COMP 3270 / Module 1 / Unit 1 / Lesson 3

Lesson 1.3 · Week 1 · Reading: Chapter 1

The AI Paradox and ORBIT Analysis

Integrate the AI Paradox as the course's organising concept, apply the ORBIT framework, distinguish efficiency from effectiveness, and define personal AI boundaries.

Suggested time: 3–4 hours Concludes Unit 1 Prepares for Units 2 & 3

Table of Contents

1. Lesson Overview

In Lesson 1.1, you explored what generative AI is and where it appears in daily life. In Lesson 1.2, you examined capabilities, opportunities, and risks, and learned why technical capability does not guarantee contextual suitability.

Now, in Lesson 1.3, you will integrate these ideas into the AI Paradox — the central organising concept of this course. The AI Paradox captures the essential tension: every significant generative AI opportunity may be accompanied by a corresponding risk. You will learn to analyse this relationship systematically using the ORBIT framework (Opportunities, Risks, Beneficiaries and Burden Bearers, Human Intervention and Oversight, Transparency and Accountability).

You will also distinguish efficiency from effectiveness, and begin to define your own personal AI boundaries — uses you consider appropriate, those that require caution, and situations where human judgment must remain central.

This lesson concludes Unit 1 and prepares you for Unit 2 (how generative AI works technically) and Unit 3 (how to make responsible AI decisions).

In this lesson you will:
  • Define the AI Paradox and explain why opportunities and risks must be examined together.
  • Apply the ORBIT framework to a bounded, low‑risk AI use.
  • Distinguish efficiency from effectiveness in AI workflows.
  • Identify beneficiaries and burden bearers in an AI‑supported activity.
  • Articulate personal AI boundaries and stop conditions.
  • Distinguish questions answered in Unit 1 from those reserved for Units 2 and 3.

2. Lesson Learning Outcomes

After completing this lesson, you will be able to:

  1. Explain the AI Paradox in your own words.
  2. Apply the AI Paradox to a bounded, low‑risk use case.
  3. Apply the introductory ORBIT framework to a low‑risk AI use.
  4. Identify beneficiaries and possible burden bearers in an AI‑supported activity.
  5. Distinguish efficiency from effectiveness in AI workflows.
  6. Identify preliminary requirements for human intervention, oversight, transparency, and accountability.
  7. Document your own initial experience, expectations, concerns, and responsible‑use boundaries.
  8. Distinguish questions answered in Unit 1 from those reserved for Unit 2 and Unit 3.

3. Before You Begin

Responsible‑use reminder: Do not enter personal, confidential, proprietary, privileged, or security‑sensitive information into any public or unapproved generative AI system. Use only synthetic, public, or generalised information for all activities.

4. Opening Scenario

Scenario: A convenient but uncertain AI assistant

Jamal is a graduate student who has just started using a new AI‑powered study assistant. The assistant promises to provide personalised explanations, generate practice questions, and offer study tips based on his course materials. Jamal is excited about the time savings and the potential to improve his understanding of difficult concepts.

He inputs his lecture notes and the AI generates a set of practice questions with answers. The questions seem relevant and the answers are clear. However, Jamal notices that some of the explanations are more detailed than his notes, and he is not sure where the AI got that information. He also worries about relying too heavily on the AI: if it makes a mistake, he might not catch it because he is still learning the material.

Jamal wonders:

  • Is it appropriate to use AI‑generated questions to prepare for a high‑stakes exam?
  • How can he be sure the AI's answers are correct?
  • If the AI suggests a study plan that is not effective, who is responsible?
  • Is he becoming too dependent on the AI, and will that affect his own learning?

Questions to consider:

  1. What opportunities does Jamal see in using the AI assistant?
  2. What risks or concerns does he have?
  3. How does the AI Paradox apply here?
  4. What human oversight and accountability are needed?
Show guided analysis of the scenario

Opportunities: Time savings, personalised support, potential to identify gaps in understanding, increased confidence through practice.

Risks: Inaccurate or misleading explanations, over‑reliance, reduced critical thinking, privacy concerns (if inputting course materials), potential for academic misconduct if answers are used improperly.

AI Paradox: The same capability (personalised explanation and question generation) creates an opportunity for enhanced learning but also carries risks of inaccuracy, dependence, and privacy exposure.

Human oversight: Jamal must verify the AI's answers against his course materials, question the AI's assumptions, and ensure he is still engaging with the material himself. He should not treat the AI's outputs as authoritative without verification.

Accountability: Jamal is responsible for his own learning and for the integrity of his exam preparation. The AI is a tool, not a substitute for his own judgment.

5. The AI Paradox

The AI Paradox is the central organising concept of COMP 327. It can be stated simply:

The AI Paradox

Every significant generative AI opportunity may be accompanied by a corresponding risk.

This is not a claim that every AI use is harmful, or that all opportunities and risks are equal, or that risks are inevitable. Rather, it is a recognition that the very capabilities that make generative AI powerful also create potential for error, harm, misuse, or unintended consequences.

The AI Paradox is context‑dependent. Whether an opportunity is realised or a risk materialises depends on many factors:

To understand the AI Paradox, it is helpful to examine specific opportunity‑risk pairs. The table below illustrates eight common pairs, each with a capability, an opportunity, a risk, stakeholders, a contextual factor, a safeguard, and a question for later study.

Table 1. AI Paradox opportunity‑risk pairs
Capability Opportunity Risk Stakeholders Contextual factor Possible safeguard Question for later study
Rapid generation Faster preliminary drafting Faster production and distribution of misinformation Writer, audience Whether content is reviewed before publication Human verification before sharing How can we detect AI‑generated misinformation?
Personalised explanation Support adapted to learner's needs Dependence, privacy exposure, inaccurate adaptation Learner, educator Accuracy and relevance of the explanation Cross‑check with authoritative sources How do models personalise content?
Summarisation Faster orientation to lengthy material Omitted context, conditions, disagreement, or uncertainty Reader, decision‑maker Complexity of the original material Compare summary to original; use multiple summaries What factors affect summary quality?
Creative generation Broader ideation and experimentation Obscured authorship, imitation, loss of personal voice Creator, audience, society Intended use and attribution Clear disclosure of AI use; retain human creative input How does AI influence creativity?
Translation Expanded language access Altered meaning, terminology, tone, cultural context Translator, audience Domain specificity (e.g., legal vs. general) Human translator review; back‑translation How do models handle low‑resource languages?
Workplace assistance Reduced repetitive drafting Hidden verification work, overreliance, skill erosion Employee, employer Criticality of the task Balanced use; preserve human skills What is the impact on professional skills?
Accessibility support Alternative formats and explanations Inaccurate or unsuitable accommodation Person with disability, support provider Specific needs and context User testing and feedback; involve experts How can AI be made more inclusive?
Personalisation Greater relevance Privacy intrusion, manipulation, narrowed choices User, platform, society Amount and sensitivity of personal data Data minimisation; user control How does personalisation affect user autonomy?
Key insight: The AI Paradox is not a reason to avoid AI, but a reason to proceed with awareness, caution, and safeguards. The responsible response to an opportunity may be to explore it further, add safeguards, verify outputs, seek accountable human review, use a non‑AI alternative, postpone action, narrow the task, or decline the use.
Self‑check: AI Paradox

Question: Which of the following statements best reflects the AI Paradox?

  • A. AI is always harmful and should be avoided.
  • B. Every AI capability creates an opportunity and a corresponding risk that must be considered together.
  • C. Risks are inevitable and cannot be mitigated.
  • D. Opportunities always outweigh risks.

Answer: B. The AI Paradox emphasises that opportunities and risks are linked and must be examined together. It does not claim that AI is always harmful (A), that risks are unavoidable (C), or that opportunities always dominate (D).

6. The ORBIT Framework

The ORBIT framework is an introductory tool for analysing AI use cases. It helps you systematically consider the key dimensions of responsible AI use. ORBIT stands for:

ORBIT Framework
O — Opportunities
The legitimate benefits an AI‑supported activity may create.
Guiding questions: What positive outcomes are possible? Who stands to gain?
R — Risks
The possible errors, harms, costs, losses, or unintended consequences associated with the use.
Guiding questions: What could go wrong? What are the negative impacts?
B — Beneficiaries and Burden Bearers
The people or groups receiving the benefits and those performing additional work or experiencing risks and consequences.
Guiding questions: Who benefits? Who bears the burden of verification, risk, or consequences?
I — Human Intervention and Oversight
The points at which people must define the purpose, check evidence, make decisions, handle exceptions, or stop the activity.
Guiding questions: Where and how do humans intervene? Who monitors the process?
T — Transparency and Accountability
The information that should be disclosed and the person or institution responsible for the final result, correction, and remedy.
Guiding questions: What needs to be disclosed? Who takes responsibility?

Low‑risk example: Using AI to generate study questions from your own notes.

Common analytical errors:

Quick ORBIT practice

Think of a low‑risk AI use you have considered (e.g., using AI to draft a personal email, generate a recipe, or summarise a public article). Try to map it using the ORBIT elements. Write down your analysis.

Show a sample ORBIT analysis for a different use

Use: Using AI to generate a weekly meal plan.

  • Opportunities: Saves time, provides variety, reduces decision fatigue.
  • Risks: Nutritional inaccuracies, food allergies not considered, unsustainable suggestions.
  • Beneficiaries: You and your family.
  • Burden bearers: You (must verify nutrition, adjust for preferences).
  • Human intervention: You review the plan, substitute items, check against dietary needs.
  • Oversight: You monitor health outcomes and satisfaction.
  • Transparency: You know AI was used; no disclosure needed.
  • Accountability: You are responsible for your meal choices.
Self‑check: ORBIT

Question: In ORBIT, what does the "I" stand for?

  • A. Innovation
  • B. Integrity
  • C. Human Intervention and Oversight
  • D. Impact

Answer: C. "I" stands for Human Intervention and Oversight, emphasising the need for human involvement and monitoring.

7. Efficiency vs. Effectiveness

A common misconception is that faster and more convenient always means better. In the context of AI, it is essential to distinguish between efficiency and effectiveness.

Key distinction
  • Efficiency concerns the speed, volume, convenience, and reduced initial effort of a process. It asks: "How quickly and easily can we produce something?"
  • Effectiveness concerns whether the activity achieves its legitimate purpose while maintaining adequate quality, evidence, appropriateness, human understanding, fairness, privacy, and accountability. It asks: "Does this actually work well and produce good outcomes?"

An AI‑generated output may be efficient (produced quickly) but ineffective if it requires extensive correction, contains errors that cause harm, or undermines human understanding. Conversely, a slower process may be effective because it ensures quality and accountability.

Example: Two workflows

Workflow A (AI‑first): Generate a report using AI (5 minutes), then spend 2 hours verifying and correcting errors. Total time: 2 hours 5 minutes. Final quality: acceptable after correction, but the user may not fully understand the content.

Workflow B (Human‑governed): Write the report from scratch (2 hours), with no verification time needed (assuming accuracy). Total time: 2 hours. Final quality: high, with full understanding.

Comparison: Workflow A was efficient in generating a draft quickly but was not necessarily effective because it required significant correction and may have reduced the user's understanding. Workflow B was less efficient in initial generation but more effective overall.

Key lesson: Efficiency does not guarantee effectiveness. The best workflow depends on the context, the stakes, and the need for human understanding and accountability.

When evaluating an AI use, consider not only the speed of generation but also the time and effort required for verification, correction, and ensuring human understanding. A fast output that is wrong or that you do not understand may not create genuine net benefit.

Self‑check: Efficiency vs. effectiveness

Question: Which of the following best describes an effective use of AI?

  • A. The AI generates a draft in 30 seconds.
  • B. The output is accurate, appropriate, and the user understands and verifies it.
  • C. The output is produced quickly with minimal effort.
  • D. The output is lengthy and detailed.

Answer: B. Effectiveness is about achieving the purpose with quality, understanding, and accountability, not just speed or volume.

8. Personal AI Boundaries

As you continue to use and engage with generative AI, it is important to define your own personal boundaries — the uses you consider appropriate, those requiring caution, and those you would avoid.

Reflecting on your boundaries

Consider the following questions:

  • What uses of generative AI do you currently consider appropriate for your own learning, work, or personal life?
  • What uses require caution — perhaps because of accuracy concerns, privacy risks, or potential dependence?
  • What information would you never enter into a public or unapproved AI system?
  • What claims or outputs would you always verify before acting on them?
  • What situations require qualified human assistance (e.g., health, legal, financial) rather than AI?
  • Is there any activity you want to perform without AI — for example, to preserve a skill or maintain personal voice?
  • What would be a stop condition — a point at which you would stop using the AI system or reject its output?

Your boundaries may evolve as you learn more. The key is to be intentional and reflective rather than passive or over‑reliant.

Important: Some boundaries are non‑negotiable: never enter personal, confidential, proprietary, or security‑sensitive information into public or unapproved AI systems. Never assign moral, legal, professional, academic, or institutional responsibility to AI.
Self‑check: Boundaries

Question: Which of the following is a sensible personal boundary when using generative AI?

  • A. Entering your social insurance number to personalise a financial plan.
  • B. Using AI‑generated medical advice without consulting a doctor.
  • C. Verifying all AI‑generated information before using it in a work project.
  • D. Assuming all AI outputs are accurate because they are generated by a sophisticated system.

Answer: C. Verifying AI‑generated information is a responsible boundary. The other options describe unsafe or irresponsible practices.

9. Worked Example: AI‑Generated Meeting Minutes

This worked example demonstrates a complete analysis of a low‑risk AI use case using the AI Paradox and the ORBIT framework.

Scenario: A project manager uses a generative AI tool to transcribe and summarise team meeting recordings. The AI generates a concise summary with action items.

Table 2. Analysis using the AI Paradox and ORBIT
Element Analysis
AI Paradox
  • Opportunity: Saves time on note‑taking; ensures consistent documentation; helps team members catch up quickly.
  • Risk: May omit important details, misinterpret discussion points, or generate incorrect action items; may reduce team members' engagement with meeting content.
ORBIT
  • O (Opportunities): Time savings, improved documentation, better follow‑up.
  • R (Risks): Inaccurate summary, missing action items, privacy of meeting content.
  • B (Beneficiaries & Burden Bearers): Team members (benefit from summary); project manager (must verify and correct summary).
  • I (Human Intervention): The project manager reviews the summary, adds context, and confirms action items with the team.
  • T (Transparency & Accountability): The team is informed that AI was used; the project manager is accountable for the final minutes.
Efficiency vs. Effectiveness The AI generates minutes quickly (efficient). However, if the summary requires significant correction or misses key points, the overall process may be less effective than having a human take minutes. The project manager must assess whether the time saved outweighs the verification effort.
Personal boundary The project manager should ensure that no confidential information is entered into the AI tool if it is not approved. The summary should not be used as the sole record; the original recording should be preserved.
Preliminary conclusion Using AI to generate a draft of meeting minutes can be efficient, but human review and verification are essential. The project manager remains accountable for the accuracy and completeness of the final minutes.
Key takeaway: The AI Paradox and ORBIT framework help ensure that the benefits of AI are realised while risks are identified and managed.
Show reflection questions for this worked example
  • What if the meeting involved sensitive strategic decisions? How would that change the risk assessment?
  • How could the project manager verify the accuracy of the AI‑generated summary?
  • What if the team members start to rely on the AI summary instead of attending meetings? Is that a risk?

10. Guided Learning Activities

Activity 10.1: AI Paradox Analysis
  • Purpose: To practise identifying capabilities, opportunities, risks, stakeholders, contextual factors, safeguards, and residual uncertainties.
  • Learning outcome: Apply the AI Paradox to a bounded, low‑risk use case.
  • Estimated time: 20–25 minutes.

Instructions: Consider the following use case: Using generative AI to draft a social media post for a personal blog.

Complete the following analysis:

  • Capability: What capability is used?
  • Opportunity: What is the potential benefit?
  • Corresponding risk: What could go wrong?
  • Factors affecting the opportunity: What makes the opportunity more or less likely?
  • Factors affecting the risk: What makes the risk more or less likely?
  • Stakeholder effects: Who is affected and how?
  • Preliminary safeguard: What safeguard would you propose?
  • Residual uncertainty: What questions remain?
Show sample analysis
  • Capability: Rapid generation of text (social media post).
  • Opportunity: Saves time, provides a starting point, helps overcome writer's block.
  • Risk: The post may be inauthentic, contain inaccurate information, or not match the author's voice; may lead to over‑reliance.
  • Factors affecting opportunity: The quality of the prompt, the author's editing skills, the audience's expectations.
  • Factors affecting risk: The importance of authenticity, the accuracy of the content, the potential for embarrassment.
  • Stakeholder effects: The author (benefits from time savings, but may feel inauthentic); readers (may be misled or notice a lack of personal voice).
  • Safeguard: The author should edit the post to add personal touches and verify any factual claims before publishing.
  • Residual uncertainty: Will the AI‑generated content sound "like me"? How much editing is needed?
Activity 10.2: ORBIT Mini‑Analysis
  • Purpose: To apply all five ORBIT elements to a bounded, low‑risk use.
  • Learning outcome: Apply the introductory ORBIT framework to a low‑risk AI use.
  • Estimated time: 20–25 minutes.

Instructions: Choose a low‑risk use case (e.g., generating a recipe, summarising a public article, or drafting a personal email). Complete an ORBIT analysis:

  • Opportunities: List at least two.
  • Risks: List at least two.
  • Beneficiaries and Burden Bearers: Identify who benefits and who bears burdens.
  • Human Intervention and Oversight: Describe where and how humans should intervene.
  • Transparency and Accountability: State what should be disclosed and who is accountable.
Show sample ORBIT analysis for "generating a recipe"
  • Opportunities: Quick meal ideas, variety, inspiration.
  • Risks: Inaccurate measurements, missing dietary restrictions, food safety issues (e.g., undercooking times).
  • Beneficiaries: The cook (time savings).
  • Burden bearers: The cook (must verify ingredients, check allergies, adjust to taste).
  • Human intervention: The cook reviews the recipe, checks ingredient availability, adjusts for personal preferences and dietary needs.
  • Oversight: The cook monitors the cooking process and outcome.
  • Transparency: No disclosure needed; but if sharing, may mention AI was used.
  • Accountability: The cook is responsible for the final meal.
Activity 10.3: Efficiency vs. Effectiveness
  • Purpose: To compare two workflows and assess overall suitability.
  • Learning outcome: Distinguish efficiency from effectiveness in AI workflows.
  • Estimated time: 20–25 minutes.

Instructions: Consider the following two workflows for writing a one‑page project update:

  • Workflow A (AI‑first): Use an AI to generate a first draft (2 minutes), then spend 15 minutes reviewing, correcting, and adding details. Total time: 17 minutes.
  • Workflow B (Human‑governed): Write the update from scratch (20 minutes), with no verification time needed. Total time: 20 minutes.

Compare the two workflows in terms of:

  • Generation time
  • Verification/correction time
  • Final quality
  • Human understanding
  • Stakeholder effect (e.g., readers)
  • Overall suitability

Which workflow is more efficient? Which is more effective? Justify your answer.

Show sample comparison
  • Generation time: A is faster (2 vs. 20 min).
  • Verification/correction time: A requires 15 min verification, B none.
  • Total time: A is slightly faster (17 vs. 20 min).
  • Final quality: Both may be high, but A's quality depends on the user's editing skills and understanding of the material.
  • Human understanding: In B, the writer has a deeper understanding because they created it from scratch. In A, the writer may not fully understand the content.
  • Stakeholder effect: Readers of A may receive a well‑edited document, but if the writer missed subtle errors, readers could be misled.
  • Overall suitability: If the update is routine and the writer is competent, A may be suitable. However, if accuracy and understanding are critical, B may be more effective despite slightly longer time.

Conclusion: A is more efficient in terms of generation speed, but B may be more effective if understanding and control are important. The best choice depends on context and stakes.

Activity 10.4: Initial Personal AI Boundaries
  • Purpose: To articulate your own boundaries for using generative AI.
  • Learning outcome: Document your own initial experience, expectations, concerns, and responsible‑use boundaries.
  • Estimated time: 15–20 minutes.

Instructions: Based on what you have learned so far, write a brief personal statement (150–250 words) that identifies:

  • Uses you currently consider appropriate.
  • Uses requiring caution.
  • Information you will not enter into an unapproved system.
  • Claims you would verify.
  • Situations requiring qualified human assistance.
  • One activity you want to perform without AI.
  • One preliminary stop condition.

Your statement is personal and not graded.

Show a reflection framework (not a sample answer)

Consider structuring your statement as follows:

"I am comfortable using AI for [tasks] because [reasons]. I will be cautious when using AI for [tasks] because [reasons]. I will never enter [types of information] into a public AI system. I will always verify [types of claims]. For [situations], I will seek a qualified human expert. I prefer to do [activity] without AI to preserve [skill/personal voice]. I will stop using AI if [condition]."

Remember: your boundaries are personal and may evolve.

11. Exercises

These exercises are designed to help you apply the concepts from this lesson. Attempt each exercise before revealing the solutions or hints.

Exercise 1: Explain the AI Paradox Using a New Example
  • Learning outcome: Explain the AI Paradox in your own words.
  • Estimated time: 10–12 minutes.
  • Task: Choose a capability of generative AI that you find interesting (not one already used in the lesson). Write a short paragraph explaining the AI Paradox using that capability. Identify the opportunity, the corresponding risk, and one contextual factor that could affect the outcome.
Show hint

Think of a capability like "code generation," "email drafting," or "data analysis." Then identify a benefit and a potential downside.

Show sample solution

Capability: Code generation — AI that can write code snippets based on natural language descriptions.

Opportunity: Accelerates software development, helps beginners learn syntax, and can suggest efficient algorithms.

Risk: The generated code may contain bugs, security vulnerabilities, or logic errors that are not immediately obvious. Over‑reliance may reduce the developer's own coding skills.

Contextual factor: The importance of security: using AI‑generated code in a banking application carries much higher risk than using it for a personal hobby project.

AI Paradox: The same capability (rapid code generation) can save time but also introduce risks that could be costly or dangerous.

Exercise 2: Apply ORBIT to a Supplied Low‑Risk Use Case
  • Learning outcome: Apply the introductory ORBIT framework to a low‑risk AI use.
  • Estimated time: 15–18 minutes.
  • Task: Apply the ORBIT framework to the following use case: Using generative AI to generate a list of potential research questions for a literature review.

Complete all five ORBIT elements. Provide specific details.

Show hint

Think about the researcher's goal, the potential benefits, the risks of unoriginal or irrelevant questions, who benefits (the researcher) and who bears the burden (the researcher must refine and validate), the need for human judgment in selecting questions, and accountability for the final research direction.

Show sample solution
  • Opportunities: Saves time brainstorming, provides a wide range of ideas, may suggest novel angles.
  • Risks: The questions may be irrelevant, too broad, too narrow, or not grounded in current literature; may lead the researcher down unproductive paths.
  • Beneficiaries: The researcher (time savings, idea generation).
  • Burden bearers: The researcher (must evaluate, refine, and validate each question).
  • Human intervention: The researcher reviews each question for relevance and feasibility, discards unsuitable ones, and rephrases as needed.
  • Oversight: The researcher monitors the quality and progress of the literature review.
  • Transparency: If the research is published, the researcher may disclose that AI was used for brainstorming.
  • Accountability: The researcher is responsible for the final set of research questions and the integrity of the review.
Exercise 3: Distinguish Efficiency from Effectiveness
  • Learning outcome: Distinguish efficiency from effectiveness in AI workflows.
  • Estimated time: 12–15 minutes.
  • Task: Describe a situation where an AI use is efficient but ineffective. Explain why it is not effective and what could be done to improve effectiveness.
Show hint

Think of a scenario where speed is achieved but quality is sacrificed, or where the user does not understand the output, or where the output requires so much correction that the overall process is not beneficial.

Show sample solution

Example: A student uses AI to generate a full essay on a complex topic. The AI produces a 2000‑word essay in 2 minutes (efficient). However, the essay contains factual errors, misinterprets key concepts, and fails to address the assignment prompt. The student must spend 4 hours correcting and rewriting the essay, and even then, the argument is weak because the student did not fully understand the topic.

Why ineffective: The student did not achieve the legitimate purpose (learning and demonstrating understanding). The final quality was poor, and the process did not build the student's knowledge. The time spent on correction could have been used to write a better essay from scratch.

Improvement: The student could use AI to generate an outline or a list of resources, then write the essay themselves, using the AI only for brainstorming and proofreading.

Exercise 4: Identify Human Intervention, Transparency, Accountability, and Stopping Requirements
  • Learning outcome: Identify preliminary requirements for human intervention, oversight, transparency, and accountability.
  • Estimated time: 12–15 minutes.
  • Task: A small business owner is considering using a generative AI system to draft customer emails for promotional campaigns. Identify at least one requirement for each of the following: human intervention, oversight, transparency, accountability, and a stop condition.
Show hint

Think about who reviews the emails, who monitors the campaign's performance, what disclosures are needed (e.g., "AI‑generated"), who takes responsibility for errors, and under what conditions the AI would not be used (e.g., if it produces too many errors).

Show sample solution
  • Human intervention: A marketing team member must review every AI‑generated email before sending, checking for accuracy, tone, and compliance with regulations.
  • Oversight: The marketing manager monitors open rates and conversion rates to assess the effectiveness of the campaign and flag any issues.
  • Transparency: The business should disclose in its privacy policy or footer that some communications are AI‑generated, to be transparent with customers.
  • Accountability: The business owner is ultimately responsible for the content and outcomes of the emails; if a customer complains or a legal issue arises, the owner must address it.
  • Stop condition: If the AI‑generated emails produce a significant number of complaints or errors, the business should stop using the AI for this purpose until the issues are resolved.

12. Homework Problems

Formative · Ungraded · Recommended preparation for Assignment 1

Problem 1: AI Paradox Application
  • Estimated time: 10–12 minutes.
  • Task: Describe the AI Paradox for the capability of "question generation" (AI generating practice questions). Identify the opportunity, the corresponding risk, and one contextual factor that could affect the balance.
Show hint

Think about how AI can generate questions for self‑testing, but they might be off‑topic or poorly worded.

Show sample solution

Opportunity: Efficient self‑testing, exposure to different question types, identification of knowledge gaps.

Risk: Questions may be irrelevant, too easy/hard, or contain errors; over‑reliance may reduce critical thinking.

Contextual factor: The quality of the source material (e.g., notes vs. textbook) and the user's ability to evaluate the questions.

Problem 2: ORBIT Classification
  • Estimated time: 10–12 minutes.
  • Task: For the following use case, identify which ORBIT element is being described: "The user reviews the AI‑generated summary and adds their own analysis before presenting it."
Show hint

This describes the role of humans in the process.

Show sample solution

Answer: Human Intervention and Oversight (I). The user is intervening to review and enhance the output.

Problem 3: Complete ORBIT Analysis
  • Estimated time: 15–18 minutes.
  • Task: Perform a complete ORBIT analysis for the following use: Using generative AI to generate a summary of a public news article for personal knowledge.
Show hint

Consider all five elements: Opportunities, Risks, Beneficiaries and Burden Bearers, Human Intervention and Oversight, Transparency and Accountability.

Show sample solution
  • Opportunities: Quick understanding of articles, saves time, helps identify key points.
  • Risks: Summary may be inaccurate, miss nuance, or reflect bias.
  • Beneficiaries: You (the reader).
  • Burden bearers: You (must verify key claims, compare with other sources).
  • Human intervention: You read the summary, check important facts, and decide if you need to read the full article.
  • Oversight: You monitor your understanding and the usefulness of the summary.
  • Transparency: Not required for personal use.
  • Accountability: You are responsible for your own comprehension and any actions based on the information.
Problem 4: Beneficiary and Burden Bearer
  • Estimated time: 10–12 minutes.
  • Task: In the context of using AI to draft performance reviews for employees, identify one beneficiary and one burden bearer. Explain why each fits that role.
Show hint

Beneficiaries might include managers (time saved) and the organisation (efficiency). Burden bearers might include employees (if reviews are unfair) and managers (who must verify and take responsibility).

Show sample solution
  • Beneficiary: The manager who saves time drafting reviews and can focus on other tasks.
  • Burden bearer: The employee whose review may contain inaccuracies or biases, and the manager who must ensure the review is fair and accurate.
Problem 5: Efficiency vs. Effectiveness
  • Estimated time: 10–12 minutes.
  • Task: A marketing team uses AI to generate 50 different ad copy variations in 5 minutes. However, most variations are not on‑brand and require significant editing. Is this use efficient? Is it effective? Explain.
Show hint

Efficiency is about speed and volume; effectiveness is about quality and achieving the goal.

Show sample solution

Efficiency: Yes, generating 50 variations in 5 minutes is efficient in terms of speed and volume.

Effectiveness: Possibly not, if most variations are not usable and the team must spend a lot of time editing. The final goal (creating effective ads) may not be achieved efficiently overall.

Conclusion: The use is efficient in generation but may be ineffective if the editing time outweighs the benefits.

Problem 6: Human Accountability
  • Estimated time: 10–12 minutes.
  • Task: A university is considering using AI to grade student essays. Who should be accountable for the final grade? What safeguards would you recommend?
Show hint

Consider that grades affect students' academic records. The instructor must take responsibility. Safeguards might include human review of a sample, appeals process, and transparency about AI use.

Show sample solution

Accountability: The instructor of record is ultimately accountable for the grades. The AI can assist but not replace the instructor's judgment.

Safeguards:

  • All AI‑graded essays should be reviewed by a human instructor, especially borderline cases.
  • Students should have the right to appeal and request a human regrade.
  • The AI system should be transparent about its criteria and limitations.
  • Regular audits of AI grading consistency and fairness should be conducted.
Problem 7: Preliminary Boundary
  • Estimated time: 10–12 minutes.
  • Task: A journalist uses AI to draft a story about a breaking news event. What preliminary boundary should the journalist set regarding verification? Why?
Show hint

Think about journalistic standards: accuracy, fact‑checking, and attribution. The boundary might be to always verify all facts before publication.

Show sample solution

Boundary: The journalist should never publish an AI‑generated story without independently verifying all facts, sources, and quotes. The AI draft should be treated as a starting point, not a final product.

Why: Journalism requires accuracy and accountability. AI may generate false information or misrepresent facts. The journalist, not the AI, is accountable for the story's veracity.

Problem 8: Responsible Refusal or Non‑AI Alternative
  • Estimated time: 12–15 minutes.
  • Task: You are a human resources manager considering using AI to screen job applications and generate a shortlist. What might cause you to refuse or postpone this use? What non‑AI alternative could you consider?
Show hint

Consider risks of bias, lack of transparency, and legal compliance. Alternatives might include human screening with structured criteria.

Show sample solution

Reasons to refuse or postpone:

  • The AI system has not been validated for fairness and may introduce bias.
  • The organisation lacks a clear policy on AI use in hiring.
  • Legal risks: hiring decisions must be defensible and transparent.
  • Lack of oversight: who reviews the AI's decisions?

Non‑AI alternative: Use a structured screening process with clear criteria, where trained HR staff review applications. This ensures human judgment and accountability, though it may be less efficient.

13. Mini‑Project: Introductory ORBIT Use‑Case Analysis

Project: Introductory ORBIT Use‑Case Analysis
  • Purpose: To integrate all concepts from Unit 1 by performing a complete ORBIT analysis on a bounded, low‑risk use case.
  • Learning outcomes: Apply the AI Paradox and ORBIT framework; distinguish efficiency from effectiveness; identify beneficiaries and burden bearers; articulate human intervention and accountability.
  • Estimated effort: 60–75 minutes.

Instructions: Select a bounded use from one of the following contexts:

  • Learning (e.g., generating practice questions, explaining concepts)
  • Professional practice (e.g., drafting routine documents, summarising information)
  • Organisational activity (e.g., generating meeting minutes, internal communications)
  • Communication (e.g., drafting emails, social media posts)
  • Creative work (e.g., brainstorming ideas, generating prompts)
  • Public information (e.g., summarising public reports, translating public notices)
  • Everyday life (e.g., meal planning, travel itineraries)

Required deliverable: Complete the following analysis (use the template below).

ORBIT Use‑Case Analysis Template
# Component Your response
1 Use‑case title
2 Human purpose
3 AI capability
4 Opportunity
5 Risk
6 Beneficiaries
7 Burden bearers
8 Human intervention
9 Oversight
10 Transparency
11 Accountability
12 Preliminary information boundary
13 One verification question
14 One stop condition
15 Efficiency vs. effectiveness assessment
16 Preliminary conclusion (choose one: suitable for further exploration; may be suitable with safeguards; requires redesign; requires authoritative review; non‑AI alternative preferable; insufficient information)
Privacy and academic integrity:
  • Use only synthetic, public, or generalised information.
  • Your analysis is for your own learning and may be used as evidence for Assignment 1.
  • Do not copy this analysis from another student or from an AI system.
Show self‑review checklist
  • Have you selected a bounded use case from the allowed contexts?
  • Have you completed all 16 components?
  • Have you identified at least one opportunity and one risk?
  • Have you identified beneficiaries and burden bearers?
  • Have you described human intervention, oversight, transparency, and accountability?
  • Have you stated an information boundary and a stop condition?
  • Have you assessed efficiency vs. effectiveness?
  • Have you selected a preliminary conclusion?
Show solution architecture (staged guidance)
  1. Use‑case title: Give a clear, descriptive name.
  2. Human purpose: State the user's goal.
  3. AI capability: Specify the capability (e.g., summarisation, generation).
  4. Opportunity: Describe the potential benefit.
  5. Risk: Describe a corresponding risk.
  6. Beneficiaries: Who gains? List specific individuals/groups.
  7. Burden bearers: Who performs verification or bears consequences?
  8. Human intervention: What must humans do? When?
  9. Oversight: Who monitors the process and outcomes?
  10. Transparency: What should be disclosed and to whom?
  11. Accountability: Who is responsible for the final output?
  12. Information boundary: What information should not be entered?
  13. Verification question: What specific claim or output must be checked?
  14. Stop condition: Under what conditions would you stop using the AI?
  15. Efficiency vs. effectiveness: Compare speed vs. quality and overall value.
  16. Preliminary conclusion: Choose one of the allowed conclusions.
Show illustrative partial example (different context)

Use‑case: Generating a draft of a routine internal newsletter.

  • Human purpose: To quickly produce a weekly newsletter for staff.
  • AI capability: Rapid generation of text based on key points.
  • Opportunity: Saves time, ensures consistency, reduces workload.
  • Risk: The draft may contain errors, miss important updates, or have an inappropriate tone.
  • Beneficiaries: The communications team and staff who receive timely information.
  • Burden bearers: The editor who must review and correct the draft.
  • Human intervention: The editor reviews and edits the draft before distribution.
  • Oversight: The manager monitors newsletter quality and staff feedback.
  • Transparency: No disclosure needed; internal use.
  • Accountability: The editor and manager are accountable for the newsletter's content.
  • Information boundary: Do not enter confidential employee information.
  • Verification question: Are all dates, names, and project updates correct?
  • Stop condition: Stop if the AI draft requires more than 30 minutes of editing per week.
  • Efficiency vs. effectiveness: The AI is efficient in generating a first draft, but effectiveness depends on the quality of the editing.
  • Preliminary conclusion: May be suitable with safeguards (human review).

14. Self‑Check Quiz

This ungraded quiz is designed to help you check your understanding of the key concepts from this lesson.

Quiz 1 — Multiple choice

What is the AI Paradox?

Select one:
Show answer and explanation

Correct answer: C.

Explanation: The AI Paradox states that every significant AI opportunity may be accompanied by a corresponding risk. It is not a deterministic claim that all AI is harmful or that risks are inevitable.

Quiz 2 — Multiple choice

In the ORBIT framework, what does the "B" stand for?

Select one:
Show answer and explanation

Correct answer: B.

Explanation: B stands for Beneficiaries and Burden Bearers — identifying who gains from the AI use and who bears the burdens of verification, risk, or consequences.

Quiz 3 — Multiple choice

Which of the following is an example of being effective but not efficient?

Select one:
Show answer and explanation

Correct answer: B.

Explanation: Effectiveness is about achieving the purpose with quality and accountability. Taking extra time to verify ensures accuracy (effective) but may not be efficient in terms of speed.

Quiz 4 — Multiple choice

Who is typically a burden bearer in an AI‑supported activity?

Select one:
Show answer and explanation

Correct answer: C.

Explanation: The human user typically bears the burden of verifying the output, correcting errors, and taking responsibility. The AI system and data are not burden bearers in the human sense.

Quiz 5 — Multiple choice

Which of the following is a responsible response to an AI opportunity?

Select one:
Show answer and explanation

Correct answer: C.

Explanation: A responsible approach is to use AI with appropriate safeguards, verification, and accountability. Avoiding AI (D) may be appropriate in some cases but is not the only responsible response.

Quiz 6 — Multiple choice

What is the purpose of defining personal AI boundaries?

Select one:
Show answer and explanation

Correct answer: C.

Explanation: Personal boundaries help you use AI intentionally and reflectively, identifying appropriate uses and areas where caution is needed.

Quiz 7 — Multiple choice

Which of the following is a stop condition?

Select one:
Show answer and explanation

Correct answer: A.

Explanation: A stop condition is a predefined point at which you would stop using the AI system, e.g., if errors exceed a threshold or if the output is not verifiable.

Quiz 8 — True or false

The AI Paradox implies that every AI use is equally risky.

Select one:
Show answer and explanation

Correct answer: False.

Explanation: The AI Paradox does not claim that all uses are equally risky. Risk depends on context, stakes, safeguards, and other factors.

15. Assignment 1 Journal Connection

Unit 1 contributes to Assignment 1: AI Learning and Reflection Journal (15% of your final grade).

Journal prompts for Lesson 1.3

After completing this lesson, consider adding to your journal with entries that address some of the following prompts:

  1. AI Paradox: Describe an AI capability you use or encounter. What is the opportunity, and what is the corresponding risk? How do you manage that risk?
  2. ORBIT analysis: Choose a low‑risk AI use you have considered. Apply ORBIT to it and write a brief analysis.
  3. Efficiency vs. effectiveness: Describe a time when you prioritised efficiency (speed) over effectiveness (quality), perhaps with an AI tool. What was the outcome?
  4. Personal boundaries: Revisit the personal boundaries you identified in the activity. Have they changed after this lesson? If so, how?
  5. Questions for Units 2 and 3: What technical questions do you have about how AI works? What questions do you have about responsible‑use decision‑making?

Your journal entries are personal and reflective. They do not need to be perfect; they are a record of your learning process.

Show journal planning checklist
  • Have you selected at least 2 prompts to address?
  • Have you written about your own experiences and perspectives?
  • Have you maintained privacy boundaries?
  • Have you kept a record of any AI interactions or observations?
  • Have you noted any questions for Units 2 and 3?

16. Key Takeaways

17. Key Terms

AI Paradox
The principle that every significant generative AI opportunity may be accompanied by a corresponding risk.
ORBIT framework
A tool for analysing AI use cases: Opportunities, Risks, Beneficiaries and Burden Bearers, Human Intervention and Oversight, Transparency and Accountability.
Opportunity (ORBIT)
A legitimate benefit that an AI‑supported activity may create.
Risk (ORBIT)
A possible error, harm, cost, loss, or unintended consequence.
Beneficiary
A person or group who receives benefits from an AI‑supported activity.
Burden bearer
A person or group who performs verification work, bears risks, or experiences negative consequences.
Human intervention
The points at which people must define purpose, check evidence, make decisions, handle exceptions, or stop the activity.
Oversight
The monitoring of the AI‑supported activity to ensure it proceeds appropriately and to intervene if needed.
Transparency
Disclosure of relevant information about the AI use, such as how it was used and what data was involved.
Accountability
The responsibility of a person or institution for the final result, correction, and remedy.
Efficiency
The speed, volume, convenience, and reduced initial effort of a process.
Effectiveness
Whether an activity achieves its legitimate purpose while maintaining quality, evidence, appropriateness, human understanding, fairness, privacy, and accountability.
Stop condition
A predefined point at which the use of an AI system should cease, often triggered by an error threshold or a change in context.

[See Appendix A for the complete Generative AI Glossary.]

18. Looking Ahead

You have now completed all three lessons of Unit 1. You have:

Before moving on to Unit 2, please complete the Unit 1 Consolidation Exercise, the Cumulative Homework Set, and the Cumulative Project (Introductory AI Paradox and ORBIT Analysis). These activities will help you solidify your understanding and prepare you for the remaining units.

In Unit 2, you will dive into the technical workings of generative AI: training data, models, tokenisation, next‑token prediction, and more. This will give you the foundational knowledge to understand how systems generate content and why they have the limitations we have discussed.

In Unit 3, you will learn how to make responsible decisions about AI use, including formal suitability assessment, risk classification, and comprehensive verification procedures.

Distinguishing Unit 1, Unit 2, and Unit 3:
  • Unit 1 (completed): Concepts, applications, capabilities, risks, AI Paradox, ORBIT framework.
  • Unit 2: Detailed technical mechanisms — how generative AI works.
  • Unit 3: Detailed responsible‑use decision‑making — how to determine if an AI use is effective and responsible.
Continue to Unit 1 Consolidation Return to Unit 1 Home

19. Sources & Further Reading

Primary source:

Supporting materials:

Additional references (for further exploration):

Note: All references and citations will be verified and updated in the final course materials. Do not treat unverified references as authoritative sources.

👤