Table of Contents
- 1. Lesson Overview
- 2. Learning Outcomes
- 3. Before You Begin
- 4. Unit 10 Retrieval Practice
- 5. Opening Scenario
- 6. Core Content
- 6.1 Reference Metadata Verification
- 6.2 Lateral Reading
- 6.3 Original-Source Tracing
- 6.4 Citation Laundering
- 6.5 Quotation Verification
- 6.6 Quantitative-Claim Verification
- 6.7 Circular Reporting
- 7. CLAIM Walkthrough
- 8. Verification-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 Lessons 11.1 and 11.2, you learned to evaluate sources and align evidence with claims. But what if the source itself is problematic—because it doesn’t exist, because its metadata are wrong, or because it’s not the original source at all? This lesson equips you with the investigative skills to verify the source itself.
You will learn to verify reference metadata, read laterally (leaving a source to investigate its publisher, author, and reputation), and trace claims, quotations, and statistics back to their original origins. You will also learn to recognise citation laundering—where a claim gains the appearance of authority through a chain of citations that ultimately do not support it.
These skills are essential for detecting fabricated citations, misattributions, and circular reporting—problems that are especially common in AI-generated content.
2. Learning Outcomes
By the end of this lesson, you should be able to:
- Verify whether a cited source exists and whether its reference metadata are correct.
- Conduct lateral reading to investigate a source’s author, publisher, reputation, and evidence base.
- Trace quotations, statistics, findings, and repeated claims to their original sources.
- Recognise circular reporting and identify when multiple sources depend on the same origin.
- Identify citation laundering patterns, including chains of dependency and misrepresentation.
- Distinguish direct evidence from summaries, reposts, and AI-generated descriptions.
- Create provenance maps that document the origin and transformations of a claim.
- Apply the CLAIM framework to verify source existence, authenticity, and support.
- Integrate TRACE error analysis with citation tracing to identify fabrication, misattribution, and omission.
3. Before You Begin
- Assigned reading: Chapter 8 of the course textbook: Verification, Evidence, and Epistemic Responsibility (sections on lateral reading, citation verification, original-source tracing, and citation laundering).
- Prerequisite knowledge: You should be able to distinguish primary, secondary, and tertiary sources; evaluate authority, relevance, and currency; align evidence with claims; and assess source independence.
- Library and search tools: You may use institutionally available library portals, public search engines (Google Scholar, Internet Archive), or open-access resources. No paid database is required.
- Optional AI tool: You may use an approved AI system for discovery (e.g., finding possible sources), but you must independently verify every citation. AI-generated citations are not reliable without verification.
- Supplied-source alternative: Every activity includes fully supplied source sets so you are never disadvantaged by lack of access.
- Privacy and copyright: Do not upload confidential, proprietary, or third-party material to any unapproved system. Use only synthetic, public, or supplied information.
- Academic integrity: All work you submit must be your own. Document any AI assistance you use.
- Estimated preparation time: 20–30 minutes to review Chapter 8 and the Unit 10 summary.
4. Unit 10 Retrieval Practice
The following questions review TRACE and claim analysis from Unit 10. Answer each before revealing the feedback.
Question 1: Using TRACE, what type of error is present if an AI response cites a study that you cannot find in any database, and the citation includes an author, journal, and year that appear plausible but do not match any real publication?
Show feedback
Answer: This is a fabrication — a completely invented citation. It may also be a hallucination if the AI generated it from patterns in its training data without real-world grounding. The key distinction is that fabrication involves a deliberate (or non-deliberate) creation of false information.
Action: You must verify the citation independently. If it doesn’t exist, it is fabricated.
Question 2: What is the difference between a fabricated citation and a misattributed citation?
Show feedback
- Fabricated: The source does not exist at all. The author, title, journal, and year are invented.
- Misattributed: The source exists, but it does not support the claim, or the claim is attributed to the wrong source.
Both are serious problems, but misattribution requires tracing the claim to determine whether the source was misread or misrepresented.
Question 3: Why is it important to prioritise claims for verification? Give one criterion for deciding which claim to verify first when multiple claims are present.
Show feedback
Answer: Prioritisation ensures that the most consequential claims receive the most attention. A criterion could be the potential impact if the claim is false (e.g., health, safety, legal, financial consequences), or the centrality of the claim to the overall argument.
5. Opening Scenario: A Claim Repeated Across Sources
This scenario runs throughout the lesson. Read it carefully and return to it as you learn each new concept.
You are a graduate student researching the health benefits of a popular supplement, omega-3 fatty acids. You find an AI-generated summary that states:
“A 2021 study published in the Journal of Clinical Nutrition by Dr. Sarah Chen and colleagues found that omega-3 supplementation reduces the risk of cardiovascular disease by 25%. This finding has been widely reported in the media and is cited in numerous health blogs and government recommendations.”
You search for the study and find:
- Website A: A health news article that reports “omega-3s cut heart disease risk by 25%” and cites “Chen et al., 2021, Journal of Clinical Nutrition.”
- Website B: A blog post that repeats the same statistic and cites Website A as its source.
- Website C: A government health page that mentions omega-3s but does not cite any specific study.
- Website D: A nutrition supplement company’s website that uses the statistic to promote its product and cites “Chen et al.”
- Website E: A scientific review article from 2023 that discusses omega-3s and lists many studies, but does not mention Chen et al. or the 25% figure.
Your task:
- Does the Chen et al. study actually exist? How would you verify this?
- If the study exists, does it actually support the 25% claim? What would you need to check?
- Which sources are independent? Which are dependent? Which are repeating the same claim without adding new evidence?
- What is the original source of the 25% claim? Is it the Chen et al. study, or could it have originated elsewhere?
Show guided analysis of the scenario
1. Verifying existence: To verify the study, you would search for “Chen, Journal of Clinical Nutrition, 2021” in a database like PubMed, Google Scholar, or the journal’s website. You would check the author list, title, and abstract. If it doesn’t appear, it may be fabricated.
2. Verifying support: If the study exists, you would read the abstract or full text to see if it actually reports a 25% reduction in CVD risk. You would check the population, study design, and whether the 25% figure applies to the entire population or a subgroup.
3. Independence vs. dependence:
- Independent: Website E (the scientific review) if it does not cite the Chen study; but it doesn’t support the claim either.
- Dependent: Websites A, B, and D all appear to derive from the same original source (Chen et al.). They are not independent. They are repetitions.
- Not relevant: Website C (government page) doesn’t support the claim; it’s a different source.
4. Original source: The chain appears to be: Chen et al. (if it exists) → Website A (news article) → Website B (blog) → Website D (company). If the Chen study doesn’t exist, the entire chain may be built on a fabricated source. Alternatively, the 25% figure might have originated in a different study and was misattributed to Chen.
6. Core Content
6.1 Reference Metadata Verification
Reference metadata are the bibliographic elements that describe a source: author(s), title, journal or publisher, date, volume, issue, pages, edition, DOI, ISBN, report number, URL, version, and access date.
Verifying metadata is the first step in source verification. You must confirm that each element is correct and that the source exists. Metadata accuracy does not prove claim support—it only proves that the reference points to a real source.
To verify metadata:
- Search for the source in library databases, Google Scholar, or the publisher’s website.
- Use the DOI to locate the source (if provided).
- Check the author’s name and affiliation.
- Verify the journal title against the publisher’s official list.
- Confirm the year, volume, and page numbers.
Common problems:
- Fabricated sources (no record exists).
- Misattributed authors (the author is real but did not write the cited work).
- Incorrect journal names (the journal exists but the article is not in it).
- Incorrect dates, volumes, or page numbers.
- Outdated versions or editions.
Example: An AI cites “Smith, J. (2023). The Science of AI. Cambridge University Press.” You search for this title and find that Cambridge University Press has no record of it. The source is likely fabricated.
6.2 Lateral Reading
Lateral reading is the practice of leaving a source to investigate its publisher, author, institution, reputation, external assessments, funding, methodology, and other coverage. Instead of evaluating a source based solely on its own presentation, you look at what other sources say about it.
Lateral reading is especially important for unfamiliar sources. A website may look professional but have no credibility. Lateral reading allows you to:
- Learn about the author’s expertise and affiliations.
- Determine the publisher’s reputation and review process.
- Identify conflicts of interest or funding biases.
- See how other experts or fact-checkers have assessed the source.
- Identify whether the source has been retracted or criticised.
How to read laterally:
- Leave the source you are evaluating.
- Search for the author’s name and affiliation.
- Search for the publisher or organisation.
- Search for the source title with terms like “review,” “critique,” or “retraction.”
- Check Wikipedia or other encyclopedias for background.
- Check fact-checking sites (e.g., Snopes, FactCheck.org, or disciplinary-specific resources).
Example: You find a study on a website that looks legitimate. You search for the author and discover they are not affiliated with any university and have no other publications. This raises a red flag. You also find a forum discussion where other researchers have questioned the study’s methodology.
6.3 Original-Source Tracing
Original-source tracing is the process of following a claim, quotation, statistic, or finding back to its ultimate origin. This involves moving from derivative sources (news articles, blogs, summaries) to the primary source (original study, official record, dataset).
Why trace to the original?
- Derivative sources may distort, simplify, or omit important context.
- The original source may contain qualifications, limitations, or definitions that are lost in summary.
- The claim may have been misattributed—it may not actually appear in the cited source.
- The original source may be inaccessible or paywalled, requiring you to document the gap.
What to trace:
- Quotations: Find the exact wording in the original context. Check for omissions, paraphrasing, or translation issues.
- Statistics: Find the original dataset or report. Verify the measure, population, denominator, time period, and calculation.
- Scientific claims: Locate the original study or systematic review. Check the study design, sample, methods, and limitations.
- Legal claims: Locate the current authoritative legal text, regulation, or case law.
- Policy claims: Locate the current official policy or government document.
- Product claims: Locate the current official product documentation.
- News claims: Trace back to original reporting, press releases, or primary records.
Example: A news article says “a study found that 50% of adults have anxiety.” You trace it back and find the original study used a different measure, had a small sample, and only found 50% in a specific population. The original context changes the interpretation.
6.4 Citation Laundering
Citation laundering occurs when a claim gains an appearance of authority by being repeated through citations that do not support it, trace only to one another, or ultimately lead to an unsupported or inaccessible source. It is the process of making a weak or fabricated claim look legitimate through a chain of citations.
Common patterns:
- Circular citation: A cites B, B cites C, and C cites A, with no actual supporting evidence.
- Press release repetition: Multiple news articles and blogs all repeat the same press release, creating the illusion of widespread support.
- Summary cites review: A summary cites a review that cites an inaccessible or unsupported claim.
- AI-generated repetition: AI-generated wording is repeatedly published and later cited as if it came from a human expert.
- Authentic source misrepresentation: A source is authentic but is misrepresented—the claim is not actually in the source.
- Dependent sources: Several sources all trace back to the same unverified original, creating the illusion of corroboration.
Why it matters: Citation laundering can make a claim appear well-supported when it is not. This is especially dangerous in AI-generated content, where AI systems may generate citations that look plausible but are part of a laundering chain.
How to detect citation laundering:
- Trace each citation back to its original source.
- Check whether the cited source actually supports the claim.
- Map the dependencies—are multiple sources all citing the same origin?
- Check for circular references.
- Look for inaccessible or paywalled sources that cannot be verified.
Example: Website A says “Study X found that Y causes Z.” Website B cites Website A. Website C cites Website B. The original Study X either doesn’t exist, doesn’t support the claim, or is inaccessible. The claim has been laundered through a chain of citations.
6.5 Quotation Verification
Quotation verification is the process of confirming the exact wording, speaker or author, source, date, context, omissions, translation, and authenticity of a quoted statement.
Quotations are often misrepresented. They may be:
- Truncated to remove important context.
- Paraphrased but presented as a direct quote.
- Misattributed to the wrong speaker.
- Translated inaccurately.
- Fabricated entirely.
- Out of context—the speaker was making a different point.
To verify a quotation:
- Search for the exact wording in quotation marks.
- Search for the speaker and key phrases.
- Find the original source (interview, article, speech, book, etc.).
- Read the surrounding context.
- Check if the quote appears in multiple reliable sources.
- Be wary of quote-aggregation websites (e.g., Quote.com, BrainyQuote) as final authorities—they often misattribute quotes.
Example: A blog post quotes Albert Einstein as saying “the definition of insanity is doing the same thing over and over and expecting different results.” This quote is widely misattributed to Einstein; there is no record of him saying it. Verification would reveal this.
6.6 Quantitative-Claim Verification
Quantitative-claim verification is the process of verifying numerical claims, including measures, numerators and denominators, populations, samples, periods, baselines, units, methods, uncertainty, and whether percentages and comparisons are calculated correctly.
Quantitative claims are particularly susceptible to error:
- Numbers may be misread or misquoted.
- Percentages may be misapplied (e.g., a 10% increase vs. a 10 percentage point increase).
- Denominators may be omitted or misreported.
- Populations may be switched (e.g., “50% of students” vs. “50% of respondents”).
- Time periods may be confused.
- Uncertainty (confidence intervals, margins of error) may be omitted.
To verify a quantitative claim:
- Trace to the original dataset or report.
- Confirm the exact measure used.
- Check the numerator and denominator.
- Confirm the population and sample.
- Check the time period.
- Look for the confidence interval or margin of error.
- Recalculate percentages to ensure accuracy.
Example: A news article says “unemployment dropped by 50%.” You trace it back and find the unemployment rate went from 6% to 3%—a 3 percentage point drop, but a 50% relative decrease. Both may be correct, but the framing changes the interpretation.
6.7 Circular Reporting
Circular reporting is a pattern where sources cite each other in a loop, creating the illusion of corroboration. A cites B, B cites C, and C cites A, or a claim is repeated across sources that all rely on the same original (often unverified) source.
Circular reporting is a form of citation laundering. It is especially common in:
- Wikipedia editing wars (where sources are misused to support a claim).
- Media reporting (where multiple outlets repeat the same wire story without verification).
- Academic citation networks (where a claim is cited repeatedly without anyone checking the original).
- AI-generated content (where the model has learned to generate chains of citations that look plausible).
How to detect circular reporting:
- Follow the citation chain backward.
- Identify whether sources are independent or whether they share a common origin.
- Check whether the original source actually exists and supports the claim.
- Look for a break in the chain—a point where the claim originated from a non-authoritative or unsupported source.
7. CLAIM Walkthrough: Citation Verification and Tracing
- Definition: Confirm that the cited source exists and that all bibliographic metadata are correct.
- Guiding questions:
- Does the source exist in a library database, publisher site, or search index?
- Are the author, title, journal, date, volume, and pages correct?
- Is the DOI active and pointing to the correct source?
- Example: Verify “Chen, S., & Lee, J. (2021). Omega-3 and cardiovascular risk. Journal of Clinical Nutrition, 45(3), 210-218.” Search the journal’s website and database to confirm existence.
- Common mistake: Assuming that a citation is correct because it looks plausible. Always verify.
- Definition: Follow the claim, quotation, or statistic back to its ultimate origin.
- Guiding questions:
- Is this source the original or is it derivative?
- What is the earliest source for this claim?
- Has the claim changed as it moved between sources?
- Example: The 25% claim in the news article traced back to Chen et al. But Chen et al. may have derived the 25% from an earlier meta-analysis. Trace further.
- Common mistake: Stopping at the first source you find without checking if it is the original.
- Definition: Identify whether a claim has been laundered through a chain of citations that do not independently support it.
- Guiding questions:
- Do the sources all trace back to the same original?
- Does the original source actually support the claim?
- Are there any breaks in the chain where the claim is unsupported?
- Example: If the Chen study does not exist, or if it does exist but does not report a 25% reduction, the claim has been laundered.
- Common mistake: Counting multiple citations as evidence without checking their independence or support.
8. Verification-Process Walkthrough
This walkthrough demonstrates how to verify a citation, trace it to the original, and detect laundering. We use the opening scenario: the claim that “omega-3 supplementation reduces the risk of cardiovascular disease by 25%” attributed to “Chen et al. (2021), Journal of Clinical Nutrition.”
- Stage 1: Verify metadata. Search for the citation in PubMed, Google Scholar, and the journal’s website. Search term: “Chen” AND “Journal of Clinical Nutrition” AND “2021” AND “omega-3.” Check the author list, title, and abstract. If the search returns no results, the citation may be fabricated. If it returns results, verify the title, volume, and pages match. Result: The search yields no results. The citation appears to be fabricated.
- Stage 2: Read laterally. Even if the citation doesn’t exist, you can read laterally about the authors or the journal. Search for “Chen” and “Journal of Clinical Nutrition” to see if other articles by that author exist. Check the journal’s reputation. Also search for “omega-3 cardiovascular disease 25%” to see if the claim appears elsewhere. Result: You find that the 25% figure is widely cited on supplement company websites and blogs, but none of them link to a credible original study. This suggests the claim may be a marketing myth.
- Stage 3: Trace the claim backward. Even if the Chen citation is fake, the 25% figure might have originated elsewhere. Search for “omega-3 25% cardiovascular” in academic databases. Look for the earliest mention of this figure. Result: You find a 2018 meta-analysis that reported a 15% reduction, not 25%. The 25% figure appears to be an exaggeration that has been laundered through supplement marketing.
- Stage 4: Map the provenance. Document the chain: 2018 meta-analysis (15% reduction, primary source) → 2019 blog post (“omega-3s may reduce risk,” summary) → 2020 supplement company (“up to 25% reduction,” exaggeration) → 2021 fake study attribution (“Chen et al. found 25%,” fabricated) → 2022–present: repeated across websites (laundering). Conclusion: The claim has been laundered. The original evidence does not support the 25% figure. The citation is fabricated.
- Stage 5: Document and correct.
- Corrected wording: “A 2018 meta-analysis found a 15% reduction in cardiovascular risk associated with omega-3 supplementation; however, the widely cited 25% figure is not supported by the evidence and appears to have originated from exaggerated marketing claims.”
- Uncertainty: The evidence for omega-3s and CVD is mixed; some studies show no effect.
- Escalation: If the claim is used in a health decision, consult a qualified health professional.
9. Worked Example: Tracing a Misattributed Quotation
Claim: “Albert Einstein once said, ‘The definition of insanity is doing the same thing over and over and expecting different results.’”
Step 1: Verify the quotation
Search for the exact wording in quotation marks. Results show many websites, but none provide a source. Some attribute it to Einstein, others to various anonymous sources.
Result: No primary source for Einstein saying this phrase.
Step 2: Read laterally about the quotation
Search for “Einstein insanity quote” and look for discussions about its origins. You find that fact-checking websites (e.g., Quote Investigator) have traced it back to a 1981 book, and that there is no record of Einstein saying it. Some sources attribute it to a Narcotics Anonymous pamphlet.
Step 3: Trace to the original
Quote Investigator traces the phrase to a 1981 book by Rita Mae Brown, and then to a 1983 pamphlet from Narcotics Anonymous. There is no evidence Einstein ever said it.
Step 4: Detect laundering
The quotation has been laundered: it was misattributed to Einstein, then repeated across countless websites, blogs, and social media posts. The laundering gives it an appearance of authority (because Einstein is a famous figure), but the attribution is false.
Step 5: Corrected wording
Corrected: “The phrase ‘the definition of insanity is doing the same thing over and over and expecting different results’ is often misattributed to Albert Einstein. It appears to have originated in a 1981 book by Rita Mae Brown and was later used in a Narcotics Anonymous pamphlet. There is no evidence Einstein ever said it.”
10. Guided Learning Activities
- Purpose: Practice verifying citation metadata.
- Learning outcome: Verify reference metadata.
- Estimated time: 15 minutes.
You are given the following citation: “Jones, P., & Williams, R. (2020). The effects of mindfulness on stress. Journal of Psychology and Health, 35(2), 112-125. DOI: 10.1007/s10900-020-00854-7.”
Using the supplied search results below, determine whether the citation exists and whether the metadata are correct.
Supplied search results:
- Journal of Psychology and Health exists (ISSN: 0887-0446).
- Volume 35, issue 2 exists and contains articles on pages 98-150.
- No article by Jones and Williams is listed in that issue.
- The DOI 10.1007/s10900-020-00854-7 resolves to a different article: “Smith, A. (2020). Anxiety and coping mechanisms. Journal of Psychology and Health, 35(2), 98-110.”
What do you conclude?
Show activity feedback
- Authors: Jones and Williams do not appear in this issue.
- Title: The title “The effects of mindfulness on stress” does not match the article listed.
- Pages: The pages (112-125) are within the issue range, but the actual article occupies 98-110.
- DOI: The DOI points to a different article entirely.
Conclusion: The citation is fabricated or misattributed. The DOI and journal issue exist, but the article does not match the citation. The metadata are incorrect.
- Purpose: Practice lateral reading.
- Learning outcome: Conduct lateral reading.
- Estimated time: 15 minutes.
You find an article on a website called “The Health Guardian” that makes bold claims about a new treatment. The site looks professional and has a .org domain. Using lateral reading, what would you do to investigate this source? List at least four steps.
Show activity feedback
Possible steps:
- Search for “The Health Guardian” to see if other sources have evaluated it (e.g., Wikipedia, fact-checking sites).
- Search for the author(s) of the article to see their credentials and other publications.
- Search for the organisation behind the website—is it a real non-profit, a think tank, or a marketing operation?
- Check if the article’s claims are supported by peer-reviewed research; search for the treatment in academic databases.
- Check if the website has a funding or conflict-of-interest disclosure.
- Look for evidence of retractions or corrections.
- Purpose: Trace a claim through its citation chain.
- Learning outcome: Trace to the original source.
- Estimated time: 20 minutes.
You have this chain:
- AI output: “According to a 2022 study, 75% of students prefer online learning.”
- Source cited by AI: “Johnson, L. (2022). Student preferences in higher education. Journal of Educational Technology, 28(4), 12-29.”
- What Johnson’s article actually says: “In a survey of 200 students at one university, 65% reported a preference for online learning over in-person instruction.”
- Johnson’s source: Johnson cites a 2021 report from the university’s institutional research office that surveyed 150 students and found 60% preferred online.
Trace the claim and identify the transformations. What is the original source? What has changed?
Show activity feedback
- Original source: 2021 institutional report: 60% preference among 150 students at one university.
- Johnson’s article: 65% preference among 200 students (different sample, slightly different percentage).
- AI output: 75% preference among “students” (population generalised, percentage changed).
Transformations: The sample size changed, the percentage increased, and the population became generalised. The original context (one university, specific time) is lost. The AI output is an overgeneralisation.
- Purpose: Map a citation-laundering chain.
- Learning outcome: Detect citation laundering.
- Estimated time: 20 minutes.
You have the following sources:
- Source 1: A press release from a supplement company: “New study shows our product reduces pain by 50%.”
- Source 2: A health blog that reports “supplement reduces pain by 50%” and cites the press release.
- Source 3: A news article that says “study shows supplement reduces pain” and cites the blog (without naming the company).
- Source 4: A Wikipedia editor adds the claim to a page about the supplement, citing the news article.
Create a provenance map showing the chain. Identify where the claim originated and where it was laundered.
Show activity feedback
Provenance map:
- Origin: Press release (company claim, unsupported by peer-reviewed study).
- Step 1: Blog repeats the claim, citing press release (no independent verification).
- Step 2: News article repeats the claim, citing blog (no independent verification).
- Step 3: Wikipedia cites news article (now the claim appears in a “reliable” source, but the original is an unsupported press release).
Laundering: Each step moves the claim further from its origin and gives it an appearance of authority. The original is an unsupported commercial claim. The Wikipedia entry gives it the appearance of established fact.
- Purpose: Verify a quotation and a statistic.
- Learning outcome: Verify quotations and statistics.
- Estimated time: 15 minutes.
You read: “As Dr. Martin Luther King Jr. said in his 1963 ‘I Have a Dream’ speech, ‘I have a dream that my four little children will one day live in a nation where they will not be judged by the color of their skin but by the content of their character.’ More than 200,000 people attended the speech, making it one of the largest political rallies in US history.”
Verify both the quotation and the statistic.
Show activity feedback
- Quotation: The quotation is accurate. It appears in the transcript of the speech, and the wording is correct.
- Statistic: The crowd estimate is approximately 200,000–250,000. This is widely reported and generally accepted. The claim “one of the largest political rallies” is subjective but defensible.
- Conclusion: The quotation is verified; the statistic is broadly accurate.
11. Exercises
Complete the following exercises. For each, show your reasoning and final answer, then check against the feedback.
- Difficulty: Introductory
- Estimated time: ~10 min
- Learning outcome: Verify reference metadata.
You are given the citation: “Taylor, M. (2021). Artificial intelligence and society. AI Review, 12(3), 45-67.” You check the journal’s website and find that volume 12, issue 3 contains articles on pages 45-67, but the authors are Davis and Kumar, not Taylor. What do you conclude?
Show answer
- Authors: Taylor is not the author of the article in that issue.
- Title: The title does not match the article listed.
- Conclusion: The citation is misattributed or fabricated. The journal and issue exist, but the article does not match the metadata.
- Difficulty: Introductory
- Estimated time: ~10 min
- Learning outcome: Identify original and derivative sources.
You have: (A) a scientific article published in a peer-reviewed journal; (B) a news article that summarises the scientific article; (C) a blog post that quotes the news article; (D) a social media post that shares a link to the blog post. Which is the original source? Which are derivative?
Show answer
- Original: A (the peer-reviewed article).
- Derivative: B (news article), C (blog post), D (social media post).
- Explanation: B, C, and D all depend on A. They do not provide independent evidence.
- Difficulty: Intermediate
- Estimated time: ~12 min
- Learning outcome: Conduct lateral reading.
You find a website “Global Health Today” that claims a new supplement cures a disease. The website has no author listed and no contact information. What lateral reading steps would you take?
Show answer
Possible steps:
- Search for “Global Health Today” to see if it is known as a reliable source.
- Search for the supplement and the disease to see if credible sources (e.g., WHO, CDC, NIH) support the claim.
- Check the website’s domain registration (whois) to see who owns it.
- Look for the website’s funding or sponsorship information.
- Check fact-checking sites for any information about the claim.
- Difficulty: Intermediate
- Estimated time: ~10 min
- Learning outcome: Trace a quotation.
You read: “Mahatma Gandhi said, ‘Be the change that you wish to see in the world.’” What would you do to verify this quotation?
Show answer
- Search for the exact wording in quotation marks.
- Search for the phrase with “Gandhi” and check the sources.
- Check quote databases for the source.
- Search for scholarly discussions of the quote’s origin.
Result (for reference): This quote is widely attributed to Gandhi, but its exact origin is uncertain. Many sources say it was paraphrased from a longer statement. Verification reveals that the quote is not directly traceable to a single authenticated source.
- Difficulty: Intermediate
- Estimated time: ~10 min
- Learning outcome: Trace a statistic.
You read: “80% of students report feeling stressed.” You want to verify this. What would you need to trace? What questions would you ask?
Show answer
Questions to ask:
- What is the source of the statistic?
- What population was surveyed? (All students? University students? High school students? A specific country?)
- What was the sample size?
- How was “stress” defined and measured?
- When was the survey conducted?
- Was the survey methodologically sound?
- What is the margin of error?
Trace back to the original survey report or article to find these details.
- Difficulty: Intermediate
- Estimated time: ~12 min
- Learning outcome: Identify citation laundering.
You have a Wikipedia article that states “A 2021 study found that X causes Y.” The Wikipedia citation links to a news article, which cites a press release, which cites a study that is not peer-reviewed and not accessible. Is this citation laundering? Explain.
Show answer
Yes, this is citation laundering.
- The Wikipedia article cites a news article (derivative).
- The news article cites a press release (derivative).
- The press release cites a study that is not peer-reviewed and not accessible (unverified original).
- There is no direct link to a credible original source.
- The claim has been laundered through a chain of citations, giving it an appearance of authority.
- Difficulty: Advanced
- Estimated time: ~15 min
- Learning outcome: Create a provenance map.
You are verifying the claim: “A 2022 study by the University of Toronto found that remote work increases productivity by 15%.” You find:
- A 2021 survey by a consulting firm found a 10% increase in productivity.
- A 2022 blog post cited the 2021 survey and changed the number to 12%.
- A 2023 news article said “remote work boosts productivity by 15%” and cited the blog post.
- A 2023 social media post repeated the 15% figure.
Create a provenance map. Where did the 15% figure come from? Was it laundered?
Show answer
Provenance map:
- Original: 2021 consulting firm survey (10% productivity increase).
- Step 1: 2022 blog post changes the number to 12% (exaggeration).
- Step 2: 2023 news article reports 15% (further exaggeration) and cites the blog post.
- Step 3: 2023 social media repeats 15%.
Laundering: Yes. The original 10% figure from an unverified consulting survey was exaggerated to 12%, then to 15%, and then attributed to a study that doesn’t exist. The chain gives the claim an appearance of authority, but the original source is not credible.
12. Homework Problems
You have the citation: “Brown, A. (2019). The future of work. Journal of Business, 45(2), 88-95.” The journal’s website shows volume 45, issue 2, pages 88-95, but the article is by “Brown, A.” and is titled “The future of work.” What do you conclude? If the metadata are correct, what is your next step?
Show solution design
- Metadata: The metadata appear correct—the author, title, journal, year, volume, issue, and pages match.
- Next step: Read the article (abstract or full text) to verify that it actually supports the claim for which it is cited. Metadata correctness does not prove claim support.
You are verifying a claim about historical events. You have: (1) a contemporary diary entry, (2) a history textbook written 50 years later, (3) a modern podcast about the event. Which is the original source? Why?
Show answer
Answer: The contemporary diary entry is the original (primary) source. The textbook and podcast are derivative (secondary and tertiary). The diary is closest to the event in time and is a direct record.
You find a website called “Climate Truth” that claims climate change is a hoax. The website cites a few articles from obscure journals. Describe your lateral reading strategy for investigating this source.
Show answer
- Search for “Climate Truth” to see its reputation.
- Search for the authors of the cited articles to see their credentials.
- Check the cited journals to see if they are legitimate (e.g., in Scopus, Web of Science).
- Search for the claims on fact-checking sites (e.g., Snopes, Climate Feedback).
- Look for funding or organisational affiliations of the website.
You read: “Nelson Mandela said, ‘Education is the most powerful weapon which you can use to change the world.’” How would you verify this quotation? What would you need to find?
Show answer
- Search for the exact wording in quotation marks.
- Search for the source (speech, interview, book) where Mandela said this.
- Check reliable sources (e.g., the Nelson Mandela Foundation).
- Look for the context—was it from a specific speech?
Result (for reference): This quote is widely attributed to Mandela and appears in many sources. It is often traced to a 1996 speech.
You read: “The unemployment rate in Canada is 6.5%.” What would you need to verify this statistic? What is the authoritative source?
Show answer
- Authoritative source: Statistics Canada (Labour Force Survey).
- Need to check: The current month’s data, the definition of unemployment, the sample size, the margin of error, and whether the 6.5% is seasonally adjusted.
- If the date of the claim is not current, you need to check the date of the statistic.
You have the following chain: A press release from a university → A news article → A blog post → A Wikipedia article. The Wikipedia article cites the blog post. The original press release claims a 50% reduction in disease risk, but the press release does not cite a peer-reviewed study. Is this laundering? Explain.
Show answer
Yes, this is laundering.
- The original press release is not a peer-reviewed source.
- The news article repeats the claim without independent verification.
- The blog post repeats it again.
- The Wikipedia article uses the blog post as a citation, giving it the appearance of legitimacy.
- The claim has been laundered through a chain that obscures its non-peer-reviewed origin.
You find that Source A cites Source B, Source B cites Source C, and Source C cites Source A. No source provides original evidence. What is this called? What should you conclude?
Show answer
Answer: This is circular reporting or a citation loop. The sources are not independent; they depend on each other. There is no original evidence. The claim is unsupported. You should not use any of these sources to support the claim.
You are tracing a claim about the health benefits of green tea. You find: (1) a 2010 meta-analysis, (2) a 2015 news article citing the meta-analysis, (3) a 2018 blog post citing the news article, (4) a 2020 supplement company website citing the blog post and adding a “50% reduction in cancer risk” claim that does not appear in the meta-analysis. Create a provenance map and identify where the laundering occurred.
Show solution design
Provenance map:
- Original: 2010 meta-analysis (credible source).
- Step 1: 2015 news article summarises the meta-analysis (derivative).
- Step 2: 2018 blog post cites the news article (derivative).
- Step 3: 2020 supplement company website cites the blog post and adds “50% reduction in cancer risk” (exaggeration, laundering).
Laundering occurred: The supplement company introduced a 50% claim that does not appear in the original meta-analysis. The chain of citations gives the claim an appearance of authority, but the original evidence does not support the 50% claim.
You have a direct quotation from a scientific paper: “In our sample of 500 participants, we observed a 20% increase in test scores.” You also have a blog post that says “a study found a significant increase in test scores.” Which is direct evidence? Which is a summary? Why does this matter?
Show answer
- Direct evidence: The quotation from the scientific paper (it provides the exact data and context).
- Summary: The blog post (it does not provide the exact figure or the sample size).
- Why this matters: The summary may omit important context, such as the sample size, population, limitations, and the exact measure. The direct evidence allows you to verify the claim precisely.
Find a claim in your field (or use the supplied omega-3 scenario) that is repeated across multiple sources. Trace the claim back to its original source. Create a provenance map showing the chain of citations. Identify whether the claim has been laundered.
Show solution design
- Identify a claim that appears in multiple sources.
- List the sources (at least 3) in chronological order.
- For each source, note the date, type (primary, secondary, tertiary), and whether it supports the claim.
- Trace the claim backward to the earliest source.
- Check if the earliest source actually supports the claim.
- Determine if the claim has been laundered (changed, exaggerated, or misattributed).
- Create a provenance map showing the chain.
13. Mini-Project: Citation and Provenance Audit
- Purpose: Apply lateral reading, citation tracing, and laundering detection skills to a synthetic claim chain.
- Learning outcomes: Verify metadata, trace claims, detect laundering, create provenance maps, and produce corrected wording.
- Estimated effort: 120–180 minutes.
Supplied claim chain
You are auditing the following claim chain:
- Initial claim (AI output): “According to a 2022 study by Dr. Michael Torres in the Journal of Environmental Science, climate change is responsible for a 30% increase in extreme weather events.”
- Source 1 (AI’s citation): “Torres, M. (2022). Climate change and extreme weather. Journal of Environmental Science, 44(2), 112-130.”
- Source 2 (news article): A 2023 news article in the Daily Climate that says “a 2022 study found a 30% increase in extreme weather” and cites “Torres (2022).”
- Source 3 (blog post): A 2023 blog post on “EcoWatch” that says “climate change causes 30% more extreme weather” and cites the Daily Climate article.
- Source 4 (environmental report): A 2024 report from the International Climate Institute that cites “Torres (2022)” and claims “25% increase in extreme weather.”
- Source 5 (Wikipedia): A Wikipedia article on climate change that states “studies suggest a 30% increase” and cites the Daily Climate article.
Required deliverable
Complete the following sections for the supplied claim chain (or use your own).
- Initial claim: State the initial claim.
- AI-generated wording: Copy the AI wording.
- Citation supplied: Record the citation.
- Reference metadata: Verify the metadata against the journal. Is the source real? (Use the supplied search results below.)
- Source existence: Does the source exist?
- Source authenticity: Is the source authentic (does it match the metadata)?
- Source type: Classify each source (primary, secondary, tertiary).
- Author and publisher: Investigate the author and publisher.
- Original or derivative: Which source is original?
- Claim-evidence alignment: Does the evidence support the claim?
- Original source traced: Trace to the original.
- Intermediate sources: List the intermediate sources.
- Source independence: Are any independent?
- Circular citation: Is there circular citation?
- Citation laundering: Is there laundering?
- Quotation accuracy: Check the quotation.
- Statistic accuracy: Check the statistic.
- Context preserved or lost: Was context lost?
- Provenance chain: Create a provenance map.
- Evidence gap: What is missing?
- Corrected citation or source: What is the correct source?
- Corrected wording: Rewrite the claim.
- Remaining uncertainty: What is uncertain?
- Escalation or stop condition: Should this be escalated?
- Preliminary conclusion: What is your final assessment?
Supplied search results (for Source 1):
- The Journal of Environmental Science exists (ISSN: 1234-5678).
- Volume 44, issue 2 exists and covers pages 100-150.
- There is no article by Torres in that issue.
- There is an article by “Torres, M.” in volume 43, issue 4 (2021), titled “Climate projections and uncertainty.”
- No article in any volume of the journal reports a 30% increase in extreme weather events.
Assessment criteria
- Accurate metadata verification
- Correct source type classification
- Thorough lateral reading analysis
- Complete provenance map
- Accurate laundering detection
- Well-justified corrected wording
- Clear documentation of uncertainty
- Professional, evidence-based conclusion
Show staged hints
Hint 1: Start with Source 1. Verify the metadata for Source 1 (Torres, 2022). Use the supplied search results. Does the article exist? Is the author correct? Is the journal correct? Is the year correct?
Hint 2: Trace the statistic. Where did the “30% increase” figure originate? Does it appear in any of the sources? If not, where did it come from? The earliest source with the 30% figure may be the news article or the AI output itself.
Hint 3: Build the provenance map. Map the chain: AI output → Source 1 (fabricated?) → Source 2 (news) → Source 3 (blog) → Source 4 (report) → Source 5 (Wikipedia). Identify which sources are derivative and which are independent.
Show illustrative partial example (different claim chain)
Claim: “A 2020 study found that 40% of adults have sleep disorders.”
Chain: AI output → 2021 news article → 2020 press release → original study (not peer-reviewed).
Metadata verification: The original study is a small survey (n=100) from a non-academic source. It is not peer-reviewed.
Provenance map:
- Original: 2020 survey (n=100, non-peer-reviewed, 40% reported sleep issues).
- Press release: “40% of adults have sleep disorders” (overgeneralisation).
- News article: Repeats the press release.
- AI output: Adds “study found” and implies it is a reliable source.
Laundering: Yes. The claim was laundered from a non-peer-reviewed survey to a “study” with implied authority. The population was generalised from the survey sample to all adults.
Corrected wording: “A 2020 survey of 100 adults found that 40% of respondents reported sleep issues; however, the survey was small, non-peer-reviewed, and not representative of the general population. This figure should not be generalised without further evidence.”
14. Self-Check Quiz
Question 1: What is the first step in verifying a citation?
Show answer and explanation
Answer: b.
Before reading the source, you must confirm that the source exists and that the metadata (author, title, journal, date, etc.) are correct. If the metadata are wrong or fabricated, the source cannot be verified.
Question 2: What is lateral reading?
Show answer and explanation
Answer: b.
Lateral reading involves going outside the source to see what others say about it.
Question 3: When tracing a quotation, which of the following is not a necessary step?
Show answer and explanation
Answer: c.
Quote-aggregation websites are not reliable primary sources. You must trace the quotation to its original context.
Question 4: What is citation laundering?
Show answer and explanation
Answer: b.
Citation laundering creates an illusion of authority through repeated citations that ultimately trace back to an unsupported or fabricated source.
Question 5: If a source’s DOI resolves to a different article than the one cited, what does this indicate?
Show answer and explanation
Answer: b.
If the DOI points to a different article, the citation is incorrect. The source may be fabricated, or the metadata may be wrong.
Question 6: What is circular reporting?
Show answer and explanation
Answer: a.
Circular reporting creates a closed loop where sources depend on each other without any original evidence.
Question 7: When tracing a statistic, which of the following is essential?
Show answer and explanation
Answer: a.
To verify a statistic, you must trace it to its original source and check the measure, population, denominator, and calculation.
Question 8: In a provenance map, what is the purpose of documenting the chain of sources?
Show answer and explanation
Answer: b.
A provenance map shows how a claim has moved through sources and whether it has changed, been exaggerated, or been misattributed.
Question 9: If an AI generates a citation that seems plausible but you cannot find it in any database, what should you do?
Show answer and explanation
Answer: b.
If you cannot verify a citation, you must not use it. Document it as unverified and escalate if necessary.
15. Assignment Connections
In Assignment 5, you will need to verify every citation in an AI-generated passage. This includes:
- Verifying reference metadata for every cited source.
- Reading laterally to assess the authority and reputation of each source.
- Tracing quotations, statistics, and findings to their original sources.
- Detecting citation laundering and circular reporting.
- Documenting provenance chains and evidence gaps.
As you complete this lesson, retain:
- Your ability to verify metadata using databases and search tools.
- Your practice in lateral reading and source investigation.
- Your skill in tracing claims to original sources.
- Your ability to detect citation laundering and circular reporting.
- Your documentation of provenance and evidence gaps.
In Assignment 5, you will produce a citation audit and provenance map for each claim. The mini-project in this lesson directly prepares you for that task.
Reflect on the following questions in your Assignment 1 journal.
- What did lateral reading reveal about a source you previously trusted? Describe an experience where leaving the source changed your evaluation.
- How does citation laundering affect trust in sources? Provide an example of a claim that appears to have been laundered.
- Why is it important to trace claims to their original sources? What can be lost when you rely only on derivative sources?
- How has your approach to verifying citations changed after this lesson? What will you do differently when you encounter a citation?
Write in your own voice; there are no model responses.
Show reflection planning checklist
- I have described a specific lateral reading experience.
- I have explained citation laundering with an example.
- I have discussed the importance of tracing claims to original sources.
- I have identified a change in my citation verification practice.
- I have connected my reflection to my own field or experience.
16. Key Takeaways
- Metadata verification is the first step: confirm that the source exists and the citation is correct.
- Lateral reading involves leaving the source to investigate its author, publisher, reputation, and context.
- Original-source tracing is essential for verifying claims, quotations, and statistics.
- Citation laundering occurs when a claim gains authority through a chain of citations that do not support it.
- Circular reporting is a form of laundering where sources cite each other in a loop.
- Quotations must be verified against the original context—they are often misattributed or taken out of context.
- Quantitative claims require careful verification of measures, denominators, populations, and calculations.
- Provenance maps help you visualise the chain of sources and identify transformations or laundering.
- A fabricated citation is one that does not exist; a misattributed citation exists but does not support the claim.
17. Key Terms
- Reference metadata
- The bibliographic elements that describe a source (author, title, journal, date, volume, pages, DOI, etc.).
- Lateral reading
- Leaving a source to investigate its author, publisher, reputation, and external assessments.
- Original-source tracing
- Following a claim, quotation, or statistic back to its ultimate origin.
- Citation laundering
- A claim gaining authority through a chain of citations that do not support it.
- Circular reporting
- Sources citing each other in a loop, creating the illusion of corroboration.
- Quotation verification
- Confirming the exact wording, speaker, source, date, context, and authenticity of a quotation.
- Quantitative-claim verification
- Verifying numerical claims, including measures, denominators, populations, and calculations.
- Provenance map
- A visual representation of the chain of sources and transformations of a claim.
- Fabricated citation
- A citation that does not exist; the author, title, and other metadata are invented.
- Misattributed citation
- A citation that exists but does not support the claim, or is attributed to the wrong source.
- Press release repetition
- Multiple sources repeating the same press release, creating an illusion of wide support.
- AI-generated repetition
- AI-generated content that is repeatedly published and later cited as if it came from a human expert.
18. Looking Ahead: Lesson 11.4
You have now learned to verify citations, trace claims, read laterally, and detect citation laundering. In Lesson 11.4, you will integrate all of these skills with the TRACE and CLAIM frameworks to:
- Apply the complete CLAIM framework to verify claims.
- Integrate TRACE error analysis with CLAIM verification.
- Produce corrected wording and communicate remaining uncertainty.
- Exercise epistemic responsibility and establish escalation, correction, withdrawal, and stop conditions.
- Complete the Unit 11 cumulative project and prepare for Assignment 5.
Lesson 11.4 is the culmination of the Unit 11 verification methodology. After completing it, you will be ready to apply these skills to any claim you encounter—whether generated by AI, found in the media, or encountered in your professional practice.
19. Sources & Further Reading
- Primary textbook: Chapter 8: Verification, Evidence, and Epistemic Responsibility (COMP 3270 course text). [PAGE NUMBER TO BE CONFIRMED]
- Supporting: Appendix C: Verification Toolkit (COMP 3270).
- Supporting: Appendix A: Generative AI Glossary (COMP 3270).
- Caulfield, M. (2017). Web Literacy for Student Fact-Checkers.
- Stanford History Education Group. (2016). “Evaluating Information: The Cornerstone of Civic Online Reasoning.”
- University of Toronto Libraries. (2024). “Evaluating Sources.”
- Quote Investigator. (n.d.). “Investigating the Origins of Popular Quotes.”
Additional authoritative sources may be provided by your instructor. If you are unable to locate a source, use the supplied-source alternatives in the activities.