Lecture: Evaluating Sources and Evidence
Lecture script · about 15–20 minutes spoken · Week 12
Opening
Picture three things you might scroll past in one evening. A photo of a flooded street, shared with the name of a nearby city. A sales page for a supplement that says "a study proves it works" and links nothing. And a chatbot reply that cites a journal article, with authors, a title, and a year, all in perfect format. Each one asks for your trust.
That request shows up everywhere. A nursing student meets it when a health post quotes numbers with no source. A business major meets it when a supplier says "a study proves it" about a product. A family member meets it when a frantic call sounds like a cousin asking for money. In every case, the useful question is the same: where did this claim come from, and has its source earned my trust?
Here's where we are. Week 1 taught you to find an argument, and Week 2 to put it in standard form. Week 7 gave you the appeal to authority and showed when it turns into a fallacy. Weeks 9, 10, and 11 taught you to judge a sample, a causal claim, and a number.
Every one of those tests assumed the evidence was real and reported fairly. This week checks that assumption. We'll cover six ideas. First come four questions for any source and the trace of a claim to where it started. Then we'll leave a page, check a photo or a quotation against its original, verify what AI hands you, and put it all together. Next week, in Science and Explanation, we'll ask what would count as a real test of a claim.
1. Four questions for any source
Let's start with two people telling you a storm is coming. One is a neighbor who noticed a dark cloud. The other is the national weather service, which tracks storms for a living. You'd trust both a little. You'd trust one of them a lot more. That difference has a name.
Credibility is how far a source deserves trust on a particular claim, judged by its relevant expertise in that field, its track record of accuracy, its independence from the outcome, and whether it shows the evidence behind the claim.
Notice that the definition holds four questions. First, relevant expertise: training and experience in the field a claim belongs to. A source can have real expertise in one field and none in the field of the claim at hand. Second, a track record: the source's history of accuracy on earlier claims, including whether it corrects its mistakes openly. Third, independence from the outcome: the source gains or loses nothing from whether you believe it. A seller praising its own product fails that one. Fourth, does the source show its evidence, or only assert the claim?
That's 4 questions, and you ask them about one claim at a time. A doctor is a strong source on a medicine and a weak one on a used car.
Now let's call back a passage from Week 7. It's illustrative, from the course's fallacy set. "A famous actor says this herbal supplement cures colds, so it works." Run the first question. Acting is not expertise in medicine, so the source fails the first question, and the passage commits the appeal to unqualified authority, ad verecundiam.
Set it beside a contrast passage that the course discusses and never keys. A national weather service forecasts a storm's path, and an event planner cites it as a reason to move an outdoor event indoors. Forecasting storms is the service's own field. Its forecasts are checked against the weather every day, in public. It gains nothing from where the event happens. And it shows maps and tracks. That source passes all 4 questions, on a question in its own field.
Pause and try this. The weather service passes all 4 questions. How many does the actor need to fail before the appeal becomes a fallacy?
Think about which question carries the actor's whole case.
Here's the answer. Failing 1 is enough here. The whole case rests on the actor's word, and the actor has no expertise in the field of the claim, so that single failure leaves the claim with no support.
One more thing to keep track of. This week keys no new passage from the fallacy set. The actor passage is 1 of the 19 keyed through Week 10, and the set holds 20, so 1 passage is still to come in a later week.
2. Trace the claim to where it started
Your roommate tells you that losing sleep hurts how you do on exams. You ask where they heard it, and they say a newsletter. Where did the newsletter get it? That second question starts a trace.
Two terms help here. A primary source is the original record of an event, study, or statement, such as the study a news story reports on or the full text a quotation was taken from. A secondary source is a source that reports, summarizes, or interprets a primary source, such as a news story about a study.
Tracing a claim means following a claim back through the sources that repeat it until you reach the primary source where it started, then checking that the original says what the retellings say.
Why bother? Because each retelling is a chance for the claim to change. A study about one group can turn into a headline about everyone. A finding that something "can" happen can turn into a claim that it "will".
Let's run it on the running question. Its strongest objection says that losing sleep can hurt exam performance. In our illustrative version, a campus newsletter repeats that claim and cites a news story, and the news story cites a study. So the chain has 3 links: the newsletter, the news story, and the study. You open the study itself and check three things: that it says what the retellings say, that it's about the same kind of people, and that it claims the effect with the same strength.
The same move works on a number. The cost premise, P4, says keeping the library open overnight for finals week costs about 1,600 dollars. Trace it back to its inputs. The library pays 2 staff for 8 hours a night for 5 nights at 20 dollars an hour. Multiply those together and you get 1,600 dollars. The figure matches the staffing numbers it rests on. That doesn't settle whether 1,600 dollars is a modest cost. That part is still a judgment the argument has to defend.
Pause and try this. In the sleep-claim chain of 3 links, how many are secondary sources, and how many are primary?
Sort each link by whether it retells or records.
Here's the answer. There are 2 secondary sources and 1 primary source. The newsletter and the news story retell the claim, and only the study is the original record.
Here's an honesty note. You won't always reach the primary source. A study can sit behind a paywall. When the trace stops, say where it stopped, and give the claim only the trust that the last source you reached has earned.
3. Leave the page before you read it
Picture a sales page for a supplement. Clean fonts, smiling faces, a seal that says "clinically studied", and the line "a study proves it works". It looks wonderful. Nothing on it tells you who wrote it, or whether the study exists.
Here's the careful move, and it can feel backward. Stop reading the page. Open new tabs. Before you trust an unfamiliar site, you leave it and check what other sources say about who runs it and how reliable it is.
An independent source is a source with no tie to the one being checked and no stake in the claim, such as a reference site or a news report about who runs a website.
So you search the site's name. Independent sources describe it as the seller's own store. Now run the questions. Independence? The seller profits if you believe the claim, so it fails. Evidence shown? "A study" is asserted, not linked or even named, so it fails. That's 2 of the 4 questions failed.
There's a real standard behind this. The Federal Trade Commission's guidance for advertisers says ads must be truthful and not deceptive, and advertisers must have evidence to back up their claims before the ad runs. Health and safety claims generally need competent and reliable scientific evidence. So "a study proves it" is a claim the seller should be able to back up.
Pause and try this. The sales page fails 2 of the 4 questions. Which 2, and why does its polished design not help?
Name the two questions before you go on.
Here's the answer. It fails independence and evidence shown. The seller profits from the claim, and the study is asserted, not shown, while a clean design answers none of the 4 questions.
A caution, though. Leaving the page tells you about the source, not the claim. You learn who's selling. You don't yet learn whether the product works.
4. Check the original context
A real photo can tell a false story. Go back to the flooded street. Suppose it's a real photo of a real flood, but the flood happened years ago, in a different city. Now a post shares it with a nearby city's name and the words "happening now". Every detail in the photo is true. The caption isn't.
The term is out of context: said of a real photo, video, or quotation shown apart from its original date, place, event, or surrounding words, so that it suggests something the original did not.
The check is the same for a photo and for a quotation. Find the original, and read what surrounds it. For the photo, you search for earlier copies. If the earliest copy names a different city and a different storm, the new caption moved the photo to a place and a time where it didn't happen.
Quotations can be cut the same way, and a cut can flip the meaning. Here's an illustrative case from the running question. A library staff member wrote, "Opening overnight would be a mistake unless the library adds staff for safety." A post quotes only "Opening overnight would be a mistake."
Pause and try this. The cut quotation keeps every word it uses exactly as written. Is it accurate, misleading, or both?
Compare what each version says about overnight hours.
Here's the answer. It's both. Cutting the condition turns conditional support into flat opposition, so the quotation is accurate word for word and misleading in meaning.
This links back to Week 7. If someone quotes the cut sentence and then argues against it, they're answering a weaker version of the staff member's view. That's how a straw man often starts. And it's why quotation marks tell you the words were said, but not what the whole sentence meant.
5. Verify what AI hands you
Back to the chatbot reply. It cites an article with two authors, a title, a journal, and a year. The format is perfect. You go looking for the article, and it isn't anywhere.
How does that happen? A chatbot produces fluent, confident text by predicting likely words. A likely-looking citation is not the same thing as a real one. Chatbots can produce text that includes invented statistics, studies, quotations, and citations.
So here are the terms. AI-generated misinformation is false or misleading content produced with an AI tool, such as a chatbot's invented statistic, study, quotation, or citation, or a cloned voice that imitates a real person. A fabricated citation is a reference to a book, article, or study that does not exist, even though its authors, title, and year are set out in correct format.
The check is plain work. You search the named journal's own site for the title, and nothing turns up. You search a library database for the title and the authors, and nothing turns up there either. The citation is fabricated, however real it looks.
Voices can be faked too. Voice cloning means using an AI tool to imitate a particular person's voice from a short recording of it. In a 2023 consumer alert, the Federal Trade Commission warned that scammers can clone a family member's voice from a short audio clip. Picture a call that sounds like a relative in trouble, asking for money right away. The alert's advice was to call the person back on a number you know is theirs before acting.
Pause and try this. Why does calling back on a number you already know work, when the voice on the line sounds exactly right?
Think about who controls each channel.
Here's the answer. A number you already trust is a channel the caller doesn't control. A cloned voice can copy how someone sounds, but it can't answer a phone the real person owns.
The same habit covers the chatbot. Its reply is a lead, never a source. A chatbot isn't always wrong, but you can't tell from the reply alone when it's right. So every figure, study, and quotation it offers stays unverified until you find the original.
6. Put it together as one source check
You now have five moves, and they fit together. A source check is the set of moves for weighing where a claim comes from. You ask the four credibility questions, trace the claim to its primary source, and see what independent sources say. Then you check the original context, and verify through a channel already trusted.
Here's a way to hold it. Think of a careful friend who's asked to vouch for a stranger. The friend asks what the stranger knows, what they've done before, what they stand to gain, and what they can show. Then the friend asks around. That's the source check, done on a person instead of a page.
Now notice what the check does not do. It tells you how much trust a source has earned on one claim. It doesn't prove the claim false. A source that fails one question can still be right, and a source that passes all 4 can still be wrong. The check is a tool for weighing evidence, not a machine that settles a case.
That matters on the running question. Tracing the sleep claim and the cost premise tells you how well those premises are supported. It doesn't decide whether the library should stay open around the clock. That question is still yours to argue, from either side.
Pause and try this. The actor fails the expertise question. Does that prove the herbal supplement doesn't cure colds?
Keep the source and the claim apart.
Here's the answer. No, it doesn't. A failed check means the actor gave you no good reason to believe the claim, and to learn whether the claim is true you need a source that does.
Closing
Let me pull it together in the order we went. Credibility is judged one claim at a time, with 4 questions: relevant expertise in the field, a track record, independence from the outcome, and evidence shown. The actor fails the first, and the weather service passes all 4 in its own field.
A primary source is the original record, and a secondary source retells it. Trace a claim to where it started, as we did with the sleep claim's chain of 3 links and the cost premise's 1,600 dollars. Leave an unfamiliar page and see what independent sources say about who runs it. Check a photo or a quotation against its original. Treat a chatbot's citation as unverified until you find the article, and verify a familiar voice by calling back on a number you already know. Put together, those moves are the source check, a tool for weighing trust, not a verdict on the claim.
Next in this module are the slide deck, the Readings and Resources page, the lecture tutorial, the practice set, and the quiz. Next week, in Science and Explanation, the source check meets the question of what evidence could show a claim to be wrong.
Built with The Course Maker · samples.thecoursemaker.com