Starvard University · Research
Find real answers instead of hot takes: how to ask good questions, search smart, judge sources, read studies, and check facts like a pro.
The whole course in brief
Twelve lessons, twelve big ideas.
The twelve lessons
Lesson 1
Most bad research starts with a search box and a vague itch. Good research starts with a sharp question written down. "Is coffee bad for me?" is mush. "Does drinking 3 cups of coffee a day raise heart disease risk in healthy adults?" is researchable. The sharper the question, the faster you find the answer, and the harder it is to fool yourself.
Write down what would change your mind before you start. If no evidence could change your answer, you are not researching, you are decorating a belief. Decide in advance what counts as an answer: a number, a mechanism, an expert consensus. Then go find it.
Lesson 2
Information degrades as it passes through hands. The study is primary. The news article about the study is secondary. The tweet about the article is tertiary, and by then the finding has often inverted. Go upstream whenever you can: read the actual paper, the actual data, the actual quote in full context.
This one habit puts you ahead of most of the internet. Headlines about studies are wrong or misleading with depressing regularity. The abstract of the paper takes two minutes to read and tells you what was actually found. Upstream is where the truth lives.
Lesson 3
Search engines reward precision. Use specific terms, not questions: "coffee cardiovascular risk meta-analysis" beats "is coffee bad for my heart." Put exact phrases in quotes. Use site filters (site:.edu, site:.gov) to restrict to serious sources. Add "review" or "meta-analysis" to find summaries of many studies instead of one cherry-picked paper.
Then read past the first page. The top results are optimized for clicks, not truth. Scholarly search tools exist for a reason: they index papers, not ads. And when you find one good paper, raid its reference list. One solid source leads to ten more. Research compounds.
Lesson 4
Every source gets four questions. Authority: who is this, and why would they know? A cardiologist on hearts, yes; a celebrity on vaccines, no. Evidence: do they show their work, with data and citations, or just assert? Motive: who pays them, and what do they gain if you believe this? Date: is this current, or has the science moved on?
No source passes all four perfectly. The point is not purity; it is calibration. A peer-reviewed paper with industry funding is still better than a blog with no funding, but you read it with one eyebrow up. Trust is not binary. It is a dial you set per source.
Lesson 5
Wikipedia is one of the best starting points ever built and one of the worst ending points. Its strength: thousands of editors, citations required, vandalism usually fixed fast. Its weakness: it reflects consensus, which can lag or be gamed on contested topics, and any given sentence might be wrong.
The pro move: read the article for orientation, then scroll to the references and click through to the primary sources. Wikipedia is a map, not the territory. Use it to find the territory, then go there yourself. Never cite it as your authority; cite what it cited.
Lesson 6
You do not need a statistics degree. Read the abstract, then check four things. Sample: how many subjects, and are they like you? Mice are not people; 20 undergrads are not the population. Design: randomized controlled trial beats observational study, which beats anecdote. Effect size: how big was the difference in real terms? Replication: has anyone else found this?
Watch for the greatest hits of bad science reporting: relative risk without absolute numbers, correlation sold as causation, and single studies treated as settled. One study is a data point. Science is the pattern across many. Wait for the pattern.
Lesson 7
A number without context is a weapon. "Crime doubled!" From what to what? Over what time? Compared to where? Always demand the absolute numbers, the timeframe, and the comparison. A 100% increase from 1 to 2 is trivia; from 10,000 to 20,000 is a crisis. Same percent, different universe.
Also demand base rates. A "99% accurate" test for a rare disease still produces mostly false positives, because the disease is rarer than the errors. And watch the denominator: rates per capita, per user, per dollar. Anyone can make a number scary by hiding what it is out of. Your job is to find the "out of."
Lesson 8
One source is a claim. Three independent sources agreeing is evidence. The key word is independent: three articles all citing the same press release are one source wearing a trench coat. Trace each claim back and make sure the lines do not converge on a single origin.
Triangulation also means mixing source types: a statistic, a document, and a human witness pointing the same way is far stronger than three statistics from the same dataset. When sources disagree, that is information too. Figure out why they differ before you pick a side.
Lesson 9
You are the easiest person to fool. Confirmation bias makes congenial evidence feel stronger and uncongenial evidence feel weaker, automatically. Motivated reasoning lets smart people reach whatever conclusion they wanted with full rigor. The bias is not a character flaw; it is factory equipment.
Countermeasures that actually work: steelman the other side (state their best case fairly before you attack it), ask "what would change my mind" and mean it, and seek out the strongest critics of your view, not the weakest. If your research never surprises you, you are not researching. You are shopping.
Lesson 10
"Trust the experts" is too simple, and "do your own research" on technical topics is usually a fantasy. The middle path: trust expert consensus where it exists, verified by credentials, track records, and institutions with reputations to lose. Consensus among climate scientists means something. One rogue doctor with a book to sell does not.
But experts have limits. They disagree at the frontier, they have biases and funding, and consensus has been wrong before. So trust with verification: look for convergence across independent experts, check whether they show their work, and hold stronger skepticism where the evidence is thin or the incentives are loud. Respect expertise without surrendering judgment.
Lesson 11
The quick protocol. One: read past the headline; most misinformation dies in paragraph two. Two: check the source's about page and track record. Three: search the core claim in quotes to see who else reported it and whether it traces to one origin. Four: check dedicated fact-checkers, but verify their reasoning too. Five: look for the primary source and read it.
Red flags that shortcut the whole process: no named author, no date, no sources, emotional manipulation ("share before they delete this"), and claims that perfectly confirm what you already believe. The more a claim flatters you, the harder you should check it.
Lesson 12
The final step of research is making it checkable. Cite your sources: links, titles, dates, enough detail that someone else can retrace your steps. Distinguish what you found from what you concluded. Admit uncertainty where it exists; "the evidence suggests" is stronger than false certainty, because it survives being wrong.
This is also how you become someone others trust. In a world of hot takes, the person who shows receipts stands out. Research is not just about being right. It is about being checkably right, so the truth does not depend on taking your word for it.
Graduation
Sharp questions, primary sources, smart searching, judged sources, skeptical study-reading, numbers with context, triangulation, checked bias, respected experts, fast fact-checks, and shown work. That is the full toolkit. The internet rewards speed and certainty. You now reward evidence and checkability. That puts you ahead of almost everyone arguing online.
Keep learning
This course is part of Starvard University, striving for greatness.
