Workflow-focused
Examine the people, information, prompts, systems, sources, decisions, actions, review gates, and consequences.
Assignment overview
The Responsible AI Case Analysis assesses your ability to examine generative AI use as a complete sociotechnical workflow rather than attributing responsibility only to a model. Select or use an approved case involving a proposed, existing, disputed, or failed AI-supported practice. Analyze the legitimate purpose, people involved, benefits and burdens, information flows, permissions, fairness, accessibility, privacy, security, authorship, integrity, oversight, accountability, contestability, and remedy.
Your analysis should distinguish facts from assumptions, identify missing evidence, compare realistic alternatives, and make a qualified recommendation. A responsible recommendation may support adoption with conditions, limited piloting, redesign, deferral, a non-AI alternative, or stopping.
Rights-aware
Address fairness, inclusion, accessibility, privacy, security, permission, authorship, integrity, and affected non-users.
Action-oriented
Propose proportionate safeguards, accountability, correction, contestability, remedy, escalation, and stopping.
Case-analysis principle: Naming an ethical principle is not enough. Explain who is responsible, what evidence is required, how the safeguard works, and what happens when the safeguard fails.
What the case analysis should demonstrate
- Define the case, purpose, scope, context, and decision requiring analysis.
- Separate established facts, claims, assumptions, unknowns, and disputed interpretations.
- Identify users, decision-makers, information subjects, affected non-users, communities, beneficiaries, and risk bearers.
- Map the complete AI-supported workflow and relevant human decisions.
- Apply FAIRER to representation, assumptions, accessibility, inclusion, participation, and unequal effects.
- Apply PROTECT to information classification, minimization, consent, retention, security, permissions, and incidents.
- Apply CREATE to copyright, licences, source permission, attribution, authorship, academic or professional integrity, and disclosure.
- Evaluate purpose, proportionality, alternatives, automation bias, and responsibility laundering.
- Identify applicable professional, organizational, community, or institutional authority.
- Design meaningful oversight, review gates, documentation, and escalation.
- Provide correction, contestability, redress, remedy, fallback, and stop conditions.
- Make a qualified, evidence-supported recommendation and state residual risk and uncertainty.
Select an appropriate case
Use a bounded case with identifiable decisions, people, information, risks, and possible remedies. The case may be supplied by the course or developed from public, synthetic, or explicitly authorized information.
Suitable case characteristics
- A clear AI-supported purpose or proposed use
- Identifiable stakeholders and affected non-users
- Meaningful fairness, accessibility, privacy, security, rights, authorship, or integrity questions
- A workflow with review, decision, action, and correction points
- Enough authoritative evidence for analysis
- Realistic alternatives and safeguards
- A scope suitable for one focused academic analysis
Cases to avoid
- Cases requiring confidential workplace, client, patient, student, or community records
- Cases so broad that affected people or decisions cannot be identified
- Live investigations of identifiable individuals
- Security testing or attempts to exploit real systems
- Cases requiring legal, health, financial, psychological, engineering, or other professional authority you do not possess
- Cases based mainly on speculation or unverifiable social-media claims
Approval point: If the case involves vulnerable people, high-consequence decisions, restricted information, Indigenous or culturally governed knowledge, or unclear authority, consult the course coordinator before proceeding.
Required analysis components
The official Assignment 3 instructions determine the final length and file format. The components below reflect the approved Module 3 learning activities and supporting course foundations.
Case definition and evidence
Describe the bounded case, legitimate purpose, context, AI role, decision, available evidence, missing information, assumptions, and non-AI alternatives.
Product: Case brief and evidence-status table.
Stakeholders and competing values
Identify beneficiaries, risk bearers, information subjects, affected non-users, decision-makers, communities, values, rights, duties, and power differences.
Product: Stakeholder and responsibility map.
FAIRER analysis
Examine representation, stereotypes, defaults, proxy variables, accessibility, intersectionality, participation, disparate effects, and corrective measures.
Product: Fairness and inclusion analysis.
PROTECT analysis
Classify information and assess purpose limitation, minimization, consent, retention, memory, secondary use, access, security, least privilege, and incident response.
Product: Information-governance and security analysis.
CREATE analysis
Distinguish permission, licensing, attribution, authorship, plagiarism, integrity, disclosure, and responsibilities for generated or source-based content.
Product: Rights, authorship, and disclosure analysis.
Recommendation, safeguards, and remedy
Compare alternatives and propose oversight, verification, approval, documentation, contestability, correction, redress, remedy, escalation, fallback, and stopping.
Product: Qualified recommendation and residual-risk statement.
Recommended analysis process
- Establish
Define the case, purpose, workflow, decision, evidence, assumptions, and alternatives. - Map
Identify people, information, authority, benefits, burdens, rights, and failure points. - Analyze
Apply FAIRER, PROTECT, CREATE, and relevant supporting concepts. - Recommend
Design safeguards, remedy, stopping, monitoring, and a qualified conclusion.
Show the evidence-to-recommendation sequence
- State the decision or recommendation question.
- Separate facts, claims, assumptions, unknowns, and disputed points.
- Map the existing or proposed workflow.
- Identify stakeholders, affected non-users, authority, and competing values.
- Apply FAIRER, PROTECT, and CREATE where material.
- Consider AI-supported and non-AI alternatives.
- Assess benefits, burdens, likelihood, severity, reversibility, and distribution.
- Identify safeguards, owners, evidence, review gates, and failure responses.
- Provide contestability, correction, remedy, fallback, and stop authority.
- Make a qualified recommendation and document residual uncertainty.
Evidence and source requirements
Use authoritative evidence
- Course readings and frameworks
- Peer-reviewed scholarship
- Applicable professional standards
- Authoritative organizational or institutional policies
- Privacy, security, accessibility, or human-rights guidance
- Official technical documentation
- Applicable laws or regulations, when verified and appropriately interpreted
Evaluate every source
- Authority and expertise
- Currency and continuing applicability
- Jurisdiction and scope
- Primary versus secondary status
- Independence and conflicts of interest
- Method and evidential strength
- Relevance to the actual case and affected population
Source rule: Do not rely on AI-generated citations, vendor claims, news summaries, or policy descriptions without checking the original authoritative source.
Suggested report organization
| Section | Primary purpose |
|---|---|
| 1. Executive summary | State the case, main findings, qualified recommendation, and major conditions. |
| 2. Case, context, and decision | Define scope, purpose, workflow, AI role, evidence, assumptions, and alternatives. |
| 3. Stakeholders and authority | Identify affected people, values, rights, duties, power, and decision authority. |
| 4. FAIRER analysis | Analyze fairness, representation, assumptions, accessibility, participation, and unequal effects. |
| 5. PROTECT analysis | Analyze information, privacy, security, permissions, retention, and incident responsibilities. |
| 6. CREATE analysis | Analyze permissions, licensing, attribution, authorship, integrity, and disclosure. |
| 7. Integrated risk and alternatives | Compare benefits, burdens, risks, non-AI options, and residual uncertainty. |
| 8. Recommendation and implementation conditions | Specify safeguards, owners, oversight, verification, contestability, remedies, and stop conditions. |
| 9. AI-use disclosure and references | Document assistance and provide verified sources. |
Follow the official instructions if they prescribe a different structure.
Participation pathways
Option A: Approved AI-supported case
Use an approved system only for bounded, low-risk analysis with public, synthetic, supplied, or authorized information.
Option B: Supplied case
Use a course-provided case containing synthetic records, prompts, outputs, policies, stakeholder information, or incident evidence.
Option C: Non-AI case analysis
Analyze an AI-related case using course readings, authoritative sources, supplied evidence, and conventional research tools without using live AI.
All pathways are assessed against the same outcomes. Live AI use is not required.
Responsible use, authorship, and boundaries
Required boundaries
- Do not upload or reproduce confidential, personal, privileged, proprietary, security-sensitive, unpublished, Indigenous, or culturally governed information without authority.
- Do not investigate, profile, diagnose, rank, or infer sensitive characteristics about real people.
- Do not test attacks, prompt injection, data extraction, surveillance, or vulnerabilities against live systems.
- Do not use AI to make consequential decisions or contact affected people.
- Do not claim legal, professional, institutional, or community authority you do not possess.
Student responsibility
- You must author the analysis, framework application, evaluation, and recommendation.
- Verify material factual and source-based claims.
- Distinguish evidence from assumptions and proposals.
- Document and disclose material AI assistance as required.
- Provide text alternatives for workflow and stakeholder diagrams.
- Preserve uncertainty and identify conditions requiring qualified guidance.
Assignment 3 resources and submission
Assessment details
Assessment: Assignment 3, Responsible AI Case Analysis
Weight: 20%
Recommended submission: End of Module 3
Official Moodle due date and time: ____________________
Required length and file format: ____________________
Before submitting
- Review the official instructions and rubric.
- Confirm all required components are present.
- Verify claims, citations, quotations, and references.
- Remove confidential or identifying information.
- Complete the required AI-use disclosure.
- Check accessibility, filename, and file format.
- Upload the correct file and verify submission status.
Course developer: replace the placeholder anchors with final Moodle URLs or companion HTML filenames.