Equity & Ethics

Establish ethical procurement standards aligned with the Blueprint for an AI Bill of Rights and applicable privacy regulations.

Copy-ready expectations

• Vendor provides a plain-language Model Card (intended use, limitations, known risks) and an ACR/VPAT (Accessibility Conformance Report / Voluntary Product Accessibility Template) demonstrating conformance with WCAG (Web Content Accessibility Guidelines) 2.1 Level AA.
• Contract includes a Data Processing Agreement (DPA) and NDPA (National Data Privacy Agreement) addendum when appropriate.
• District retains IP (intellectual property) rights to teacher-created content and student work.

Sample RFP (Request for Proposal) prompts

• Describe your model’s training data sources and steps taken to reduce bias.
• Confirm no use of our tenant data for model training; attach your DPA/NDPA.
• Provide your ACR/VPAT and describe how you meet WCAG 2.1 Level AA.

Evaluate vendors on their commitment to algorithmic fairness. Include fairness criteria in procurement to ensure transparent, equitable AI tool evaluation.

What to ask vendors

• Do you publish model cards or impact assessments? Provide links or PDFs.
• How do you test for disparate impact across protected groups? Summarize findings.
• Can users contest or correct system decisions? Describe your appeals process.

District practice: Run a small fairness pilot with diverse sample data; document outcomes and mitigations.

• Ensure vendors comply with COPPA — Children’s Online Privacy Protection Act, FERPA — Family Educational Rights and Privacy Act, and Wisconsin statutes (Ch. 43.30 and Ch. 134.98).
• Prohibit use of PII (Personally Identifiable Information) for training models unless specifically authorized.
• Mandate permanent data deletion and written certification upon contract termination.
• Prohibit unauthorized third-party sharing and ad targeting.

Privacy by design (admin checklist)

• PIA/DPIA (Privacy/Data Protection Impact Assessment) completed for each new tool.
• Role-based access; encryption in transit (TLS 1.2+) and at rest (e.g., AES-256).
• Parent/family notices in plain language, with translation options.

All contracts should explicitly include Generative AI usage terms and expectations, ensuring compliance with district standards and legal requirements.

Require vendors to notify the district of any new Generative AI features or capabilities added to existing tools to ensure transparency and informed decision-making.

• Written change notice within 10 business days; provide opt-out where feasible.
• Updated ACR/VPAT and privacy documentation with each major release.

Encourage tools that minimize environmental impact through energy-efficient infrastructure and transparent reporting.

• Ask for basic sustainability metrics (e.g., carbon reporting approach, efficient hosting regions).
• Prefer vendors with published sustainability commitments and datacenter efficiency practices.

Use a formal vetting process (like the Selecting an AI Tool flowchart) to ensure tools:

• Align with district vision and instructional goals
• Meet age-appropriateness and data security requirements
• Protect intellectual property rights of the school, library, or tribal community

One tool that can help support research of these items is the EdTech Index, which showcases validations for safety, evidence-base, inclusivity, usability, and interoperability.

Adopt comprehensive strategies to address and prevent bias while promoting responsible data practices throughout the AI lifecycle.

• Require vendor bias documentation and publish a district “Bias Review” quick check (representation, sources, harms, mitigations).
• Use diverse, representative examples during pilots; collect feedback from students and families.

Teach how systems rely on patterns/probabilities rather than true understanding. This helps staff and learners spot confident but incorrect answers (“hallucinations”).

• Post “Use with Care” reminders near AI tools: verify claims, check dates, ask “What’s missing?”

Promote active questioning and corroboration with credible sources.

• Adopt a 3-source rule for factual claims; log changes made after verification.
• Use lateral reading (open new tabs from reputable sources) before sharing AI-generated information.

Ensure humans remain in decision-making roles—especially for grading, discipline, eligibility, or placement.

• Human–AI–Human workflow: humans define task → AI assists → humans evaluate and decide.
• Flag and route edge cases to a human reviewer; document outcomes and reasoning.

Provide professional learning on AI ethics, bias recognition, responsible use, and societal impacts.

• Short PD modules: prompt safety, privacy by design, bias checks, accessibility basics.
• Scenario practice: evaluate sample AI outputs; annotate risks and revisions.

• Improve infrastructure (devices, bandwidth, IT support) and offer non-AI alternatives when needed.
• Support multilingual and neurodiverse learners via captions, alt text, transcripts, and plain-language summaries aligned to WCAG 2.1 Level AA.
• Track participation and outcomes by subgroup; adjust support to close gaps.

Embed privacy from the start of Generative AI projects.

• Run PIA/DPIA before launch; redact PII in prompts and uploads.
• Publish parent/family notices explaining data flows, purposes, and choices.

Engage families and stakeholders on content moderation, data practices, and equitable Generative AI use.

• Hold public demos of AI tools with Q&A; post recordings and FAQs in multiple languages.
• Prepare transparent communications and response templates for potential data incidents (see Wis. Stat. 134.98 on breach notification).

Develop formal Generative AI policies addressing bias, privacy, vendor accountability, and transparency; include guidance for compliance monitoring and content attribution.

Copy-ready line: “Students and staff must disclose when AI is used, verify factual claims with credible sources, and submit final work in their own voice.”