Learning Objectives

After this 180-minute seminar, participants will articulate how AI design, deployment, and governance choices affect equitable access to public services for communities of color, tribal nations, rural populations, seniors, people with disabilities, LGBTQ+ Americans, limited English proficiency speakers, immigrants, and low-income households. Learners will map the Blueprint for an AI Bill of Rights principle on algorithmic discrimination protections, EEOC technical assistance, OFCCP affirmative action obligations, the Justice40 Initiative, the Digital Equity Act, and Title VI of the Civil Rights Act to specific agency AI deployments. Participants will apply NIST AI RMF MEASURE function subgroup evaluation, OMB M-24-10 rights-impacting minimum practices, and GAO AI Accountability Framework equity elements to case studies including IRS ID.me, Michigan MIDAS, Allegheny County child welfare, Houston HISD EVAAS, COMPAS, and the Dutch childcare benefits scandal. Learners will design community engagement strategies that embed tribal consultation under EO 13175, linguistic access under EO 13166, disability access under Section 508 and the ADA, and rural inclusion through the Affordable Connectivity Program legacy and Broadband Equity Access and Deployment. Finally, learners will draft an equity impact assessment template, a subgroup monitoring dashboard, and an accessible appeals pathway for a rights-impacting AI in their own agency.

Key Topics Covered

The seminar covers equity dimensions in AI including race, ethnicity, gender, sexual orientation, age, disability, language, geography, and income; legal frameworks including Title VI, Title VII, Title IX, ADA, ADEA, Rehabilitation Act Section 504 and 508, Executive Order 13166 on limited English proficiency, Executive Order 13175 on tribal consultation, Executive Order 13985 on racial equity, EO 14110 on AI, OMB M-24-10, the Justice40 Initiative, and the Digital Equity Act; technical practices including subgroup evaluation under NIST AI RMF MEASURE, fairness metrics such as demographic parity, equalized odds, and calibration, data representativeness auditing, accessible UI design under WCAG 2.1 AA, and multilingual model considerations; community engagement through tribal consultation, language access committees, disability advisory councils, and community advisory boards in the Allegheny County model; case studies of IRS ID.me biometric exclusion, Michigan MIDAS unemployment failures, COMPAS and Houston HISD disparate impact, the Dutch childcare benefits scandal, CHIP AI enrollment barriers, and VA claims equity; and governance practices for equity impact assessments, subgroup monitoring dashboards, accessible appeals, and public transparency.

Why This Matters for Government

Government serves every citizen. That is the constitutional commitment. AI threatens and reinforces that commitment simultaneously. When a face verification service like ID.me cannot recognize darker skin tones reliably, it excludes Black Americans. When an AI voice bot supports only English and Spanish, it excludes speakers of Mandarin, Vietnamese, Hmong, Arabic, Somali, Navajo, and dozens of other languages routinely spoken in US households. When a mobile-first benefits application assumes smartphone ownership and broadband access, it excludes seniors, rural residents, and low-income households. When a recidivism scoring tool like COMPAS is trained on historically biased arrest data, it reproduces racial disparities at scale. When a child welfare screening algorithm uses public benefits receipt as a proxy for risk, it penalizes poverty. Government AI leaders must recognize these patterns, design against them, measure them, and remediate when they occur.

The legal framework is extensive. Title VI of the Civil Rights Act of 1964 prohibits discrimination on the basis of race, color, or national origin in programs receiving federal financial assistance. Title VII covers employment. The Americans with Disabilities Act and Rehabilitation Act Section 508 require accessibility. Executive Order 13166 requires agencies to take reasonable steps to provide meaningful access to persons with limited English proficiency. Executive Order 13175 requires consultation with tribal governments. Executive Order 13985 on racial equity directs agencies to advance equity and support underserved communities. Executive Order 14110 on AI incorporates equity protections. OMB Memorandum M-24-10 identifies rights-impacting AI and requires pre-deployment testing, ongoing monitoring, and public notice. The Blueprint for an AI Bill of Rights articulates the Algorithmic Discrimination Protections principle. The EEOC has published technical assistance on AI hiring tools under Title VII and the ADA. The Department of Justice Civil Rights Division has initiated investigations into AI discrimination under ADA and Title VI.

The Justice40 Initiative commits that at least 40 percent of the benefits of certain federal investments flow to disadvantaged communities. The Digital Equity Act of 2021 funds state digital equity plans. The Broadband Equity Access and Deployment program funds broadband infrastructure in unserved and underserved areas. Each of these programs intersects government AI deployment. Agencies deploying AI for benefits, licensing, or enforcement must measure Justice40 impact. Agencies deploying AI for services must ensure that digital divides do not become AI divides.

Case studies anchor the seminar. The IRS ID.me rollout in 2022 required biometric facial recognition for taxpayer authentication. After civil rights objections from the Electronic Privacy Information Center, members of Congress, and advocacy groups, Treasury reversed course. Key concerns included accuracy disparities across skin tones documented by NIST, exclusion of taxpayers without smartphones, and inadequate alternative pathways. Michigan MIDAS issued tens of thousands of false unemployment fraud determinations. Working-class families, disproportionately Black, were devastated. The system had no adequate appeals pathway, and human oversight was minimal. A class action settlement followed. COMPAS was analyzed by ProPublica in 2016, which found that the recidivism tool was twice as likely to falsely flag Black defendants as future criminals compared with white defendants. Houston Federation of Teachers v HISD struck down the EVAAS teacher evaluation model in 2017 for violating due process, with disproportionate impact on teachers of color. The Dutch childcare benefits scandal of 2021 exposed how the SyRI-like fraud risk model targeted dual-nationality families, overwhelmingly Moroccan and Turkish Dutch, driving twenty six thousand families into wrongful fraud accusations and collapsing the Rutte government. The Allegheny County Department of Human Services child welfare screening tool took a different path: published validation, community advisory boards, and shadow mode before production. Criticism remains, but the process illustrates a more equitable approach.

Subgroup evaluation is technical practice. Demographic parity checks whether selection rates are similar across groups. Equalized odds checks whether error rates are similar across groups. Calibration checks whether predicted probabilities match actual rates across groups. No single metric captures all fairness concerns, and they can trade off against each other. Practitioners must select metrics aligned with the specific decision context, document rationale, and test empirically. NIST AI RMF MEASURE function provides the governance anchor. EEOC's four fifths rule offers a legal threshold for adverse impact in selection procedures.

Community engagement is not a checkbox. Meaningful tribal consultation under EO 13175 requires government-to-government engagement, not stakeholder meetings. Language access under EO 13166 requires vital document translation, qualified interpreters, and multilingual digital services. Disability access requires Section 508 conformance under WCAG 2.1 AA, assistive technology testing, and human alternatives when AI fails for users with disabilities. Rural access requires attention to broadband availability documented in FCC broadband maps and BEAD plans. LGBTQ+ inclusion requires attention to name, pronoun, and gender marker handling. Immigrant inclusion requires attention to fear of government interaction and alternative documentation pathways. Community advisory boards with real authority, budgets, and term limits bring lived expertise to deployment design.

Governance practices include equity impact assessments pre-deployment, subgroup monitoring dashboards during deployment, accessible appeals with plain-language notice and explanation, tribal and community consultation records, and public transparency in the AI use case inventory. The GAO AI Accountability Framework Performance component supports this, as does the Blueprint principle on Notice and Explanation. Agencies that skip these practices reproduce the pattern of exclusion that AI is supposed to close. Agencies that invest in these practices build trust across communities.

Overview

The seminar opens with a framing that AI can widen or narrow equity gaps depending on design. It draws on NIST face recognition accuracy studies, DOL O*NET exposure data, and Pew Research broadband access data. The Blueprint for an AI Bill of Rights principle on Algorithmic Discrimination Protections anchors the discussion.

Legal Frameworks and Equity Obligations

Participants map Title VI, Title VII, Title IX, ADA, Rehabilitation Act Sections 504 and 508, EO 13166 on limited English proficiency, EO 13175 on tribal consultation, EO 13985 on racial equity, EO 14110 on AI, OMB M-24-10, the Justice40 Initiative, the Digital Equity Act, and BEAD. Each is tied to specific government AI deployment decisions.

Subgroup Evaluation and Fairness Metrics

The session introduces demographic parity, equalized odds, and calibration, discussing tradeoffs and context. Participants analyze subgroup evaluation under NIST AI RMF MEASURE function and consider EEOC's four fifths rule and adverse impact doctrine.

Community Engagement Models

Participants examine the Allegheny County community advisory model, tribal consultation protocols, language access committees, disability advisory councils, and community-informed procurement. Case studies include CHIP enrollment AI barriers and VA claim equity assessments.

Equity Governance Practices

Participants draft an equity impact assessment template, subgroup monitoring dashboard, and accessible appeals pathway. Each is peer-reviewed. The session concludes with commitments to civil rights officer coordination and ongoing community engagement.

Start Your CLUB Certification

This lecture is part of L5: AI Visionary, 160 hours of advanced training aligned with NIST AI RMF, OMB M-24-10, EO 14110, and the Blueprint for an AI Bill of Rights.

Related Lectures

L5 5.1.7 AI and National Workforce Transformation, 180 minutes. L5 5.1.9 Long Term AI Futures, 180 minutes. L4 4.3.1 Fairness in Government AI, 120 minutes, workshop. L3 3.3.5 AI Equity Impact Assessments, 90 minutes. L2 2.2.3 Civil Rights and AI, 90 minutes.