COMP 3270 / Module 2 / Unit 6 / Lesson 1

Lesson 6.1 · Week 4 · Reading: Chapter 6

AI Roles and Human-AI Complementarity

Define the six AI roles, map complementary human and AI strengths, distinguish appropriate assistance from problematic substitution, and recognise why an AI role never confers professional, legal, moral, or institutional authority.

Estimated study time: 150–180 min Prerequisites: Units 1–5; SCOPE & PCAFE-V Module 2 · Week 4

Table of Contents

1. Lesson Overview

Before we can design meaningful human-AI collaboration, we must understand what roles an AI system can play and what complementarity means in practice. This lesson introduces the six AI roles defined in Chapter 6—Tool, Assistant, Critic, Simulator, Recommender, and Agent—and examines how each role shapes risk, responsibility, and review. You will learn to distinguish appropriate assistance from problematic substitution, map complementary human and AI strengths, and recognise why an AI system's assigned role never confers professional, legal, moral, or institutional authority.

This is the first of four lessons in Unit 6. It establishes the conceptual foundation for Task Allocation (Lesson 6.2), Oversight and Bias (6.3), and the full PARTNER integration (6.4). It also prepares you for Assignment 2: Prompt Engineering Portfolio by helping you classify roles and articulate boundaries in your own use cases.

Why this matters: Role confusion is one of the most common causes of overtrust and responsibility gaps in AI-supported work. If you cannot say precisely what role the AI is playing—and what it is not permitted to do—you cannot design meaningful oversight, and you cannot defend the outputs of the workflow.

2. Learning Outcomes

By the end of this lesson, you should be able to:

  1. Define and distinguish the six AI roles: Tool, Assistant, Critic, Simulator, Recommender, and Agent.
  2. Classify an AI-supported activity according to its primary role and identify mixed-role or role-drift situations.
  3. Explain why role classification matters for permissions, risk, review, and accountability.
  4. Identify the capabilities, limits, and principal risks associated with each AI role.
  5. Distinguish appropriate AI assistance from inappropriate substitution in learning, professional, and everyday contexts.
  6. Map complementary human and AI contributions for a bounded task and articulate role boundaries.
  7. Apply the P (Purpose) element of PARTNER to a human-AI collaboration scenario.

3. Before You Begin

4. Retrieval Practice: Concepts from Units 3–5

Draw on Units 3–5 to answer these questions. Check your understanding by expanding the feedback.

Question 1: What is the AI Paradox, and why does it matter for responsible use?

Show answer

The AI Paradox is the observation that generative AI systems appear highly capable and fluent while also being unreliable, inconsistent, and prone to error. This creates a dangerous mismatch between perceived and actual reliability, making it easy to overtrust outputs. Responsible use requires constant verification, clear boundaries, and human oversight.

Question 2: In the SCOPE framework, what is the purpose of the source hierarchy?

Show answer

The source hierarchy establishes which sources the AI system should prioritise (e.g., supplied documents over general knowledge, verified sources over unverified ones). It provides a structured way to control context and reduce hallucinations by telling the AI what to use and in what order.

Question 3: What does PCAFE-V stand for, and which element most directly affects how you classify an AI's role?

Show answer

PCAFE-V stands for Purpose, Context, Audience, Format, Evaluation, and Version. The Purpose element most directly affects role classification because it defines why you are using AI and what you expect it to do, which in turn determines whether the system acts as a tool, assistant, critic, etc.

5. Opening Scenario: "The All-Purpose AI Adviser"

📌 Scenario — The All-Purpose AI Adviser

A marketing team adopts a new generative-AI platform described by the vendor as a "strategic adviser, creative assistant, and decision-support tool." The team uses it to draft campaign copy, critique messaging, simulate customer reactions, recommend budget allocations, and even generate final approval memos. One team member says, "The AI is basically a junior strategist now." Another responds, "It's just a tool—we still make the final calls."

Questions:

  1. Which AI role(s) is the system actually performing? Is the role stable across the workflow?
  2. What authority is being implied by calling it a "strategic adviser" versus a "tool"?
  3. What risks arise from role confusion in this scenario?
Show guided analysis

1. Roles performed: The system acts as an Assistant (drafting copy), Critic (critiquing messaging), Simulator (simulating customer reactions), and Recommender (budget allocations). It may also cross into Agent territory if it generates approval memos without explicit human direction at each step. The role is not stable—it shifts across tasks, which increases confusion about what the system can and cannot do.

2. Authority implications: "Adviser" implies professional judgment and influence; "tool" implies bounded, mechanical function. Using both interchangeably obscures the fact that the AI has no legitimate decision authority. The team risks treating AI outputs as expert guidance rather than provisional material.

3. Risks: Role confusion can lead to overtrust, unverified recommendations being adopted, blurred accountability, and insufficient review. Team members may not know when to stop, escalate, or apply specialist judgment because they have not clearly defined what the AI is permitted to do.

6. AI Role Classification

An AI system's role defines what it is permitted to do, how it interacts with users, what outputs it produces, and what oversight is required. The same underlying model can play different roles depending on how it is prompted, configured, and integrated into a workflow. Role classification is a design choice, not a fixed property of the technology.

Table 1. Six AI roles — definitions, examples, and boundaries
Role Definition Typical output / action Suitable example Unsuitable example Principal risk
Tool Bounded function under direct human control; does not define purpose or make decisions. Reformatting, keyword generation, classification, test-case suggestion. Reformatting a bibliography to a required style. Using the tool to select which sources to include. Over-reliance on automated formatting without verifying accuracy.
Assistant Supports a person across one or more tasks while the person directs the work. Drafting, organising, outlining, question preparation. Drafting three alternative email openings for a supervisor to edit. Submitting the draft without review or revision. Passive acceptance of drafts; loss of personal voice.
Critic Evaluates material against supplied criteria; identifies weaknesses or gaps. Checklists, weakness identification, clarifying questions. Checking a proposal against a required section list. Using the critique as the final quality assurance sign-off. Treating provisional feedback as authoritative.
Simulator Creates hypothetical interactions, scenarios, or environments for exploration or practice. Fictional dialogues, role-play scenarios, projected outcomes. Simulating a job interview for practice. Presenting a simulated outcome as an actual prediction or assessment. Blurring fiction with reality; overconfidence in simulated data.
Recommender Compares alternatives or proposes a course of action based on supplied criteria. Rankings, comparisons, suggested choices. Comparing three software tools against a feature checklist. Making the final procurement decision based on the recommendation alone. Treating recommendations as decisions; losing contextual judgment.
Agent Pursues a multi-step goal, may use tools, retrieve information, or perform external actions. Workflow orchestration, multi-step drafting, scheduling under constraints. Organising a multi-stage drafting workflow with human checkpoints. Automatically sending emails or modifying records without human approval. Unauthorised actions; permission creep; loss of human control.
Remember: An assigned role does not give the AI system professional, moral, legal, academic, institutional, or community authority. The human remains accountable for all outputs and decisions.
Section summary
  • There are six AI roles: Tool, Assistant, Critic, Simulator, Recommender, and Agent.
  • Role classification is a design choice and shapes permissions, risk, and oversight.
  • An assigned role never confers authority on the AI system.

7. Complementarity

Complementarity is the deliberate combination of different strengths and responsibilities between humans and AI systems. It is not about replacing human capability but about augmenting it while preserving human judgment, relationships, and accountability.

Table 2. Complementary contributions (illustrative — not exhaustive)
AI contributions Human contributions
Rapid generation, variation, and reformatting Defining purpose, context, and values
Pattern-based comparison and classification Evaluating evidence, exercising professional judgment
Provisional summarisation and organisation Interpreting exceptions, recognising relational/cultural meaning
Repetitive transformation and simulation Making value-laden decisions, authorising actions, correcting harm
Generating alternatives for consideration Providing care, direct service, and accountability
Key insight: These are not fixed categories. System capabilities and human expertise vary by context. The goal is to design a workflow where each party does what it does best, and where the human retains authority over decisions that matter.
Section summary
  • Complementarity combines AI speed and consistency with human judgment and accountability.
  • It is about augmentation, not replacement.
  • The human retains authority over decisions that matter.

8. Assistance vs. Substitution

Assistance occurs when AI supports a human capability without replacing the human activity that provides purpose, learning, judgment, relationship, or accountability. Substitution occurs when AI replaces an activity that should remain human.

Table 3. Assistance vs. substitution — examples
Appropriate assistance Problematic substitution
Using AI to draft an outline that you revise and expand Submitting AI-generated content as your own assessed work
Using AI to compare product features before you decide Letting the AI choose without your contextual review
Using AI to summarise a source you have already read Replacing source reading with an unverified AI summary
Using AI to simulate a conversation for practice Using AI to replace a real conversation with a vulnerable person
Using AI to generate alternatives for a professional report Using AI to generate the final professional recommendation without review
Test: Ask whether the AI is displacing purpose, learning, judgment, relationship, or accountability. If it is, you are likely crossing into substitution.
Section summary
  • Assistance supports human capability; substitution replaces it.
  • Substitution is problematic when it displaces purpose, learning, judgment, relationship, or accountability.
  • The assistance/substitution line depends on context and the nature of the activity.

9. Role Boundaries and Role Drift

Every AI role must have clear boundaries: what it is permitted to do, what it is prohibited from doing, what review is required, and what permissions are necessary. Role drift occurs when an AI system gradually expands beyond its intended role—often because users ask it to do more, or because the system is configured with broader permissions. Role drift increases risk, blurs accountability, and can lead to unauthorised actions.

Common indicators of role drift:
  • An assistant begins making recommendations without being asked.
  • A tool is used to make decisions that should be made by a human.
  • A critic's feedback is treated as final quality assurance.
  • A simulator's outputs are used as empirical predictions.
  • An agent performs actions that were not explicitly approved.

To prevent role drift, document the role as part of your workflow design, communicate boundaries to all users, and review permissions regularly.

Section summary
  • Every AI role must have explicit boundaries and permissions.
  • Role drift occurs when the AI gradually expands beyond its intended role.
  • Document the role, communicate boundaries, and review permissions regularly.

10. PARTNER Walkthrough — Purpose

PARTNER is the framework for human-AI responsibility and collaboration used throughout Unit 6. In this lesson, we focus on the first element: P — Purpose.

Purpose asks: What legitimate human purpose does the collaboration serve?

To define purpose, consider:

Table 4. Guiding questions for Purpose
Question Why it matters
What human purpose does the collaboration serve? Ensures AI is used for a legitimate, meaningful goal.
Why is AI being considered? Prevents "AI for AI's sake"; clarifies the value proposition.
What must improve? Identifies the desired benefit (e.g., speed, consistency, scale).
What must not be lost? Preserves what matters: human judgment, relationships, learning, accountability.
Is collaboration preferable to human-only or non-AI alternatives? Forces a comparison and justifies the use of AI.

Common failures in defining purpose:

🧩 Brief application

Think of a task you might use AI for (e.g., drafting an email, summarising an article, generating ideas). Write a one-sentence purpose statement that includes the problem, the intended outcome, and the boundary on AI involvement.

Show example purpose statement

"To draft a polite and clear email to a colleague requesting feedback on a project proposal, where the AI generates alternative openings and closings but does not determine the substantive content or tone, which remains my responsibility."

This statement identifies the problem (requesting feedback), the outcome (a draft email), the AI's boundary (alternatives only), and preserves human authority over content and tone.

Section summary
  • Purpose is the first element of PARTNER.
  • It defines the legitimate human goal and the boundaries on AI involvement.
  • A clear purpose statement includes the problem, outcome, and AI boundaries.

11. Process Walkthrough: From Purpose to Role to Boundaries

This walkthrough shows how to move from a human purpose to an AI role classification, articulate complementarity, and define boundaries. We use a bounded academic task: generating discussion questions for a lesson.

Step 1 — Define the human purpose

"I need to generate 5–7 discussion questions for a 300-level course on digital ethics. The questions should provoke critical thinking, connect to the week's readings, and be suitable for a 45-minute small-group discussion."

Outcome: A set of discussion questions. Boundary: AI generates draft questions; I will review, revise, and select the final set. I remain responsible for pedagogical quality.

Step 2 — Classify the AI role

The system generates alternatives based on supplied readings and criteria. It does not make decisions, act externally, or simulate interactions. This fits the Assistant role: it supports my work while I direct, review, and make the final choices.

Alternative: If the system also evaluated the quality of the questions against a rubric, it would also play a Critic role. In this case, we keep it as a pure Assistant.

Step 3 — Map complementarity

  • AI contributes: Rapid generation of varied question phrasings, formatting, and organisation.
  • Human contributes: Selection of readings, criteria definition, pedagogical judgment, revision, and final approval.

Step 4 — Define role boundaries

  • Permitted: Generate draft questions based on supplied readings and criteria.
  • Prohibited: Select the final questions, determine the discussion structure, or assess student learning.
  • Required review: Every question must be reviewed for relevance, clarity, and pedagogical value.
  • Required permission: The instructor (me) must approve the final set.
  • Stop condition: If the AI repeatedly produces irrelevant or off-topic questions, stop and revise the context or discontinue use.

Step 5 — Assistance vs. substitution check

Assistance: The AI helps me generate drafts; I retain full control over selection and revision. My pedagogical judgment and relationship with students are preserved.

Not substitution: I am not delegating the intellectual work of crafting meaningful questions; I am using the AI to spark ideas and save time on formatting.

Failure points to watch: If I start using the AI's questions without review, or if I let the AI determine which readings to use, I have crossed from assistance to substitution. If the AI's role expands to suggesting grading criteria, that is role drift.

12. Worked Example: "Job Description Drafting Assistant"

A human resources coordinator uses an AI system to draft job descriptions. This worked example shows the complete classification and boundary-setting process.

Human purpose: To draft a clear, inclusive job description for a new software-engineer position, based on a supplied role profile and organisational templates. The HR coordinator will review, revise, and publish the final version.

AI role: Assistant — the system generates draft text based on structured inputs (role, skills, responsibilities) and follows formatting templates. It does not define the role, make hiring decisions, or act on the description.

AI contributions: Generating 3–5 draft versions, formatting according to template, suggesting inclusive language alternatives, and organising sections.

Human contributions: Defining the role profile, selecting key responsibilities, reviewing drafts for accuracy and tone, making final edits, approving the description, and assuming accountability for the content.

Boundaries:

  • Permitted: Generate drafts, suggest language variations, format text.
  • Prohibited: Make decisions about role requirements, salary, or qualifications; publish the description; respond to candidates.
  • Required review: Every draft must be reviewed by the HR coordinator and, if needed, the hiring manager.
  • Stop condition: If the AI repeatedly generates non-compliant or inaccurate descriptions, escalate to the HR director and consider using a non-AI template.

Assistance vs. substitution: This is assistance because the AI supports drafting and formatting while the HR coordinator retains full authority over the content and final approval. The coordinator's professional judgment is not replaced.

Show analysis of role classification

Why Assistant, not Tool? A tool typically performs a single bounded function (e.g., reformatting). Here, the system generates multiple drafts, suggests alternatives, and organises content—more than a simple tool, but still under direct human direction.

Why not Agent? The system does not act externally, maintain state across sessions, or pursue a multi-step goal without human intervention.

Why not Recommender? It does not compare alternatives or rank options; it generates text based on supplied criteria. Thus, Assistant is the most accurate classification.

13. Self-Check Quiz

Seven questions · Formative · Ungraded. Expand each item to see feedback.

Question 1: Which AI role is best described as "performs a bounded function under direct human control, without defining purpose or making decisions"?

Select one:
Show answer and explanation

Answer: b) Tool.

Explanation: The definition matches the Tool role exactly. An Assistant supports across tasks but still operates under human direction; a Critic evaluates; a Recommender compares alternatives.

Question 2: An AI system generates three possible project timelines and ranks them by feasibility, but a project manager makes the final choice. Which role(s) is the system playing?

Select one:
Show answer and explanation

Answer: d) Assistant and Recommender.

Explanation: Generating timelines is an Assistant function; ranking them introduces a Recommender element. The system is playing a mixed role. This is common in practice and requires careful boundary-setting.

Question 3: Which of the following is an inappropriate substitution of AI for human activity?

Select one:
Show answer and explanation

Answer: c) Using AI to generate a final, unrevised recommendation for a client.

Explanation: This substitutes human professional judgment with AI output. The other options are assistance (with appropriate review) or practice.

Question 4: What is role drift?

Select one:
Show answer and explanation

Answer: b) When an AI system gradually expands beyond its intended role.

Explanation: Role drift is a design and governance failure. It often happens incrementally, making it harder to detect and correct.

Question 5: An AI simulator creates a fictional patient interaction for medical students to practice diagnosis. Which boundary is most important to communicate?

Select one:
Show answer and explanation

Answer: a) The simulation must not be represented as an actual patient.

Explanation: Simulators create hypothetical interactions. The critical boundary is that they are not real—they cannot substitute for actual patient interaction or professional assessment.

Question 6: Which of the following is not a typical human contribution in a complementarity map?

Select one:
Show answer and explanation

Answer: b) Rapid generation of alternatives.

Explanation: Rapid generation is typically an AI contribution. Humans define purpose, exercise judgment, and assume accountability.

Question 7: An AI system is described as a "strategic decision-support tool." Which risk is most directly associated with this label?

Select one:
Show answer and explanation

Answer: a) Role confusion about whether the AI is a tool, assistant, or recommender.

Explanation: "Decision-support" implies recommendation, while "tool" implies a bounded function. The label is ambiguous and may lead to overtrust or unclear accountability.

14. Exercises

Complete these five exercises to apply what you have learned.

📝 Exercise 6.1.1: AI Role Classification
  • Learning outcome: Classify AI roles
  • Estimated time: 10 min

Scenario: A financial analyst uses an AI system to scan thousands of news articles, flag those mentioning specific companies, and generate a one-paragraph summary of each flagged article. The analyst reads each summary, selects relevant ones, and incorporates them into a weekly report.

Instructions: Classify the AI's primary role. Provide a brief justification.

Show hint

Consider whether the system is generating new content, evaluating content, simulating, recommending, or acting across steps. Here, the system performs a bounded, direct function under human control.

Show sample answer

Answer: Tool (with some Assistant characteristics).

Justification: The system performs a bounded function—scanning, flagging, and summarising—under direct human control. The analyst directs the task, reviews all outputs, and makes the selection decision. It is not making recommendations, simulating, or acting independently. If the system also ranked the articles, it might move toward Recommender.

📝 Exercise 6.1.2: Distinguish the Six Roles
  • Learning outcome: Distinguish the six roles
  • Estimated time: 15 min

Instructions: For each of the following, identify the primary AI role and explain why.

  1. A system that reformats a bibliography according to APA style.
  2. A system that generates three draft marketing emails for a manager to review and edit.
  3. A system that reviews a project proposal and highlights sections that lack evidence.
  4. A system that creates a fictional customer-service interaction for training purposes.
  5. A system that ranks five potential vendors based on supplied criteria and suggests a preferred choice.
  6. A system that organises a multi-stage workflow, retrieves data from a database, and generates a final report after human approval at each stage.
Show hint

Map each activity to the definitions: Tool (bounded function), Assistant (supports across tasks), Critic (evaluates), Simulator (creates hypotheticals), Recommender (compares/ranks), Agent (multi-step, uses tools, may act).

Show sample answers
  1. Tool — bounded reformatting under direct control.
  2. Assistant — supports drafting; human directs and reviews.
  3. Critic — evaluates against criteria (evidence).
  4. Simulator — creates a fictional interaction for practice.
  5. Recommender — compares and suggests a choice.
  6. Agent — multi-step, uses tools, maintains state, but with human approval.
📝 Exercise 6.1.3: Identify Role Boundaries
  • Learning outcome: Define role boundaries
  • Estimated time: 12 min

Scenario: A public-health communicator uses an AI Assistant to draft social-media posts about health guidelines. The Assistant generates text based on verified sources.

Instructions: Identify:

  1. Permitted AI activity
  2. Prohibited AI activity
  3. Required review
  4. One stop condition
Show sample answer
  • Permitted: Generate draft posts based on verified sources; suggest phrasing variations.
  • Prohibited: Select which sources to use; make claims not supported by verified sources; publish posts; respond to comments.
  • Required review: Every draft must be reviewed by the communicator for accuracy, tone, and public-health alignment.
  • Stop condition: If the Assistant repeatedly generates content that contradicts verified sources, stop and escalate to a supervisor.
📝 Exercise 6.1.4: Create a Complementarity Map
  • Learning outcome: Map complementarity
  • Estimated time: 15 min

Scenario: A project manager uses an AI system to generate weekly status reports. The system pulls data from project-management software, drafts a narrative summary, and identifies potential risks. The manager reviews, edits, and distributes the report.

Instructions: Create a complementarity map showing at least three AI contributions and three human contributions.

Show sample map
Table 5. Complementarity map for weekly status report
AI contributions Human contributions
Aggregate data from multiple sources Define report purpose and audience
Draft narrative summary Interpret project context and risks
Flag potential risks based on patterns Make decisions on risk prioritisation and response
Format report consistently Add contextual insights and relationship considerations
Assume accountability for the final report
📝 Exercise 6.1.5: Distinguish Assistance from Substitution
  • Learning outcome: Distinguish assistance vs. substitution
  • Estimated time: 12 min

Instructions: Classify each case as appropriate assistance, possible overreliance, or inappropriate substitution. Justify your answer.

  1. A student uses AI to generate a first draft of an essay, then revises it extensively, adding original analysis and citations.
  2. A manager uses AI to generate performance-review comments for all team members and submits them without changes.
  3. A researcher uses AI to translate a source from French to English, then verifies the translation against the original.
  4. A clinician uses AI to generate a treatment plan and reviews it with the patient before finalising.
Show sample answers
  1. Appropriate assistance — the student revises and adds original work; learning is preserved.
  2. Inappropriate substitution — the manager submits AI-generated reviews without review, replacing professional judgment and personal feedback.
  3. Appropriate assistance — translation is a bounded task, and verification preserves accuracy and accountability.
  4. Appropriate assistance — the clinician reviews and discusses with the patient; professional judgment and patient relationship are preserved.

15. Homework Problems

Formative · Ungraded · Recommended for portfolio preparation and Assignment 1 evidence.

🧠 Problem 6.1.1: AI Role Classification

A data analyst uses an AI system to clean a dataset: it identifies missing values, suggests imputation methods, and formats the data for analysis. The analyst reviews and approves each step.

Classify the primary AI role. Justify your answer.

Show solution design

Answer: Assistant (with some Tool characteristics).

Justification: The system supports the analyst across multiple bounded tasks (identification, suggestion, formatting) while the analyst directs and reviews. It is not making decisions or acting independently. If it only formatted data, it would be a Tool.

🧠 Problem 6.1.2: Mixed-Role Identification

A legal researcher uses an AI system to: (1) summarise case law, (2) highlight contradictions between cases, and (3) suggest which cases are most relevant to the current legal question. The researcher verifies all sources and makes the final decision on which cases to cite.

Which roles is the system playing? Provide a rationale.

Show solution design

Answer: Assistant (summary), Critic (highlighting contradictions), and Recommender (suggesting relevance).

Rationale: The system is playing a mixed role across the workflow. Each function corresponds to a different role. This is common and requires careful boundary-setting for each stage.

🧠 Problem 6.1.3: Role Drift Detection

An AI system is deployed as a Tool to reformat text. Over time, users start asking it to "suggest improvements" and "generate alternative versions." The system begins to make stylistic recommendations without being asked.

What type of issue is this, and what risks does it introduce?

Show solution design

Answer: This is role drift — the system has expanded beyond its intended Tool role into Assistant territory.

Risks: Users may begin to treat the system's recommendations as authoritative; accountability becomes blurred; the system may make unverified changes; permissions may expand without review.

Mitigation: Redefine the role, reset boundaries, communicate limitations, and ensure human review of all recommendations.

🧠 Problem 6.1.4: Role-Boundary Specification

A teacher uses an AI Assistant to generate quiz questions based on supplied learning objectives. The teacher reviews and selects questions for the final quiz.

Specify: (a) permitted activity, (b) prohibited activity, (c) required review, (d) required permission, (e) one stop condition.

Show solution design
  • Permitted: Generate draft quiz questions based on supplied objectives and criteria.
  • Prohibited: Select final questions; determine grading; assess student performance.
  • Required review: Teacher reviews every question for accuracy, clarity, and alignment.
  • Required permission: Teacher approves the final quiz.
  • Stop condition: If the Assistant generates questions that consistently misalign with objectives, stop and revise the prompts or escalate to the department head.
🧠 Problem 6.1.5: Complementarity Mapping

A journalist uses an AI system to transcribe interviews, identify key quotes, and generate a draft article. The journalist reviews, verifies, and revises the draft before publication.

Create a complementarity map showing at least three AI contributions and three human contributions.

Show solution design
Table 6. Complementarity map for journalism task
AI contributions Human contributions
Transcribe audio to text Conduct interviews and ask questions
Identify and extract key quotes Verify quotes against original audio
Generate a draft narrative structure Add context, interpretation, and analysis
Suggest headline alternatives Assume accountability for accuracy and fairness
🧠 Problem 6.1.6: Assistance vs. Substitution — Scenario

A small business owner uses an AI system to: (a) generate ideas for social-media posts, (b) write the posts, and (c) schedule and publish them automatically. The owner reviews the ideas but publishes the posts without further review.

Is this assistance or substitution? Justify your answer.

Show solution design

Answer: This is substitution for the writing and publication steps.

Justification: While idea generation is assistance, the owner has delegated writing and publication to the AI without review. This substitutes the owner's editorial judgment and accountability. The business owner remains responsible for the content but has ceded control over its production and release.

🧠 Problem 6.1.7: Assistance vs. Substitution — Classification

Classify each as assistance or substitution:

  1. Using AI to generate a list of potential research questions, then selecting and refining the best ones.
  2. Using AI to write a full research proposal and submitting it as your own.
  3. Using AI to translate a document, then verifying the translation with a fluent speaker.
  4. Using AI to generate patient-education materials and distributing them without clinical review.
Show solution design
  1. Assistance — human selects and refines.
  2. Substitution — human submits without revision or attribution.
  3. Assistance — human verifies and assumes accountability.
  4. Substitution — human distributes without clinical review, replacing professional judgment.
🧠 Problem 6.1.8: Unsuitable Delegation

A manager delegates to an AI system the task of responding to employee grievances. The AI generates responses based on a policy database, and the manager approves them without reading the full grievance file.

What makes this delegation unsuitable? Identify at least three concerns.

Show solution design
  • Loss of human judgment: Grievances often involve nuanced interpersonal and emotional factors that AI cannot interpret.
  • Inadequate review: The manager approves without accessing the full evidence, making oversight nominal.
  • Relationship erosion: Employees may feel unheard when responses are generated by AI, damaging trust.
  • Accountability gap: If a response is inappropriate, it is unclear who is responsible—the manager, the AI, or the policy database.
🧠 Problem 6.1.9: Agent Permission Analysis

A project manager wants to use an AI Agent to: (a) retrieve data from a project database, (b) generate a status report, (c) email the report to stakeholders, and (d) update the project schedule based on the report.

Which of these actions would you permit without explicit human approval, and which would you require? Justify.

Show solution design

Permitted (with human review at each stage): (a) retrieve data, (b) generate report draft.

Require explicit human approval: (c) emailing stakeholders (external communication), (d) updating the schedule (record modification).

Justification: External communication and record modification have direct consequences. The human must review and authorise these actions to maintain accountability and prevent unauthorised changes.

🧠 Problem 6.1.10: Transfer Problem — Applying to a New Context

You are an instructional designer developing a new online course. You are considering using an AI system to generate draft discussion prompts and assignment instructions.

Apply the concepts from this lesson: What role would the AI play? What boundaries would you set? What would be assistance vs. substitution in this context?

Show solution design

AI role: Assistant — generates draft prompts and instructions for the instructional designer to review and refine.

Boundaries:

  • Permitted: Generate drafts based on learning objectives and course material.
  • Prohibited: Determine learning outcomes, select readings, or make final decisions on assignment weight or grading.
  • Required review: Every prompt and instruction must be reviewed by the instructional designer for pedagogical quality and alignment.

Assistance vs. substitution: This is assistance if the designer reviews, revises, and assumes accountability. It becomes substitution if the AI-generated materials are used without review or if the designer cedes pedagogical judgment to the AI.

16. Mini-Project: AI Role and Complementarity Profile

📁 Project: AI Role and Complementarity Profile
  • Purpose: Develop a detailed profile of one bounded AI-supported activity, classifying the AI role, mapping complementarity, defining boundaries, and assessing assistance vs. substitution.
  • Estimated effort: 75–90 minutes.
  • Deliverable: A structured document (approx. 500–800 words) addressing 15 items.

Instructions

  1. Select one bounded AI-supported activity from your learning, professional, or personal life (e.g., drafting emails, summarising articles, generating ideas, formatting documents).
  2. Do not use sensitive, confidential, or institutional information.
  3. Use the structure provided.
  4. Complete the self-review checklist before submitting (if required for credit).

Required Sections

  1. Human purpose
  2. AI role
  3. Reasons for role classification
  4. AI contributions (list 3–5)
  5. Human contributions (list 3–5)
  6. Role boundaries — permitted
  7. Role boundaries — prohibited
  8. Required information
  9. Review requirement
  10. Permission requirement
  11. Assistance vs. substitution assessment
  12. Capability-preservation concern
  13. Relationship concern
  14. Stop condition
  15. Preliminary recommendation (Proceed / Proceed with safeguards / Redesign / Do not proceed)

Self-Review Checklist

  • Have I clearly stated the human purpose?
  • Is the AI role classification justified?
  • Have I identified at least three AI contributions and three human contributions?
  • Are the boundaries (permitted / prohibited) specific and actionable?
  • Have I assessed whether the use is assistance or substitution?
  • Have I identified at least one capability-preservation and one relationship concern?
  • Is the stop condition concrete?
Show Hint 1: Selecting a use case

Choose something you actually do or plan to do with AI. It can be as simple as "using AI to generate topic ideas for a blog post" or "using AI to summarise meeting notes." The key is that it is bounded—not a whole project, but a specific task.

Show Hint 2: Role classification

Refer to the definitions in section 6. Ask: What is the AI actually doing? Is it generating, evaluating, simulating, recommending, or acting across steps? Most everyday tasks fall under Assistant or Tool.

Show Hint 3: Boundaries

Think about what the AI should not do. What would cross the line? For example, if you use AI to draft an email, the AI should not send it, choose the recipient, or decide the final content.

Show illustrative partial example (different use case)

Use case: Using AI to generate recipe ideas based on ingredients I have at home.

  • Human purpose: To generate 3–5 recipe ideas using available ingredients, reducing food waste and saving time on meal planning.
  • AI role: Assistant — the AI generates ideas and instructions; I choose, adapt, and cook.
  • AI contributions: Generate recipes, suggest substitutions, estimate cooking times.
  • Human contributions: Select ingredients, judge taste preferences, adapt instructions, cook, assume accountability.
  • Boundaries — permitted: Generate recipes, suggest alternatives. Prohibited: Buy ingredients, decide final meal, cook.
  • Assistance vs. substitution: Assistance — I remain in control of selection, adaptation, and execution.

Your project will be more detailed and specific to your chosen context.

17. Key Takeaways

Remember these points
  • Six AI roles: Tool, Assistant, Critic, Simulator, Recommender, Agent. Each has distinct definitions, capabilities, risks, and boundary requirements.
  • Role matters: Classification affects permissions, risk assessment, review intensity, and accountability. Role confusion leads to overtrust and blurred responsibility.
  • Complementarity: Effective collaboration combines AI strengths (speed, pattern recognition, generation) with human strengths (purpose, judgment, relationships, accountability).
  • Assistance ≠ Substitution: Assistance supports human capability; substitution replaces it. Problematic substitution occurs when AI displaces learning, judgment, relationships, or accountability.
  • Boundaries are essential: Every role must have clear permitted activities, prohibited activities, review requirements, and stop conditions. Role drift must be actively monitored.
  • No authority: An assigned AI role never confers professional, legal, moral, or institutional authority. The human remains accountable.
  • PARTNER Purpose: Start every collaboration by defining the legitimate human purpose, the intended outcome, what must improve, and what must not be lost.

18. Key Terms

Tool
An AI role that performs a bounded function under direct human control.
Assistant
An AI role that supports a person across one or more tasks while the person directs the work.
Critic
An AI role that evaluates material against supplied criteria and identifies weaknesses or gaps.
Simulator
An AI role that creates hypothetical interactions, scenarios, or environments for exploration or practice.
Recommender
An AI role that compares alternatives or proposes a course of action based on supplied criteria.
Agent
An AI role that pursues a multi-step goal and may use tools, retrieve information, or perform external actions.
Complementarity
The deliberate combination of different strengths and responsibilities between humans and AI systems.
Assistance
AI supporting a human capability without replacing the human activity that provides purpose, learning, judgment, relationship, or accountability.
Substitution
AI replacing an activity that should remain human.
Role drift
The gradual expansion of an AI system beyond its intended role.
PARTNER
A framework for human-AI responsibility and collaboration, beginning with Purpose.

19. Looking Ahead: Lesson 6.2

You have now established the conceptual foundation for human-AI collaboration. In Lesson 6.2: Task Allocation and Human Authority, you will learn to decompose workflows, allocate tasks among humans and AI, and map different types of authority. You will also extend your PARTNER analysis with Allocate Tasks and Retain Human Authority.

Before moving on, ensure you have completed the self-check quiz, exercises, and mini-project. These are essential preparation for Assignment 2 and the remaining Unit 6 lessons.

Continue to Lesson 6.2: Task Allocation and Human Authority Return to Unit 6 Home

20. Sources & Further Reading

Note: All page numbers, specific definitions, and direct quotations from the textbook are to be confirmed from the supplied course materials. [PAGE NUMBER TO BE CONFIRMED]

⚠️ Note

Some textbook content, definitions, examples, and figures are marked with placeholders. These will be supplied by the course coordinator or subject-matter expert.

👤