// learn.shawon.ch / research-methods / spotting-weak-studies STUDY GUIDE
← Research methods

Research methods · Putting it to work

Spotting weak studies & hype

Try this first

You have 60 seconds and one bold health claim in front of you: "clinically proven to boost energy 40%." Before reading on, write down what you'd check — and the order you'd check it in. Which question would you ask first?

Here's the freeing part: you almost never need to read the study. Most over-hyped claims collapse in the first ten seconds of a fixed checklist, the same way a flimsy chair gives out the moment you lean on the first leg. You don't inspect all four legs — you sit, it fails, you move on. The skill isn't deep reading. It's running a short triage in a fixed order and letting the claim fall at the first hurdle it can't clear.

The one idea

Run a fixed checklist, in order, and stop at the first failure. Humans or animals? Controlled or just observed? Hard outcome or a surrogate? Absolute numbers or a relative percent? Big and long enough? Independent or paid for? Most hype dies on the first one or two questions — you never reach the rest.

The six-question filter

Order matters. Cheaper, more decisive questions go first, so the weakest claims fail fastest. Each question has a "no" side that should make you stop and stay skeptical, and a "yes" side that lets the claim advance to the next rung. Survive all six and you've earned the right to actually read the paper.

The triage, top to bottom
Ask"No" means
Humans?Mice, cells in a dish, or a worm — stop. It may never translate.
Controlled / randomized?Just watched a group over time — can't separate cause from "healthy-user" habits.
Hard outcome?A surrogate (a blood marker, a scan, self-reported "energy") stood in for what you care about.
Absolute effect shown?Only a relative percent ("40% more") with no baseline to anchor it.
Big and long enough?A handful of people for a few weeks — noise, not signal.
Independent funding?The seller ran or paid for the study that conveniently flatters the product.
Humans? NO → mouse / dish — stop Controlled? NO → just observed — wary Hard outcome? NO → surrogate — wary Absolute #? NO → relative % — wary Independent? NO → seller paid — wary PASSES — READ DEEPER
Each "no" peels off into skepticism; only a clean run earns a real read.

Work one, then finish one

Worked: a tweet says "clinically proven to boost energy 40%!" Run the filter. Humans? The brand's own page links a study in mice — already wobbling; if you push past it, keep going. Controlled? No control group; everyone took the product and was asked how they felt — fails. Hard outcome? "Energy" is self-reported on a questionnaire, a textbook surrogate primed for the placebo effect — fails. Absolute? "40%" is a relative number with no baseline; 40% more than what? — fails. By now you've spent maybe twenty seconds and the claim has fallen at three separate hurdles. You never needed the funding question. Dismiss and move on.

Your turn: triage the headline "mouse study shows compound reverses aging." Where does it fall, and how far do you have to read? (First hurdle: it's a mouse study, so "humans?" already fails. Stack on a surrogate outcome — a biomarker of aging, not lifespan — and a giant leap from one rodent result to "reverses aging" in people. You stop at question one; the rest is hype piled on a finding that may never translate.)

Why this matters

This is the everyday tool — the one you actually reach for. Say you're staring at a $45 bottle of a "longevity" supplement that a podcaster swears by and a Reddit thread is buzzing about. You don't have time to read trials, and you shouldn't have to. You run the filter on the loudest claim: human trial or mouse study? randomized or just a survey? did anyone live longer, or did one blood marker move? a real effect size or a scary-sounding percent? funded by the company selling it? Sixty seconds later you either have a reason to look harder or you've saved yourself $45 and a daily pill that does nothing. Multiply that over every ad, thread, and podcast claim you meet in a year, and a fast, fixed filter is worth more than any single fact you could memorize.

Recall check · no peeking

  1. List the fast study-level red flags in the order you'd check them, and why that order saves time.
  2. Name the person-and-promotion red flags — the tells in who is making the claim, not the data.
  3. What is the single error every one of these weak claims shares?

Explain it back

In one line, name the universal tell: what does every over-hyped claim do with the evidence sitting underneath it?

Learn · Shawon Chowdhury · a study guide, kept rough on purpose