to answer those questions honestly using evidence. The goal is not to conclude that AI is good or bad; it is to develop the analytical capacity to evaluate specific systems with precision. In this lesson you will conduct a full civic impact analysis of a real AI system of your choosing. Before you begin the analysis, this lesson walks you through the framework you will use and illustrates each dimension with examples.

The Civic Impact Analysis Framework

A civic impact analysis examines an AI system across five dimensions. Dimension 1 — Function and Scale: What does the system do? What are its inputs and outputs? At what scale is it deployed — how many decisions per day, how many users affected, in how many countries? Scale matters because a system that harms one in a thousand users at a thousand decisions per day harms fewer people than one that harms one in a million users at a billion decisions per day. Dimension 2 — Information Environment Effects: Does the system affect what information people receive? Does it amplify certain content and suppress others? Does it contribute to personalization effects? Does it generate or distribute synthetic content? These questions apply most directly to recommendation systems, search engines, and content-generating AI, but any system that shapes what people know is relevant here. Dimension 3 — Rights and Equity: Does the system affect people's rights — to privacy, to due process, to equal treatment, to free association, to free expression? Does it affect different groups differently? Are there documented or plausible disparate impacts along lines of race, gender, age, disability, religion, national origin, or political affiliation? Dimension 4 — Power and Accountability: Who built the system? Who controls it? Who can modify, audit, or shut it down? Is there meaningful human oversight of consequential decisions? Who is harmed when it fails, and what recourse do they have? Is the system transparent enough that oversight is possible? Dimension 5 — Resilience and Governance: Is there adequate governance of this system? What laws or regulations apply? Are they adequate? Is there independent auditing? What mechanisms exist for correcting errors or harms? What would adequate governance look like if it does not currently exist?

To illustrate the framework, consider a brief worked example: a social media content moderation AI that automatically removes posts flagged as violating community guidelines. Function and Scale: The system reviews hundreds of millions of posts per day, making removal or retention decisions with limited human review. The scale makes individual case-by-case review of every decision impossible. Information Environment Effects: Automated moderation can suppress both genuinely harmful content and legitimate speech that shares surface features with harmful content. If the system has lower accuracy for certain languages or dialects (common in models primarily trained on English data), speakers of those languages lose access to expression that English speakers retain. Rights and Equity: Free expression is affected by removal decisions. The right to appeal is technically available but practically limited at scale. Evidence suggests automated systems are less accurate for African American Vernacular English and for content from certain regions, raising disparate impact concerns. Power and Accountability: The system is controlled entirely by the platform. No independent authority can review its decision logic. Users who are wrongly silenced may have no meaningful recourse. Regulatory pressure can shift platform behavior but is limited by First Amendment concerns in U.S. contexts. Resilience and Governance: Most jurisdictions lack AI-specific moderation oversight. The EU Digital Services Act imposes some transparency and audit requirements on large platforms; these represent the most developed governance framework currently in force.