Science is not perfect. It is not a religion. It is the best error-correction system humans have built for separating what feels true from what survives reality.
01 — The problem
The honest reason most people feel lost about health, nutrition, supplements, religion, politics and everything in between is not that they lack information. It is that nobody taught them how to weigh it.
Open any feed and you will be told that a peptide will rebuild your knees, that a podcast guest has cracked the code on longevity, that one nutrient is the secret, that another is poison, that a supplement is “science-backed,” that an institution is corrupt, and that a single study changes everything. Some of these claims are real. Most are exaggerated. A few are dangerous.
Underneath the noise, the situation is genuinely hard:
The problem is not that people lack information. The problem is that people lack a method for judging information.
This page is the method — the one I would have wanted at twenty. It is not a manifesto for blind trust in “the science,” and it is not a vibes-based rejection of institutions either. It is a working toolkit for adults who have to make decisions without certainty.
02 — What it actually is
When most people say “the science,” they mean “the current best summary of what has survived testing so far.” The summary is not the thing. The process underneath it is.
The scientific method is a disciplined loop — a way of asking questions that forces ideas to put themselves at risk before they can be called knowledge.
Notice something in the world that wants explaining.
Frame it precisely enough that an answer can be wrong.
A specific guess that predicts what should happen.
Decide what is being changed, measured and held constant.
Set up a procedure that could prove the hypothesis wrong.
Run the test honestly, without nudging the result.
Use statistics carefully. Distinguish noise from signal.
Publish, share methods and data, and let others attack the work.
A different team should be able to find the same result.
Hold conclusions provisionally. Revise as evidence accumulates.
Science is not truth itself. Science is a disciplined process for reducing error.
03 — Why it changes
People often say “they keep changing their minds, so it must be fake.” The opposite is closer to the truth. A field that never updates is a field that has stopped looking.
When recommendations shift, it is usually because at least one of these has happened:
A worked example: nutrition. The mid-20th-century food pyramid mixed real evidence with public-policy choices, agricultural lobbying and a heavy bet against dietary fat that turned out to be too simple. Decades later, large reviews and reassessments have moved the consensus toward whole foods, fibre and unsaturated fats, with caution around ultra-processed foods. That movement is not science failing. It is science doing the only thing it is allowed to do: respond to better evidence.
Two things can be true at the same time. Nutrition science is genuinely difficult — humans are messy, diets are complex, long-term studies are expensive, and food guidelines are partly public policy. And the modern recommendations are still better grounded than what they replaced. Hard fields move slowly. Slow movement is not the same as no movement.
04 — When it’s wrong
Honest defenders of science don’t pretend it has a clean record. The argument is not that science is always right. The argument is that science has the only built-in mechanism for catching its own mistakes.
In the 1840s, Ignaz Semmelweis showed that doctors washing their hands between the dissection room and the maternity ward dramatically cut deaths from childbed fever. He had no germ theory to explain why it worked. The medical establishment rejected him for decades. The data was right; the community wasn’t ready. Eventually, germ theory caught up and his protocol became standard.
Doll and Hill’s 1950 study and the long British Doctors Study built up a clear link between smoking and lung cancer over years. The tobacco industry funded counter-research, manufactured doubt, and delayed regulation. Evidence eventually overwhelmed the noise; the 1964 U.S. Surgeon General’s report consolidated the case. It took decades because the system that funded the wrong answer was powerful, not because the science couldn’t see.
Starting around 2010, large efforts in psychology and biomedicine showed that many flagship results could not be reproduced by independent teams. The response was not denial; it was reform: pre-registration, registered reports, sharing data, larger samples, and stricter statistics. The crisis itself is the error-correction mechanism doing its job out loud.
A Cochrane methodology review of pharmaceutical and device studies found that industry-sponsored research is more likely to report results favourable to the sponsor than independent research, even when methodological quality looks similar. A separate JAMA Internal Medicine analysis traced how the sugar industry, in the 1960s, shaped the framing of dietary fat versus sugar in heart-disease research. Funding doesn’t make a finding wrong, but it shifts the prior.
Science can be wrong at one point in time. The scientific process has a built-in mechanism to expose that wrongness. Most belief systems do not.
05 — Why it’s still the best
Science is not in competition with religion, tradition or intuition for the same job. They answer different questions. But when the question is “is this claim about the physical world likely to be true?”, science wins on the only metric that matters — how it treats its own mistakes.
Science (at its best) asks
Most other belief systems ask
This is not arrogance about science. Religion, philosophy, ethics and lived experience answer questions science cannot — how to live, what to value, how to grieve, what we owe each other. But for empirical claims about the physical world, the comparison is one-sided.
Most belief systems protect their conclusions. Science, at its best, puts its conclusions at risk.
06 — Evidence ladder
The phrase “science says” can mean anything from a single mouse study to a decade of replicated human trials. Always ask which level of evidence is being used. The ladder below is the rough hierarchy clinicians, regulators and methodologists actually use.
A single person’s experience. Useful for generating questions, almost worthless for proving general claims.
Example: “It worked for me.”
A plausible biological or physical reason something should work. Necessary, but not enough — bodies don’t care about elegant mechanisms.
Example: “This compound binds to that receptor, therefore…”
Useful for safety screens and biology, but mice are not small humans. Most things that work in mice fail in human trials.
Example: “In rats, the molecule extended lifespan.”
Tracks people in the real world. Can show patterns and correlations, but confounding (other things that travel with the variable) is a major issue.
Example: “People who eat X tend to live longer.”
Better than observation because the researcher controls who gets the intervention. But small samples produce noisy results — underpowered trials over- and under-state effects routinely.
Example: “In 30 volunteers, a marker improved.”
A big sample, randomly assigned, ideally double-blinded and pre-registered. The closest most fields ever get to clean cause-and-effect in humans.
Example: a multi-thousand-person blinded trial of a drug versus placebo.
Different teams, different settings, similar findings. Replication is the part most consumer headlines skip and the part that matters most.
Example: three independent labs find the same effect within a similar range.
Pools many studies under explicit rules. Strong only when the underlying studies are strong; garbage in, garbage out is real here too.
Example: a Cochrane review on intervention X.
A formal recommendation built on transparent evidence review using systems like GRADE. Strongest when the process is clean — weaker when politics, funding or rushed timelines shape it.
Example: a national clinical guideline citing graded evidence.
Watch for the level mismatch
Most viral “science-backed” claims are sitting at L2 or L3 while implying L7 or L8. The single most useful question you can ask is: at what rung is this evidence, really?
07 — A practical checklist
Run any strong claim through this list before letting it shape what you do, what you buy or what you tell someone you love. None of these questions require a PhD.
Was this in humans, or in animals or cells?
How many people were studied?
Was there a control group?
Was it randomised?
Was it blinded? Single, double, or not at all?
Is the result clinically meaningful, or only statistically significant?
Has it been replicated by independent researchers?
Who funded the study?
Are the people promoting it selling something?
What is the downside if the claim turns out to be wrong?
Is the claim being made stronger than the evidence supports?
Is the outcome a real-world endpoint, or just a biomarker?
Is this a broad consensus, or one exciting paper?
You will not get clean answers to all thirteen questions on every claim. The point is not perfection. The point is that running the questions, even partially, blocks most of the bad claims that fly past ordinary attention.
08 — A grounded look
Wellness culture is the highest-volume area where the words “science-backed” are used and the lowest-density area where actual high-rung evidence exists. Treat it accordingly.
A few honest points to hold in mind:
Australia’s TGA, the U.S. FDA and other regulators have publicly warned about unapproved peptides being marketed for cosmetic, performance or anti-ageing use. That isn’t the regulators being slow. That is the regulators noticing what the wellness market is doing and saying: we have not seen the evidence you’re implying.
The question is not “does it sound scientific?” The question is: how strong is the evidence, how safe is it, who benefits from the belief, and what is the cost if I’m wrong?
09 — Belief framework
The goal is not to never be wrong. The goal is to be wrong less often, and to update faster when you are.
Hold beliefs with confidence levels, not certainty. “Probably true,” “leans true,” “genuinely uncertain,” “leans false” are real categories.
Separate what is known, what is likely, what is unknown and what is speculative. Almost no one does this in conversation, and it is the single biggest upgrade.
Update when better evidence arrives, not when louder voices arrive.
Be more skeptical when someone is selling something — a product, a worldview, a side.
Be more careful when the downside is large or irreversible. The bar for “probably fine” should scale with what happens if you’re wrong.
Prefer claims that have survived criticism over claims that have only survived applause.
Do not confuse confidence with truth. Loud certainty is a vibe, not evidence.
Do not confuse skepticism with intelligence. Reflexive doubt of everything is just contrarianism with better branding.
Do not reject all institutions just because some institutions failed. The replacement for a flawed institution is a better-functioning one, not no institution.
10 — The thesis
Science is not perfect because humans are not perfect. But it remains the best system we have because it does not ask us to believe forever. It asks us to test, criticise, replicate and update.
The goal is not to worship science. The goal is to become harder to fool.
11 — Further reading
Primary sources where possible. If a claim on this page felt strong, the source it leans on is here. Read these; they are better than any single summary.