Table of Contents
- 1. Lesson Overview
- 2. Learning Outcomes
- 3. Before You Begin
- 4. Retrieval Practice
- 5. Opening Scenario
- 6. Core Lesson Content
- 6.1 False, Unsupported, or Uncertain?
- 6.2 Material Omissions
- 6.3 Misleading Framing
- 6.4 Outdated Information
- 6.5 Contextual Mismatch
- 6.6 How Selective Truth Can Still Mislead
- 6.7 Summary of Error Categories
- 7. TRACE Walkthrough
- 8. Error-Analysis Process Walkthrough
- 9. Worked Example
- 10. Guided Learning Activities
- 11. Exercises
- 12. Homework Problems
- 13. Mini-Project
- 14. Self-Check Quiz
- 15. Assignment Connections
- 16. Reflective Journal Prompts
- 17. Key Takeaways
- 18. Key Terms
- 19. Looking Ahead
- 20. Sources & Further Reading
1. Lesson Overview
In Lesson 10.1, you learned to identify claims, separate atomic from compound claims, and detect hallucinations and fabrications. In this lesson, you will extend that foundation to examine factual errors, unsupported claims, material omissions, misleading framing, outdated information, and contextual mismatches.
You will learn that an AI-generated response can contain statements that are technically true but still misleading—through selective emphasis, missing context, or framing that distorts interpretation. A claim may be factually correct yet materially incomplete, or accurate in one jurisdiction but entirely wrong in another. You will develop the skills to detect these more subtle forms of error and to assess whether a statement is false, unsupported, uncertain, or incomplete.
This lesson establishes the second stage of the TRACE framework: Review the claims (or, as operationalised here, examine claims for factual support, omissions, framing, currency, and contextual fit). By the end of this lesson, you will be able to evaluate not just what an AI system says, but also what it leaves out, how it says it, and whether it fits the situation—all of which are essential for responsible use.
2. Learning Outcomes
By the end of this lesson, you should be able to:
- Distinguish unsupported claims from demonstrably false claims.
- Identify material omissions that affect interpretation or decision-making.
- Identify misleading framing, including selective emphasis and loaded wording.
- Identify outdated information and explain why currency matters.
- Recognise contextual, disciplinary, audience, jurisdictional, temporal, and cultural mismatches.
- Explain how a statement that is technically true can still be materially misleading.
- Apply the second element of the TRACE framework to a synthetic AI-generated passage.
- Prioritise claims for verification based on consequence and context.
3. Before You Begin
Chapter 7: Hallucination, Error, and Uncertainty (sections on factual error, omission, framing, context, and currency). Selected sections from Chapter 2 (How Generative AI Works) and Chapter 3 (Effective and Responsible AI Use) may also be helpful.
You should have completed Lesson 10.1 and be able to:
- Identify and list claims in an AI-generated passage.
- Separate atomic from compound claims.
- Distinguish hallucination and fabrication from other error types.
- Recognise credibility features that do not establish accuracy.
- Explain why fluency does not guarantee reliability.
You may use an institutionally approved generative AI system for exploration, but you are not required to do so. For every activity that suggests using a live AI system, a supplied-output alternative is provided.
Privacy reminder: Do not enter personal, confidential, proprietary, or security-sensitive information into any AI system.
- Do not treat generated citations as verified.
- Do not label a claim false merely because a source was not found immediately.
- Do not publish or act on unverified consequential claims.
- Document your verification process for Assignment 5 preparation.
4. Retrieval Practice
These questions review concepts from Lesson 10.1 that are essential for this lesson.
Question 1: List three types of claims (e.g., factual, interpretive) and provide a brief example of each.
Show answer
Example answer:
- Factual: “Water freezes at 0°C.”
- Interpretive: “The author’s use of metaphor suggests a critique of industrialisation.”
- Predictive: “The unemployment rate will decline next quarter.”
- Normative: “Governments should fund universal basic education.”
Question 2: What are two warning signs that a citation may be fabricated?
Show answer
- An unusual or placeholder DOI prefix (e.g., 10.1234).
- A journal name that does not appear in reputable databases.
- The reference cannot be found in a library catalogue or scholarly database.
- The title and author names seem generic or mismatched.
Question 3: Why does a confident, fluent tone in an AI-generated response not guarantee accuracy?
Show answer
Fluency and confidence are stylistic features that the AI system is optimised to produce. The system is trained to generate coherent, plausible text, not to track truth. A confident tone may reflect the system’s probabilistic confidence in the next token, not the evidential support for the claim. Therefore, confidence is not a substitute for independent verification.
5. Opening Scenario
The following passage was generated by an AI system to inform residents about a new community benefit program. Read it carefully. Some parts are accurate; others are problematic in ways that you will analyse in this lesson.
“The City of Rivermouth is pleased to announce the launch of the ‘Healthy Homes Initiative,’ a new program designed to help residents improve their home energy efficiency. Starting January 1, 2026, eligible homeowners can apply for grants of up to $5,000 to cover the cost of insulation, window replacements, and heat pump installations. The program is funded by a federal climate action grant and is expected to reduce household energy costs by 15–20%. To qualify, applicants must have a household income below $75,000 and must own their home. The program has been widely praised by environmental groups, and a pilot study in 2024 showed that participants saved an average of $800 per year on their utility bills. Applications will be processed on a first-come, first-served basis, and homeowners are encouraged to apply early to secure funding. Additionally, the city is offering free energy audits to all applicants. Similar programs in other jurisdictions have been shown to reduce carbon emissions by up to 30%. This initiative represents a significant step forward in our city’s commitment to sustainability.”
Questions for initial reflection:
- Which statements in this passage are factual claims? Which are interpretations or recommendations?
- What information might be missing that would be important for a resident to know?
- Does the passage present information in a way that might be misleading? If so, how?
- Is the information current? Are there any assumptions about jurisdiction or context that might not apply to all readers?
Show guided analysis (initial)
Initial observations:
- Accurate-sounding statements: The program launch date, grant amount, and eligibility criteria are presented with specificity.
- Potential omissions: The passage does not mention how many grants are available, the total funding cap, the deadline for applications, or whether the $75,000 income limit is based on gross or net income.
- Potential framing issues: The phrase “widely praised by environmental groups” is vague and may be misleading if only one group praised it. The claim that “similar programs ... have been shown to reduce carbon emissions by up to 30%” lacks a citation and may be overstated.
- Contextual assumptions: The passage assumes the reader is a homeowner, lives in Rivermouth, and has an income below $75,000. It also assumes the federal grant is confirmed (it may be “expected” but not finalised).
- Outdated information risk: The passage refers to a 2024 pilot study; if this text is used in 2026 or later, the data may be outdated.
Key takeaway: Even if every specific claim in the passage is true, the passage may still mislead through omission (e.g., not mentioning that funds are limited) or framing (e.g., overstating the benefits). This is the central theme of Lesson 10.2.
6. Core Lesson Content
6.1 False, Unsupported, or Uncertain?
When you evaluate a claim, one of the first questions to ask is: Is this claim false, unsupported, or uncertain? These categories are distinct and require different responses.
- False claim (factual error):
- A claim that conflicts with reliable evidence. It can be definitively shown to be incorrect.
- Unsupported claim:
- A claim that is plausible but lacks sufficient evidence. It may be true or false; you cannot determine without further information.
- Uncertain claim:
- A claim for which the evidence is ambiguous, incomplete, or conflicting. The truth value cannot be determined with confidence.
| Category | Example | Appropriate response |
|---|---|---|
| False (factual error) | “The capital of Canada is Toronto.” | Correct the error with the accurate fact (Ottawa). |
| Unsupported | “84% of students prefer online learning.” (no source) | Do not use without verification. Flag as requiring evidence. |
| Uncertain | “The policy may reduce costs, but the evidence is mixed.” | Communicate the uncertainty clearly; do not overstate. |
Do not assume a claim is false just because it is unsupported. Absence of evidence is not evidence of absence. An unsupported claim may be true, but you should not rely on it until it has been verified. Label it as unsupported and seek evidence.
Classify each of the following claims as false, unsupported, or uncertain.
- “The Eiffel Tower is located in Berlin.”
- “A majority of Canadians support universal basic income.” (no source provided)
- “The impact of the new policy on employment is still being studied, and results are inconclusive.”
- “The study found a correlation between screen time and sleep disruption, but the mechanism is not fully understood.”
Show answers
- False — the Eiffel Tower is in Paris, not Berlin. This is a factual error.
- Unsupported — the claim may be true, but no evidence is provided. It requires verification.
- Uncertain — the evidence is inconclusive; the claim acknowledges uncertainty.
- Uncertain — the correlation is established, but the mechanism is unknown, leaving uncertainty.
6.2 Material Omissions
A material omission occurs when an AI-generated response excludes information that is required for a reasonable interpretation or decision. The omission is “material” if it would affect how a reader understands the content or acts upon it.
An omission is material when the missing information is:
- Essential to understanding — without it, the claim is incomplete or misleading.
- Relevant to decision-making — a reader would make a different choice if they had the information.
- Reasonably expected — a competent reader would expect this information to be included.
Common types of material omissions:
- Conditions and exceptions: “You are eligible for the benefit” (omits that eligibility depends on income level or citizenship).
- Limitations: “The tool is effective” (omits that it is only effective for certain types of tasks).
- Disagreement or alternative views: “Experts agree that X is true” (omits that a minority of experts disagree).
- Uncertainty: “The policy will reduce costs by 20%” (omits that this is a forecast with a wide margin of error).
- Costs or risks: “The program will save $10 million” (omits that it costs $5 million to implement).
- Affected stakeholders: “This policy benefits workers” (omits that it may disadvantage part-time employees).
- Dates and currency: “The data shows a trend” (omits that the data is from 2010).
- Jurisdiction: “This is the legal standard” (omits that it applies only in a specific country).
- Review requirements: “This analysis is complete” (omits that it has not been peer-reviewed).
Read the following claim and identify two material omissions:
“The new software will increase productivity by 25% and is recommended by industry experts.”
Show omissions
- Omission 1: The claim does not specify which industry experts or how many—“recommended by experts” is vague and may be a handful of individuals.
- Omission 2: The claim does not mention the cost of the software, the implementation time, or whether the productivity increase is sustainable.
- Omission 3: The claim does not specify what kind of productivity (e.g., per employee, per team) or the timeframe (e.g., over a year).
6.3 Misleading Framing
Misleading framing occurs when the way information is presented—through word choice, selection, ordering, or emphasis—distorts interpretation, even if each individual fact is technically true.
- Selective emphasis: Highlighting some facts while downplaying others.
- Loaded wording: Using emotionally charged or value-laden language.
- False balance: Presenting two sides as equally credible when one is clearly better supported.
- Omission of alternatives: Presenting only one option or interpretation.
- Unsupported certainty: Presenting a possibility as a certainty.
- Correlation as causation: Implying that because two things are related, one causes the other.
- Presenting one perspective as consensus: Using phrases like “experts agree” when there is significant disagreement.
- Hiding uncertainty: Omitting qualifications, margins of error, or alternative explanations.
- Inappropriate comparison: Comparing apples to oranges to make one option look better.
Consider the same event described in two different ways:
Version A (Neutral)
“The city’s unemployment rate increased from 5.2% to 5.7% between January and December. This represents a rise of 0.5 percentage points, or approximately 1,200 additional people seeking work.”
Version B (Misleading)
“Unemployment in the city has skyrocketed by nearly 10% this year, leaving thousands of families struggling to make ends meet. The local government has failed to address the jobs crisis.”
Analysis: Version B uses loaded words (“skyrocketed,” “crisis,” “failed”), frames the increase as a percentage change (5.7% is about a 10% relative increase from 5.2%), and implies government failure without evidence. Version A presents the same facts more neutrally. Both are “true,” but Version B is misleading.
What framing devices are used in the following statement?
“Despite the overwhelming scientific consensus, a small minority of so-called ‘experts’ continue to cast doubt on climate change, delaying much-needed action.”
Show analysis
- Loaded wording: “so-called ‘experts’” implies the minority are not legitimate experts.
- Selective emphasis: Highlights the minority view to frame them as obstructionist.
- Value judgment: “delaying much-needed action” assigns blame and urgency.
- Unsupported characterisation: “overwhelming scientific consensus” is a claim that requires evidence.
The statement is not neutral—it uses framing to persuade rather than inform.
6.4 Outdated Information
Outdated information is a claim that was once true (or was believed to be true) but is no longer current. This is a common issue in AI-generated content because the training data may not include recent developments.
- Laws, regulations, and policies change frequently.
- Scientific guidance evolves with new research.
- Product features, prices, and availability change.
- Personnel, institutional structures, and contact information become outdated.
- Statistical data (e.g., population, crime rates) changes over time.
Common areas where information becomes outdated:
- Laws and policies: “The minimum wage is $15 per hour” — this may have changed.
- Standards and guidelines: “The WHO recommends X” — guidelines are periodically updated.
- Personnel: “The CEO is John Smith” — the CEO may have changed.
- Prices: “The product costs $99” — prices change.
- Scientific guidance: “Dietary guidelines recommend limiting sugar” — recommendations are revised.
- Software versions: “The software supports feature X” — new versions may add or remove features.
- Schedules and dates: “The conference is in May” — the date may have changed.
- Institutional procedures: “Applications must be submitted by mail” — procedures may have changed.
Why might the following claim be outdated, and what would you need to verify?
“According to the 2020 census, the population of the city is 450,000.”
Show analysis
Why it may be outdated: The census is from 2020. Populations change due to births, deaths, and migration. By the time you read this, the population may be significantly different.
What to verify: Check for more recent population estimates (e.g., from a 2024 or 2025 census or statistical agency report). If no newer data is available, note the date and treat the claim as potentially outdated.
6.5 Contextual Mismatch
A contextual mismatch occurs when a response is inaccurate or unsuitable because it does not fit the relevant context: jurisdiction, date, discipline, organisation, learner level, professional setting, culture, language, accessibility needs, task, or intended audience.
- Jurisdictional mismatch:
- Using laws, regulations, or standards from one jurisdiction in another. Example: Citing US case law in a Canadian legal context.
- Temporal mismatch:
- Using information from a different time period without acknowledging change.
- Disciplinary mismatch:
- Using concepts, methods, or standards from one discipline in another where they do not apply.
- Audience mismatch:
- Using language, examples, or assumptions that are inappropriate for the intended audience (e.g., overly technical for beginners).
- Cultural mismatch:
- Assuming cultural norms, values, or practices that do not apply to the target audience.
- Accessibility mismatch:
- Failing to account for accessibility needs (e.g., using colour-only distinctions, assuming visual ability).
- Organisational mismatch:
- Using procedures, structures, or terminology from one organisation in another where they differ.
- Task mismatch:
- Providing information that does not address the specific task or question at hand.
A generally accurate statement may still be inappropriate in a specific context. For example, recommending a medical treatment that is approved in one country but not in another, or suggesting a teaching strategy that works in one educational system but not in another. Always consider the context before accepting or using AI-generated content.
Identify the type of contextual mismatch in the following scenario:
“An AI system, responding to a query from a high school student, provides an answer using graduate-level statistical terminology and complex mathematical notation, making it inaccessible to the student.”
Show answer
Audience mismatch. The response is not tailored to the high school student’s level of understanding. The AI failed to consider the audience’s prior knowledge and accessibility needs.
Additional concerns: This may also be an accessibility issue if the student has a learning disability or language barrier.
6.6 How Selective Truth Can Still Mislead
One of the most important lessons in this lesson is that a statement can be technically true and still be materially misleading. This can happen through:
- Omission: Leaving out critical context, conditions, or counter-evidence.
- Framing: Presenting facts in a way that suggests a particular conclusion.
- Selective emphasis: Highlighting evidence that supports one side while downplaying evidence that contradicts it.
- Out-of-context quotes: Quoting someone in a way that distorts their original meaning.
- Unsupported certainty: Presenting a tentative finding as a definitive conclusion.
- Inappropriate comparison: Comparing data points that are not comparable.
- False balance: Giving equal weight to two sides when one is overwhelmingly supported.
A statement can be factually correct yet still function as misinformation if it is strategically incomplete or framed to mislead. For example:
- “The study found a correlation between X and Y.” (True, but it omits that the correlation is weak and may be due to confounding variables.)
- “The policy reduced costs by 10%.” (True, but it omits that the policy cost 20% more to implement.)
- “The candidate supports education funding.” (True, but it omits that they propose cutting other programs to pay for it.)
This is why critical analysis must go beyond checking individual facts. You must also examine what is not said, how it is said, and whether it fits the context.
6.7 Summary of Error Categories Covered in This Lesson
| Category | Definition | Key Question |
|---|---|---|
| Factual error | Claim conflicts with reliable evidence. | Is the claim demonstrably false? |
| Unsupported claim | Plausible but lacks sufficient evidence. | What evidence would support this? |
| Material omission | Missing information needed for reasonable interpretation. | What is missing that a reader needs to know? |
| Misleading framing | Wording, selection, or emphasis distorts interpretation. | How is the information being presented, and why? |
| Outdated information | Claim that was once true but is no longer current. | Is this information still current? |
| Contextual mismatch | Information does not fit the relevant jurisdiction, audience, etc. | Does this information apply in this context? |
7. TRACE Walkthrough
In Lesson 10.1, you learned the first element of TRACE: Trace (identify and separate claims). In this lesson, we introduce the second element: Review the claims (or, as operationalised here, examine claims for factual support, omissions, framing, currency, and contextual fit).
Definition: [INSERT AUTHOR-APPROVED TRACE EXPANSION]
Until the exact author-approved expansion is confirmed, we treat Review operationally as the process of:
- Determining whether each claim is supported by evidence, unsupported, false, or uncertain.
- Identifying material omissions that affect interpretation.
- Analysing framing and its potential to mislead.
- Checking the currency of the information.
- Assessing contextual fit (jurisdiction, audience, discipline, culture, etc.).
- Determining whether a technically true statement is still misleading.
- Is the claim factually accurate? (Can it be confirmed against reliable evidence?)
- If it is not false, is it supported by evidence? If not, what evidence is needed?
- Is any material information missing that would change how I interpret the claim?
- How is the claim framed? Does the wording, selection, or emphasis distort the truth?
- Is the information current? Has it changed since the AI’s training data was cut off?
- Does the claim fit the context—the relevant jurisdiction, audience, discipline, culture, and task?
- Could this claim be technically true but still misleading? If so, why?
Many students stop after checking whether a claim is “true” or “false.” However, a claim can be true and still be misleading. The Review step requires you to go beyond simple fact-checking to examine omissions, framing, currency, and context. This is what distinguishes thorough error analysis from superficial verification.
8. Error-Analysis Process Walkthrough
In Lesson 10.2, the error-analysis process focuses on claim → factual support, omission, framing, currency, and contextual fit. This is the second stage of a full TRACE audit.
- Begin with the claim inventory (from Lesson 10.1).
- For each claim, ask: Is this claim true, false, unsupported, or uncertain?
- Check for omissions: What important information is missing that would affect interpretation?
- Analyse framing: How is the claim worded, and what is emphasised or downplayed?
- Check currency: Is the information current? Has it changed?
- Assess contextual fit: Does the claim apply to the relevant jurisdiction, audience, discipline, culture, and task?
- Determine if the claim is technically true but misleading: If so, document the issue.
- Flag for verification: Note what evidence is needed to confirm or correct the claim.
- Assess consequence: What is the impact if the claim is false, misleading, or incomplete?
- Determine priority: Which claims should be verified or corrected first?
9. Worked Example
“All employees are required to complete the new mandatory training module by December 31, 2025. The training, which was developed in partnership with a leading international consulting firm, has been shown to improve workplace safety by 45%. Compliance with this training is a condition of continued employment, and failure to complete it may result in disciplinary action. The training is available in English only, and employees can access it through the company’s internal portal. The training is based on the latest occupational health and safety standards, which are consistent across all jurisdictions in North America.”
Step-by-step analysis
Step 1: Claim inventory
- All employees are required to complete training by December 31, 2025.
- The training was developed with a leading international consulting firm.
- The training has been shown to improve safety by 45%.
- Compliance is a condition of continued employment.
- Failure may result in disciplinary action.
- The training is available in English only.
- Employees can access it via the internal portal.
- The training is based on the latest health and safety standards.
- Those standards are consistent across all North American jurisdictions.
Step 2: Factual support and errors
- The 45% improvement claim: This is unsupported. No source is provided. It may be a fabrication or an exaggeration.
- “Consistent across all jurisdictions”: This is a contextual mismatch. Occupational health and safety standards vary by province/state and country. The claim that they are “consistent across all North American jurisdictions” is false—there are significant differences between, say, Ontario and Texas.
Step 3: Omissions
- Material omission: The passage does not mention how long the training takes, what it covers, or whether employees will be paid for their time.
- Material omission: It does not specify what “disciplinary action” means—is it a warning, suspension, or termination?
- Material omission: It does not mention accommodations for employees who do not speak English.
Step 4: Framing analysis
- Selective emphasis: The passage emphasises the “45% improvement” (which is unsupported) and the threat of discipline, but it downplays the time commitment and accessibility issues.
- Loaded wording: “Mandatory,” “condition of continued employment,” “disciplinary action” create a sense of urgency and threat.
- False authority: “Leading international consulting firm” is vague and implies credibility without evidence.
Step 5: Currency
- The deadline (December 31, 2025) is a date that will pass. If the passage is read after that date, the information is outdated.
- The “latest” standards may become outdated as new standards are released.
Step 6: Contextual fit
- Jurisdictional mismatch: The claim that standards are “consistent across all jurisdictions” is false. Health and safety standards are not uniform.
- Accessibility mismatch: The training is only available in English, which may exclude employees with limited English proficiency.
Step 7: Technically true but misleading?
- The claim that employees are “required” to complete the training may be true, but the urgency and threat of discipline may be disproportionate.
- The claim about the 45% improvement may be technically true in some context, but without evidence, it is misleading.
Step 8: Verification priorities
- Priority 1 (highest): Verify the 45% improvement claim. If it is false or unsupported, the entire justification for the training is undermined.
- Priority 2: Verify the jurisdictional consistency claim. If false, correct it to reflect actual standards.
- Priority 3: Clarify the disciplinary action and accommodations for non-English speakers.
Step 9: Preliminary corrections
- Replace “45%” with “a significant improvement” (or remove if unverified).
- Add a note that standards vary by jurisdiction and that employees should consult their local regulations.
- Add information about training duration, content, and accommodations.
10. Guided Learning Activities
- Purpose: To practise distinguishing false, unsupported, and uncertain claims.
- Learning outcome: Distinguish unsupported claims from demonstrably false claims.
- Estimated time: 15–20 minutes.
Instructions: Examine each claim below. Classify it as false, unsupported, or uncertain. For each, explain your reasoning and state what evidence would be needed to confirm or disprove it.
- “The world’s population reached 8 billion in 2022.”
- “A 2023 study found that 73% of Canadians support universal pharmacare.” (No source is provided.)
- “The new drug reduces symptoms in most patients, but the long-term effects are still being studied.”
- “The current temperature in Vancouver is 22°C.” (You are reading this in January.)
- “The company’s CEO, Dr. A. K. Lee, has a PhD in computer science.” (No source is provided.)
Show model analysis
- True claim — The world population reached 8 billion in November 2022. If the claim is made after that date with the year 2022, it is accurate. (If the year were wrong, it would be a factual error.)
- Unsupported — The study may exist, but without a source, we cannot verify the 73% figure. It is an unsupported quantitative claim.
- Uncertain — The claim acknowledges that long-term effects are still being studied. The “reduces symptoms in most patients” part is a claim that requires verification, but the overall statement expresses uncertainty.
- False (outdated information) — If you are reading this in January and the temperature is not 22°C, the claim is false. It is also potentially outdated because the temperature changes.
- Unsupported — The CEO’s PhD is an unsupported claim. It may be true, but no evidence is provided. It requires verification.
- Purpose: To identify material omissions in an AI-generated response.
- Learning outcome: Identify material omissions.
- Estimated time: 15–20 minutes.
Instructions: Read the following AI-generated email response to a student’s query about a scholarship. Identify at least five material omissions that would be important for the student to know.
“Dear Student,
Thank you for your inquiry about the University Excellence Scholarship. This scholarship is awarded to students with outstanding academic achievement. To apply, please submit your transcripts, a personal statement, and two letters of reference. The scholarship covers tuition and fees for one academic year. We encourage you to apply as soon as possible.
Best regards,
Financial Aid Office”
Show omissions
- Omission 1: The GPA requirement or academic standing needed to be eligible.
- Omission 2: The application deadline (specific date).
- Omission 3: The number of scholarships available or the competition level.
- Omission 4: Whether the scholarship is renewable or only for one year.
- Omission 5: What “tuition and fees” actually covers (e.g., does it include books, housing, meal plans?).
- Omission 6: The selection criteria beyond “outstanding academic achievement” (e.g., extracurriculars, leadership, financial need).
- Omission 7: Whether there are any conditions attached (e.g., maintaining a minimum GPA, full-time enrolment).
- Omission 8: The submission process (online portal, mail, email) and where to find the application form.
Conclusion: The email is polite but lacks the critical information a student needs to make a decision about applying. The omissions are material because they could affect the student’s ability to apply successfully.
- Purpose: To analyse how framing can change interpretation.
- Learning outcome: Identify misleading framing.
- Estimated time: 15–20 minutes.
Instructions: Two different AI-generated summaries of the same event are provided below. Compare them and answer the questions.
Summary 1
“A recent audit revealed that the city’s public transit system is operating at 85% of pre-pandemic ridership levels. While ridership has not fully recovered, the system has maintained service levels and has not raised fares. The audit also noted that customer satisfaction ratings have remained stable, with 78% of riders reporting satisfaction with the service.”
Summary 2
“The city’s public transit system continues to struggle, with ridership still 15% below pre-pandemic levels—a failure that has cost taxpayers millions in subsidies. Despite ongoing investment, customer satisfaction is only 78%, leaving nearly a quarter of riders dissatisfied. City council must take immediate action to address this crisis.”
Questions:
- What facts are presented in both summaries?
- What is the key difference in framing?
- Which summary is more neutral? Which is more misleading? Why?
- Identify at least three framing devices used in Summary 2.
Show analysis
Facts presented in both:
- Ridership is at 85% of pre-pandemic levels (i.e., 15% below).
- Customer satisfaction is 78%.
Key framing differences:
- Summary 1 frames the situation as stable and manageable; Summary 2 frames it as a “crisis” and “failure.”
- Summary 1 emphasises what has been maintained (service levels, fares); Summary 2 emphasises what is lacking (ridership, taxpayer cost).
- Summary 1 presents 78% satisfaction as “stable”; Summary 2 presents it as “only 78%” and highlights the 22% dissatisfied.
Framing devices in Summary 2:
- Loaded wording: “struggle,” “failure,” “crisis,” “taxpayers millions.”
- Selective emphasis: Focusing on the negative (15% below, 22% dissatisfied) and omitting the positives (service maintained, fares stable).
- Value judgment: “must take immediate action” implies urgency and government failure.
- Unsupported implication: “cost taxpayers millions” may be true, but it is presented without context (e.g., what the money was used for).
Conclusion: Summary 1 is more neutral. Summary 2 is more misleading because it uses loaded language and selective emphasis to create a negative impression while still presenting the same basic facts.
- Purpose: To identify contextual mismatches in an AI-generated response.
- Learning outcome: Recognise contextual, disciplinary, audience, jurisdictional, temporal, and cultural mismatches.
- Estimated time: 15–20 minutes.
Instructions: The following AI-generated response was provided to a Canadian small business owner seeking advice on employee leave policies. Identify at least four contextual mismatches and explain why they are problematic.
“Under the Family and Medical Leave Act (FMLA), eligible employees are entitled to up to 12 weeks of unpaid leave per year for family and medical reasons. Employees must have worked for the employer for at least 12 months and have logged 1,250 hours of service. The leave is available for the birth or adoption of a child, for a serious health condition, or to care for a family member. Employers are required to maintain health insurance coverage during the leave. Employees should provide 30 days’ notice for foreseeable leave. For more information, consult the US Department of Labor’s website.”
Show mismatches
- Jurisdictional mismatch: The FMLA is a US law. It does not apply to Canadian businesses. This is a major jurisdictional mismatch.
- Audience mismatch: The small business owner is in Canada, so US-specific advice is not relevant.
- Legal mismatch: Canadian employees have different rights under provincial and federal employment standards legislation (e.g., the Canada Labour Code, ESA).
- Cultural mismatch: The advice assumes a US-style legal framework and may not reflect Canadian values or expectations around leave.
- Source mismatch: “US Department of Labor” is not the relevant authority in Canada.
Conclusion: This response is not just unhelpful—it is potentially harmful because it provides legal information that does not apply. The business owner could rely on this and make incorrect decisions. The appropriate response would be to refer to Canadian employment standards.
11. Exercises
- Difficulty: Introductory
- Estimated time: ~8 min
- Learning outcome: Distinguish unsupported claims from demonstrably false claims.
Classify each claim as false, unsupported, or uncertain. Explain your reasoning.
- “The moon is made of green cheese.”
- “8 out of 10 dentists recommend this toothpaste.” (No source provided.)
- “The results of the study are inconclusive, and further research is needed.”
- “The company’s revenue doubled last year.” (No financial reports are cited.)
Show answers
- False — the moon is not made of green cheese. This is a factual error (and also a fabrication, but the key point is it is demonstrably false).
- Unsupported — the claim is plausible but lacks evidence. It may be true or false; we cannot determine without a source.
- Uncertain — the statement itself expresses uncertainty. The evidence is inconclusive.
- Unsupported — the claim is quantitative and lacks evidence. It requires verification from financial reports.
- Difficulty: Intermediate
- Estimated time: ~8 min
- Learning outcome: Identify material omissions.
Identify three material omissions in the following AI-generated product description:
“The ‘CleanAir 3000’ air purifier is the most advanced air purification system on the market. It removes 99.9% of airborne particles and is recommended by doctors. With its sleek design and quiet operation, it is the perfect addition to any home. Order now and get free shipping!”
Show omissions
- Omission 1: The cost of the product (price).
- Omission 2: The size, coverage area, or specifications (what size room does it work for?).
- Omission 3: The maintenance cost (e.g., filter replacements) and frequency.
- Omission 4: Which doctors recommended it, and in what context.
- Omission 5: The warranty, return policy, and customer support.
- Omission 6: Whether the 99.9% claim is based on independent testing or the manufacturer’s own tests.
Note: These omissions are material because they affect the consumer’s decision to purchase the product.
- Difficulty: Intermediate
- Estimated time: ~8 min
- Learning outcome: Identify misleading framing.
Identify four framing devices in the following statement:
“In a stunning reversal, the government has admitted that its flagship policy has failed to deliver the promised results. Despite spending billions of taxpayer dollars, the program has produced nothing but disappointment. Critics have rightly condemned this waste of public funds, and it is clear that the government has lost its way.”
Show framing devices
- Loaded wording: “stunning reversal,” “failed,” “disappointment,” “waste,” “lost its way.”
- Selective emphasis: Focuses on the negative outcome and omits any potential benefits.
- Emotional appeal: “taxpayer dollars” and “public funds” invoke emotional responses.
- False authority: “Critics have rightly condemned” implies that criticism is justified, but the critics are not named.
- Broad generalisation: “the government has lost its way” is a sweeping value judgment.
- Difficulty: Introductory
- Estimated time: ~5 min
- Learning outcome: Identify outdated information.
Why might the following claim be outdated, and what would you need to verify?
“The current CEO of Apple is Tim Cook.”
Show analysis
Why it may be outdated: Tim Cook has been CEO of Apple since 2011. However, if you are reading this in 2026 or later, it is possible that a new CEO has been appointed. The information may still be accurate, but it should be verified.
What to verify: Check Apple’s official website or a reliable news source for the current CEO.
Note: As of this writing (2026), Tim Cook is still CEO, but this is a good example of a claim that should be checked for currency.
- Difficulty: Intermediate
- Estimated time: ~8 min
- Learning outcome: Recognise contextual mismatches.
Identify the type of contextual mismatch in each scenario and explain why it is problematic.
- An AI provides information about UK health and safety regulations to a Canadian construction company.
- An AI responds to a query from a primary school teacher with a lesson plan that uses university-level academic references.
- An AI provides a recipe that assumes the reader has a stand mixer, but the reader is a student living in a dormitory without kitchen equipment.
- An AI provides a summary of Indigenous land rights using only colonial legal sources, without referencing Indigenous perspectives or legal frameworks.
Show answers
- Jurisdictional mismatch. UK regulations do not apply in Canada. The information is not relevant and could lead to legal non-compliance.
- Audience mismatch. The content is too advanced for a primary school teacher. The teacher may be unable to use the material effectively.
- Contextual mismatch (accessibility/resource). The assumption that the reader has a stand mixer is inappropriate for a student in a dormitory. The recipe should be adapted.
- Cultural and epistemic mismatch. The response privileges one legal framework over another and omits Indigenous perspectives. This is not only incomplete but potentially harmful and disrespectful.
- Difficulty: Advanced
- Estimated time: ~10 min
- Learning outcome: Explain how a statement may be technically true but materially misleading.
The following statement is technically true but misleading. Rewrite it to be more accurate and less misleading, preserving the key information.
“The new policy reduced costs by 15%, saving the department $50,000 annually.”
Show rewrite
Original problem: The statement omits that the policy cost $100,000 to implement, and that the $50,000 saving is based on a pilot study with a small sample. It also does not mention that the savings are projected, not guaranteed.
Possible rewrite: “Based on a pilot study with a sample of 50 cases, the new policy is projected to reduce costs by 15%, which would represent an estimated annual saving of $50,000. The policy required an initial investment of $100,000. These projections are subject to change and should be verified with full-scale implementation data.”
Explanation: The rewrite adds context (pilot study, sample size, investment cost) and qualifies the savings as “projected” rather than guaranteed, making the statement more accurate and less misleading.
12. Homework Problems
Identify the factual error in the following claim and correct it:
“The Great Wall of China is visible from space with the naked eye.”
Show answer
Factual error: The claim is false. The Great Wall of China is not visible from space with the naked eye. It is difficult to see even from low Earth orbit, and requires optical aid. This is a common myth.
Correction: “The Great Wall of China is not visible from space with the naked eye; it requires optical aid and even then is difficult to distinguish from other geological features.”
Classify the following claim and explain what evidence would be needed to support it:
“The majority of Canadians support the government’s climate change policy.”
Show answer
Classification: Unsupported claim.
Why: The claim is plausible but lacks evidence. It may be true or false, but we cannot determine without data.
Evidence needed: A representative public opinion poll with a clear methodology (sample size, margin of error, question wording). The source of the poll should be cited.
Identify four material omissions in the following AI-generated job posting:
“We are hiring a Software Developer. The position offers a competitive salary, flexible work hours, and opportunities for professional development. Apply now by sending your resume to hr@company.com.”
Show omissions
- Omission 1: The specific salary range or competitive salary amount.
- Omission 2: The required skills and qualifications (programming languages, experience level).
- Omission 3: The job location (remote, hybrid, or in-office).
- Omission 4: The application deadline or review process.
- Omission 5: The company name and industry.
- Omission 6: The benefits package beyond “flexible work hours.”
Analyse the framing in the following statement and identify four framing devices:
“Despite the critics’ baseless claims, the evidence is crystal clear: our product is the safest and most reliable on the market. With thousands of satisfied customers and a track record of excellence, there is simply no better choice for discerning consumers.”
Show framing devices
- Loaded wording: “baseless claims,” “crystal clear,” “excellence,” “discerning.”
- Dismissal of criticism: “critics’ baseless claims” frames dissent as illegitimate.
- Overstatement: “safest and most reliable on the market” is a superlative claim without evidence.
- Appeal to popularity: “thousands of satisfied customers” implies popularity equals quality.
- Exclusionary framing: “discerning consumers” implies that anyone who chooses another product is not discerning.
Identify why the following claim is likely outdated and what you would need to verify:
“The current population of India is 1.2 billion.”
Show analysis
Why it is likely outdated: India’s population has grown significantly over the past decade. As of 2026, the population is estimated to be over 1.4 billion. The claim of 1.2 billion is from approximately 2011–2015.
What to verify: Check the latest population estimates from the United Nations, World Bank, or the Indian government’s census data.
Note: This is an example where the date of the information matters. A source from 2011 would be outdated by 2026.
Identify the type of contextual mismatch and explain why it is problematic:
“The patient is exhibiting symptoms consistent with the flu. The doctor should prescribe Tamiflu, which is the standard treatment for influenza in the United States.”
Show analysis
Type of mismatch: Jurisdictional and professional mismatch.
Why it is problematic: The advice is written for a US context. In Canada, prescribing guidelines may differ. Additionally, the advice is presented as if the “doctor” is the user, but if the user is not a medical professional, this is inappropriate medical advice.
Additional concerns: The advice does not consider the patient’s specific medical history, allergies, or other medications. It is a generalised statement that should not be used for clinical decision-making.
Appropriate response: The response should be qualified: “In the US, Tamiflu is a standard treatment, but prescribing decisions should be made by a qualified healthcare professional based on the patient’s individual circumstances.”
Explain why the following statement, though technically true, is misleading. Then rewrite it to be more accurate.
“The company’s profits increased by 5% this year.”
Show analysis and rewrite
Why it is misleading: The statement does not provide context. The profit increase may be due to inflation, a one-time event, or cost-cutting that reduced future capacity. It also does not mention whether the 5% increase is relative to a year in which profits were exceptionally low.
Possible rewrite: “The company’s net profits increased by 5% this year, from $100 million to $105 million. This increase was driven primarily by a one-time sale of assets. Excluding that sale, profits declined by 2%.”
Explanation: The rewrite provides context (the base amount, the driver of the increase) and qualifies the claim, making it more accurate and less misleading.
Analyse the following AI-generated passage from a travel blog. Identify at least five issues (errors, omissions, framing, outdated info, or context mismatches) and explain how you would correct them.
“Vancouver is the most beautiful city in North America. The weather is perfect year-round, with temperatures rarely dropping below 15°C in winter. The city is also one of the most affordable places to live, with average rent for a one-bedroom apartment at just $800 per month. Visitors can enjoy the famous Stanley Park, Granville Island, and the Museum of Anthropology, which has the world’s largest collection of Indigenous art. The city is also home to the headquarters of Lululemon, a leading international brand.”
Show analysis
- Framing/overstatement: “Most beautiful city in North America” is a subjective claim presented as fact.
- Outdated/factual error: “Temperatures rarely dropping below 15°C in winter” — Vancouver’s winter average is around 3–8°C. This is a factual error.
- Outdated information: “Average rent ... at just $800 per month” — Vancouver rent is significantly higher (often over $2,000 for a one-bedroom). This is outdated or false.
- Omission: The Museum of Anthropology does have significant Indigenous art, but claiming it has “the world’s largest collection” is likely false or unverifiable.
- Contextual mismatch: The passage is written for a travel audience, but the affordability claim is misleading and could cause readers to make poor decisions.
- Selective emphasis: The passage highlights positive features and omits any negatives (rain, high cost of living, traffic).
Correction approach: Verify the weather data, rent prices, and the museum’s collection claim. Remove subjective superlatives (“most beautiful”) or qualify them as opinions. Add context about Vancouver’s climate (rainy winters) and cost of living.
Rewrite the following claim to be more accurate and less misleading. Identify what information you would need to verify before doing so.
“The new teaching method has been proven to improve student outcomes by 30%.”
Show rewrite and verification needs
Verification needed:
- The source of the “30%” claim (which study, sample size, methodology).
- What “student outcomes” means (test scores, engagement, retention?).
- The context (which subjects, which grade levels, which student populations).
- Whether the study was peer-reviewed and replicated.
Possible rewrite: “A 2024 study of 500 high school students in Ontario found that students using the new teaching method scored an average of 15% higher on standardised tests compared to a control group. Further research is needed to confirm these findings in other contexts.”
Key improvements: Adds source details, qualifies the claim, avoids overgeneralisation, and acknowledges the need for further research.
The following passage is from an AI-generated press release. Identify three material omissions and three framing devices, and explain how they combine to mislead.
“InnovateTech Inc. is proud to announce a record-breaking quarter with revenue growth of 25%. This achievement reflects our commitment to innovation and our industry-leading team. With a strong balance sheet and a robust pipeline of new products, the company is well-positioned for continued success. Shareholders can be confident in the company’s future.”
Show analysis
Material omissions:
- Does not mention profitability (revenue growth does not equal profit growth).
- Does not mention the costs associated with the revenue growth (e.g., marketing spend, R&D investment).
- Does not mention risks or challenges the company faces.
- Does not provide a comparison to the previous quarter or to competitors.
Framing devices:
- Selective emphasis: Focuses on revenue growth (positive) and omits any negatives.
- Loaded wording: “record-breaking,” “commitment to innovation,” “industry-leading,” “strong balance sheet,” “robust,” “continued success.”
- Appeal to shareholder confidence: “Shareholders can be confident” is a reassurance that may be unwarranted.
How they combine to mislead: The omissions and framing create a picture of unqualified success, even though the company may have significant challenges. A reader (especially an investor) would be misled into thinking the company is in better health than it may be.
13. Mini-Project: Context and Omission Review
- Purpose: To integrate the analysis of factual errors, unsupported claims, material omissions, misleading framing, outdated information, and contextual mismatches in a structured review of a synthetic AI-generated response.
- Learning outcomes: All Lesson 10.2 outcomes.
- Estimated effort: 75–90 minutes.
Passage for analysis
“The World Health Organization (WHO) has declared that the COVID-19 pandemic is officially over as of 2025. This decision is based on the widespread availability of vaccines and the development of herd immunity, which now protects the global population. Recent studies have shown that immunity from vaccination lasts for at least five years, and breakthrough infections are extremely rare—occurring in less than 1% of vaccinated individuals. The WHO recommends that all countries immediately lift all remaining restrictions, including mask mandates and travel bans, to return to pre-pandemic normalcy. The organisation has noted that long COVID is a minor concern that affects only a small subset of patients and that most people recover fully without medical intervention. According to Dr. Margaret Chan, a former WHO official, ‘The pandemic is a thing of the past, and we must focus on other priorities.’ Governments should now divert all public health funding away from COVID-19 and toward other areas of healthcare.”
Required deliverable
A structured review report that includes the following sections:
- Human purpose: What is the intended purpose of this passage?
- Intended audience: Who is the likely audience?
- Claim inventory: List every claim in the passage (briefly).
- Factual errors: Identify any claims that are demonstrably false and explain why.
- Unsupported statements: Identify any claims that are unsupported and explain what evidence would be needed.
- Material omissions: Identify at least three material omissions and explain why they matter.
- Framing choices: Identify at least four framing devices and explain their effect.
- Date and currency: Assess whether any information is likely outdated.
- Jurisdiction: Does the passage assume a specific jurisdiction? If so, is it appropriate?
- Disciplinary context: Is the information accurate from a public health perspective?
- Accessibility or audience fit: Is the passage appropriate for the intended audience?
- Possible consequences: What might happen if this passage is used as a basis for policy or personal decisions?
- Required verification: What needs to be verified before this passage can be trusted?
- Proposed correction: How would you correct or qualify the most problematic claims?
- Remaining uncertainty: What uncertainty remains after correction?
- Escalation: When would you escalate this to a subject-matter expert?
- Preliminary recommendation: Would you use this passage? Why or why not?
Self-assessment criteria
- All claims are inventoried and classified.
- Factual errors are correctly identified and explained.
- Unsupported claims are flagged with clear evidence needs.
- Material omissions are identified and justified as material.
- Framing devices are analysed with specific examples.
- Currency, jurisdiction, and discipline are assessed appropriately.
- Consequences and verification priorities are reasonable.
- Corrections are evidence-based and do not introduce new unsupported claims.
Show staged hints
Hint 1 (Factual errors): The claim that the WHO declared the pandemic “officially over” is a key claim to check. Has the WHO actually made such a declaration? What does “over” mean in this context?
Hint 2 (Unsupported claims): The claim that immunity lasts “at least five years” and that breakthrough infections occur in “less than 1%” are quantitative claims that require sources. They may be unsupported.
Hint 3 (Omissions): Does the passage mention vulnerable populations? Does it discuss variants? Does it mention that some countries may not have widespread vaccine access?
Hint 4 (Framing): The passage uses phrases like “a minor concern” and “a thing of the past.” These are framing devices that downplay the ongoing impact of COVID-19.
Hint 5 (Context): The recommendation to “divert all public health funding away from COVID-19” is a policy recommendation that may not be appropriate in all contexts, especially for countries still experiencing outbreaks.
Show solution architecture
- Purpose: The passage appears to be a summary of the global COVID-19 situation, possibly intended for policy makers or the general public.
- Audience: General public or public health officials.
- Key claims (partial list):
- WHO declared the pandemic over in 2025.
- Vaccines are widely available.
- Herd immunity protects the global population.
- Immunity lasts at least five years.
- Breakthrough infections occur in less than 1% of vaccinated individuals.
- All restrictions should be lifted.
- Long COVID is a minor concern.
- Most people recover fully without intervention.
- Dr. Margaret Chan said the pandemic is “a thing of the past.”
- All funding should be diverted from COVID-19.
- Factual errors:
- The WHO has not officially declared the pandemic “over” — as of 2026, it is still a global health concern. The claim is false.
- Dr. Margaret Chan’s quote is likely fabricated or taken out of context.
- Unsupported claims:
- Immunity lasts at least five years — lacks evidence; actual duration is still being studied.
- Breakthrough infections in less than 1% — depends on variant and population; likely an oversimplification.
- Omissions:
- Vulnerable populations are not mentioned.
- Variants (e.g., new strains) are not mentioned.
- Global inequity in vaccine access is not mentioned.
- Framing:
- “a minor concern” — downplays long COVID.
- “a thing of the past” — dismisses ongoing challenges.
- “divert all funding” — extreme policy recommendation.
- Recommendation: Do not use this passage. It contains factual errors, unsupported claims, omissions, and misleading framing. It should be withdrawn.
Show illustrative partial example (based on a different passage)
Illustrative example (Passage G — different from the main passage):
Passage: “The Bank of Canada has announced that interest rates will remain at 0.25% until 2024. This policy is expected to stimulate economic growth and create jobs.”
Partial review:
- Factual error: The Bank of Canada’s interest rate is not 0.25% in 2026 (it has changed). The claim is outdated and factually incorrect for the current date.
- Unsupported: The claim that the policy will “stimulate economic growth and create jobs” is an unsupported causal claim.
- Omission: Does not mention the potential risks of low interest rates (inflation, asset bubbles).
- Contextual mismatch: The passage may be relevant to a Canadian audience, but the time frame is outdated.
Conclusion: The passage should not be used without updating the interest rate and qualifying the causal claims.
14. Self-Check Quiz
Question 1: What is the key difference between a false claim and an unsupported claim?
Show answer and explanation
Answer: b.
A false claim can be demonstrated to be incorrect. An unsupported claim is one for which we lack evidence; it may or may not be true. The distinction is about the relationship to evidence, not intent or severity.
Question 2: Which of the following is an example of a material omission?
Show answer and explanation
Answer: b.
This is a material omission because the missing information (the conditions) would affect how a reader interprets the claim.
Question 3: Which of the following is an example of misleading framing?
Show answer and explanation
Answer: b.
Selective emphasis—highlighting one type of information while omitting another—is a form of misleading framing.
Question 4: When is information considered “outdated”?
Show answer and explanation
Answer: b.
Outdated information is information that is no longer accurate because the facts have changed. The age of the information alone is not the criterion.
Question 5: Which of the following is an example of a jurisdictional mismatch?
Show answer and explanation
Answer: b.
This is a jurisdictional mismatch because the information is from the wrong legal jurisdiction.
Question 6: How can a statement be technically true but still misleading?
Show answer and explanation
Answer: b.
A statement can be factually correct but still misleading if it omits important context, highlights only favourable information, or uses framing that distorts interpretation.
Question 7: What is the primary purpose of the Review element in the TRACE framework?
Show answer and explanation
Answer: b.
The Review element involves evaluating the claims you have traced for accuracy, completeness, framing, currency, and contextual fit.
Question 8: You are a student in a business course. Your AI-generated draft for a case study analysis states: “The company’s market share increased by 15% last year, making it the industry leader.” You know the market share data is from a report that you cannot access. What should you do?
Show answer and explanation
Answer: b.
The claim is quantitative and attributed to a source that you cannot verify. You should flag it as unsupported and attempt to find the original report. Do not accept it or assume it is false; it may be true but needs verification.
15. Assignment Connections
Assignment 5: Verification and Fact-Checking Assignment requires you to verify an AI-generated passage, evaluate evidence quality, identify errors and omissions, and produce corrected wording.
This lesson has prepared you for the following aspects of that assignment:
- Omission analysis: Identifying what is missing from the passage and why it matters.
- Framing analysis: Recognising how wording and selection can distort interpretation.
- Contextual mismatch: Assessing whether the information fits the relevant jurisdiction, audience, discipline, and culture.
- Currency: Checking whether the information is current.
- Distinguishing false from unsupported: Knowing the difference and responding appropriately.
What to retain:
- Your omission and framing analyses from the exercises and mini-project.
- Your notes on contextual mismatches and currency issues.
- Your preliminary corrections and qualifications.
You will build on these in Lesson 10.3 (inference and precision) and Lesson 10.4 (TRACE, correction, and communication).
Select at least two of the following prompts to reflect on after completing this lesson:
- How has this lesson changed your understanding of misinformation? Did you previously think that “true” information cannot be misleading?
- Think of a time when you were misled by an omission or framing. How did that happen, and how might you avoid it in the future?
- What is the most common contextual mismatch you have encountered in AI-generated content? How did you recognise it?
- How will you approach AI-generated content differently after learning about omissions and framing?
- Describe a situation in your field or discipline where outdated information could cause harm. How would you verify currency?
Show reflection-planning checklist
- Have you connected the concepts to your own experience with AI-generated text?
- Have you considered how omissions and framing might affect decision-making in your field?
- Have you identified specific strategies you will use to detect omissions and framing in the future?
16. Key Takeaways
- A false claim conflicts with reliable evidence; an unsupported claim lacks evidence but may be true. Distinguishing them is essential.
- Material omissions leave out information needed for a reasonable interpretation or decision.
- Misleading framing uses word choice, selection, emphasis, or balance to distort interpretation.
- Outdated information was once true but is no longer current; currency matters in time-sensitive domains.
- Contextual mismatches occur when information does not fit the jurisdiction, audience, discipline, culture, or task.
- A statement can be technically true but still misleading through omission, framing, or selective emphasis.
- The second element of TRACE, Review, involves examining claims for support, omissions, framing, currency, and context.
- Responsible use of AI requires going beyond simple fact-checking to assess completeness, framing, fit, and currency.
17. Key Terms
- Factual error
- A claim that conflicts with reliable evidence and can be demonstrated to be false.
- Unsupported claim
- A claim that is plausible but lacks sufficient evidence; it may be true or false.
- Uncertain claim
- A claim for which the evidence is ambiguous, incomplete, or conflicting; the truth value cannot be determined.
- Material omission
- Missing information that is required for a reasonable interpretation or decision.
- Misleading framing
- The use of wording, selection, emphasis, or balance that distorts interpretation, even if each fact is true.
- Outdated information
- A claim that was once true but is no longer current.
- Contextual mismatch
- When information is inaccurate or unsuitable because it does not fit the jurisdiction, audience, discipline, culture, or task.
- Selective truth
- Presenting only part of the truth in a way that misleads, while omitting context or counter-evidence.
- Loaded wording
- Emotionally charged or value-laden language used to influence interpretation.
- Jurisdictional mismatch
- Using laws, regulations, or standards from one jurisdiction in another where they do not apply.
- Audience mismatch
- Using language, examples, or assumptions that are inappropriate for the intended audience.
- Currency
- The timeliness of information; whether it is current and up to date.
18. Looking Ahead: Lesson 10.3
In Lesson 10.3, you will examine invalid inference, false precision, and uncertainty. You will learn to:
- Distinguish evidence, assumptions, inferences, and conclusions.
- Identify invalid inference, overgeneralisation, and correlation-causation errors.
- Identify false precision and unjustified numerical specificity.
- Distinguish linguistic confidence from evidential support.
- Communicate uncertainty using language proportionate to available evidence.
This will prepare you for the final stage of the TRACE framework in Lesson 10.4.
19. Sources & Further Reading
- Primary source: COMP 3270 Chapter 7: Hallucination, Error, and Uncertainty.
- Secondary source: Chapter 3: Effective and Responsible AI Use.
- Secondary source: Chapter 6: Human-AI Collaboration.
- Appendix C: Verification Toolkit.
- Appendix D: Responsible-Use Toolkit.
- Further reading: [INSERT AUTHOR-APPROVED VERIFICATION AND ERROR-ANALYSIS REFERENCE]