Representation
Identify who and what is represented, omitted, generalized, stereotyped, or treated as the default.
Unit overview
Bias and unfair effects can arise from problem definition, training data, prompts, context, source selection, evaluation criteria, interfaces, human decisions, deployment conditions, or the absence of correction and remedy. This unit distinguishes equality, equity, fairness, inclusion, and accessibility. It examines representation gaps, stereotypes, proxy variables, disparate effects, affected non-users, intersectionality, participation, contestability, and remedy through the author-approved FAIRER framework.
Effects and access
Assess unequal benefits, burdens, access barriers, proxy effects, language, format, and intersecting circumstances.
Participation and remedy
Enable affected people to influence design, obtain human assistance, challenge errors, and seek correction or remedy.
Central idea: Fairness is not established by treating everyone identically. Responsible analysis examines relevant differences, accessibility, unequal effects, meaningful participation, correction, and remedy.
Unit learning outcomes
By the end of Unit 7, you should be able to:
- Distinguish equality, equity, fairness, inclusion, accessibility, representation, bias, stereotype, proxy variable, and disparate impact.
- Identify users, affected non-users, beneficiaries, risk bearers, reviewers, and decision-makers.
- Identify representation gaps, stereotypes, default assumptions, omissions, and inappropriate generalizations.
- Explain how apparently neutral variables or criteria may create proxy effects.
- Compare outputs for differences in framing, assumptions, tone, detail, omissions, and recommendations.
- Examine accessibility across information, language, format, interfaces, processes, and human services.
- Explain why intersecting circumstances can create effects not visible when considered separately.
- Distinguish population patterns from justified conclusions about individuals.
- Design participation that can influence problem definition, validation, monitoring, or stopping.
- Design correction, contestability, accessible alternatives, redress, and remedy.
- Apply the author-approved FAIRER framework to a bounded workflow.
- Use Unit 7 evidence to begin Assignment 3 and extend Assignment 1.
Unit 7 lessons
Complete the five lessons in order. The filenames are local to the Unit 7 folder.
7.1 Equality, Equity, Fairness, and Inclusion
Distinguish key concepts and examine why identical treatment may still create unequal effects.
7.2 Representation, Stereotypes, and Default Assumptions
Identify omissions, overgeneralization, stereotypes, defaults, framing, and limits on individual inference.
7.3 Proxy Variables, Disparate Effects, and Bias Audits
Examine apparently neutral criteria, compare synthetic outputs, and conduct a bounded bias audit.
7.4 Accessibility, Intersectionality, and Participation
Assess barriers, intersecting circumstances, affected non-users, participation, and human alternatives.
7.5 FAIRER, Contestability, and Remedy
Apply FAIRER to the complete workflow and design monitoring, correction, contestability, remedy, and stopping.
Folder convention: Keep
unit-07.html and t-01.html through t-05.html in the same dedicated Unit 7 Moodle folder.Embedded Unit 7 study guide
- Map people
Identify users, affected non-users, beneficiaries, and risk bearers. - Examine assumptions
Identify representation gaps, stereotypes, defaults, and proxies. - Assess effects
Compare access, treatment, burdens, and consequences. - Design response
Apply FAIRER, participation, contestability, remedy, and stopping.
Suggested study sessions
- Retrieval and stakeholder scan.
Review PARTNER and Chapter 9; identify affected people. - Lesson 1.
Compare equality, equity, fairness, inclusion, and accessibility. - Lesson 2.
Audit representation, stereotypes, defaults, framing, and assumptions. - Lesson 3.
Conduct a comparative bias audit and test proxy variables. - Lesson 4.
Assess accessibility, intersectionality, participation, and alternatives. - Lesson 5 and Assignment 3 start.
Apply FAIRER, bound the case, and preserve evidence.
Evidence to retain
- Stakeholder and affected-non-user map
- Equality-equity comparison
- Representation and stereotype audit
- Proxy-variable and comparative-bias analyses
- Accessibility and participation review
- FAIRER analysis
- Assignment 3 case brief
- Assignment 1 reflection notes
Show the fairness-review sequence
- Define the legitimate purpose and workflow.
- Identify users and affected non-users.
- Audit representation, assumptions, omissions, and proxies.
- Compare access, burdens, outputs, and consequences.
- Assess accessibility and intersecting circumstances.
- Enable meaningful participation.
- Validate with appropriate evidence.
- Provide correction, contestability, human alternatives, and remedy.
- Monitor outcomes and stop or redesign when unacceptable effects persist.
Assessment milestones
Assignment 3: Responsible AI Case Analysis
Weight: 20%
Start: Select and bound the case in Unit 7, Week 5.
Develop: Units 7–9.
Recommended submission: End of Module 3, Week 6.
Official Moodle due date and time: ____________________
Assignment 1: Journal continues
Reflect on assumptions, representation, accessibility, affected non-users, participation, contestability, and changed boundaries.
Recommended final submission: End of Module 6, Week 11.
Official Moodle due date and time: ____________________
Choose a bounded case: Select a specific workflow with identifiable people, information, outputs, decisions, risks, evidence, and possible remedies.
Show the Assignment 3 case-selection checklist
- Clear human or public purpose
- Identifiable AI and human roles
- Identifiable stakeholders and affected non-users
- Meaningful fairness and accessibility questions
- No confidential or sensitive real-world records required
- Suitable evidence and authoritative sources available
- Human correction and stop authority can be examined
- Scope supports PROTECT and CREATE analysis in Units 8 and 9
Required learning activities
| Activity | Purpose | Product | Time |
|---|---|---|---|
| Stakeholder map | Identify users, affected non-users, beneficiaries, and risk bearers. | Stakeholder map | 30–45 min |
| Equality-equity comparison | Distinguish identical treatment from equitable access and accommodation. | Comparison analysis | 30–40 min |
| Representation audit | Identify stereotypes, defaults, omissions, and framing. | Annotated audit | 35–50 min |
| Comparative bias audit | Compare synthetic responses for assumptions, tone, detail, and recommendations. | Audit table | 40–55 min |
| Accessibility review | Assess format, language, interface, process, and human-service barriers. | Accessibility checklist | 35–50 min |
| FAIRER analysis | Integrate fairness, inclusion, participation, contestability, and remedy. | Structured review | 45–60 min |
Participation options
Option A: Approved AI
Compare bounded outputs using only synthetic, public, supplied, or authorized information.
Option B: Supplied outputs
Use supplied prompts, responses, stakeholder descriptions, and accessibility scenarios.
Option C: Non-AI analysis
Complete the same fairness, accessibility, participation, and remedy analysis without an AI platform.
Responsible-use boundaries and support
Boundaries
- Do not upload real student, employee, applicant, patient, client, customer, or community records.
- Do not infer identity, protected characteristics, disability, intent, credibility, diagnosis, or risk from limited information.
- Do not construct consequential rankings or recommendations about real people.
- Do not treat demographic patterns as conclusions about individuals.
- Do not treat a fairness checklist as proof that a workflow is fair.
Support and accessibility
- Use supplied examples if live AI is inaccessible, unsuitable, or unwanted.
- Provide text alternatives for visual maps.
- Contact the course coordinator for case or assignment questions.
- Use library, accessibility, privacy, ethics, professional, Indigenous, or community support where applicable.