Define the claim
What exactly is being asserted? About whom, where, compared with what, and over what period?
An aspirational field guide / inquiry 001
The good
scientist.
Not the person who is always right. The person who builds a way to discover when they are wrong.
A character ideal, not a biography. This page explores a way of investigating the world that anyone—including its author—can aspire to practise. Admiring the method is not evidence of living it.
03 RIVALS
∞ WAYS TO BE SURPRISED
Laboratory hours: whenever the claim becomes too comfortable.Edition 01 · 23 October 2026
Prologue / The door left open
The central virtue is not intelligence, credentials or a dramatic talent for scepticism. It is the willingness to let reality interrupt a beloved explanation.
An investigator can be imaginative without treating an interesting story as a discovered fact. They can hold convictions without turning them into untouchable property. They can be kind to people and difficult with claims, beginning with their own.
01 / THE METHOD
Build a route for correction
Give the idea
an exit.
A belief can guide a life. A hypothesis has another duty: to risk meeting observations that do not cooperate with it. The honest investigator learns to tell the two apart without pretending the distinction is always clean.
Name rivals
Write at least one plausible alternative before meeting the data. A single story has an unfair advantage.
Risk a prediction
Specify what you would expect if each explanation were true—and what would strain it.
Inspect the source
Ask how the evidence was produced, selected, measured and interpreted. Facts have supply chains.
Revise in public
Keep a record of what changed your view. A revised claim may be the most honest result.
A distinction worth keeping: This five-part sequence is an editorial practice framework, not a validated research method or a checklist sufficient for scientific rigor. Research designs vary by question and discipline. See the sources for the wider methodological conversation.
What makes an idea testable?
“Our customers liked the new display” could mean a dozen different things. Did they say they liked it? Did more customers stop? Did repeat purchases rise, relative to comparable days? A testable version identifies an observable outcome, an interval, a comparison, and a plausible route from cause to effect. “It was destined to work” protects itself from all possible results. It may be poetry; it is not a risk-bearing prediction.
Yet an honest investigator does not mistake every vague statement for nonsense. Ethical commitments, aesthetic judgments and questions of meaning may be important without reducing to a laboratory measurement. The point is to use the right tool for the kind of claim being made—not to declare the unmeasurable worthless.
02 / COMPETING TRADITIONS
A method with a history, not a priesthood
Science disagrees
about science.
The familiar image of a lone genius following a single universal method is a poor guide to how diverse disciplines actually investigate their questions.
Make claims vulnerable.
Karl Popper stressed falsifiability: a theory should exclude some possible observations. Surviving a hard test can strengthen its standing without proving it forever. He also knew that measurement and auxiliary assumptions complicate real-world refutation.
The tension: A single anomaly rarely tells us what, exactly, failed.
Look at the community.
Thomas Kuhn described periods of normal science, where practitioners work within shared problems and standards, alongside episodes of more radical conceptual change. What counts as an important problem is itself historically shaped.
The tension: Scientific change is neither a frictionless march of facts nor merely a popularity contest.
Degrees, not switches.
Bayesian epistemology offers a framework for revising degrees of belief in light of evidence. Prior expectations matter; evidence should be judged partly by how differently rival hypotheses predict it.
The tension: Priors, likelihoods and model assumptions can be disputed or poorly specified.
These are not three buttons with which to select one final philosophy of science. They are overlapping insights and live disagreements. Experimental controls, ethnography, archival work, simulation, observational studies, mathematical proof and replication play different roles in different fields. The question is not whether inquiry looks impressive, but whether its claims are accountable to an appropriate kind of scrutiny.
There is no uniform ritual that makes a conclusion true. There are practices that make error easier to notice.
That is an interpretation of methodological literature, not a quotation from Popper, Kuhn, or any research institution. The scientific enterprise is also social: cooperation, incentives, peer criticism, funding, access and institutional power shape which questions are asked and which findings travel. The ideal investigator must examine the machinery of knowledge as well as the data.
03 / THE INTERACTIVE CENTREPIECE
The claim-testing
laboratory.
Choose a fictional case, name your provisional explanation, inspect evidence cards and decide which to admit into your working notebook. Watch how a deliberately simple comparison changes. You can disagree with the instrument; its rules are exposed.
No simulated result is a real study, diagnostic test, probability estimate or judgment about you. This is an exercise in asking better questions—not a machine that discovers truth by adding coloured squares.
CASE FILE 01
Did the window display cause the sales dip?
A fictional neighbourhood shop changes a display. Sales fall the following week. The owner suspects the rearrangement.
01 · Rival explanations
Choose your opening theory02 · Evidence archive
0 of 6 admittedInspect ≠ endorse · Admit ≠ prove
03 · Working interpretation
Non-statistical / provisionalWhat the admitted cards suggest
Begin by admitting evidence. The ledger is empty.
A claim you can revise
Read the laboratory's rules and limitations
Every scenario, record and observation is invented. For each evidence card, the author assigns a qualitative direction to each of three explanations: strongly supports (+2), supports (+1), neutral (0), challenges (−1), or strongly challenges (−2). The display translates admitted-card directions into supportive/challenging clues. A provisional leader is reported only after at least three cards are admitted and one explanation leads the next by at least two illustrative sign-units; otherwise the result stays unresolved. Scores are ordinal design choices, not measured likelihoods, calibrated confidence, independent observations or a substitute for proper research design. No responses are sent to a server. Reset clears the current case; switching case files starts that case afresh. The only stored preference is whether you've dismissed the introduction.
04 / BEFORE THE HEADLINE
A base rate is not a footnote
The denominator
has a story.
A striking result means something different when the underlying event is rare. Attention tends to run toward the glowing finding; reasoning must travel back to the population from which it came.
Imagine a fictional quality-control check across 100 items. Ten actually contain a flaw. The check flags eight of those ten—but it also flags nine of the ninety good items. The device caught most of the real flaws, yet among the 17 flagged items, only eight actually contain a flaw.
Notice the shift: “the check catches eight of ten flawed items” and “eight of seventeen flagged items are genuinely flawed” describe different questions. If you only hear the first, you may mistake sensitivity for the chance that a particular alert is true.
These numbers are deliberately invented for illustration. In real investigations, the baseline frequency may be unknown, changing, or measured badly. A denominator can be politically, economically, or practically difficult to define. Ignoring it is not sophistication; neither is pretending to know it.
100 fictional items / one check
8 true flags 9 false flags 2 missed flaws
Among the 17 flagged items, 8 contain the flaw. The other 83 unflagged items comprise 81 correctly unflagged good items and 2 missed flaws.
Important: An illustration of conditional proportions—not an estimate of any real product, test or population.
The motive sneaks into the denominator.
If a report covers only successful trials, you need the number of unsuccessful trials. If a creator celebrates the handful of people who responded warmly, you need to know who saw the work and did not respond. If you conclude that a personal routine “always works,” ask which days were forgotten, and whether other changes happened at the same time. The monkey counts bananas that landed in the basket; the investigator remembers the ones that bounced into the bushes.
05 / THE SOCIAL TEST
Replication · transparency · modesty
One compelling result
is a beginning.
Knowledge grows more durable when someone else can retrace the steps, question the choices and attempt the inquiry again.
Reproducibility is not the same as replication.
In the terminology used by the US National Academies, reproducibility concerns obtaining consistent computational results using the same data and methods. Replicability concerns whether independent studies addressing the same scientific question produce consistent results using their own data. The latter is subtler: context, populations, interventions and random variation can matter.
Publishing code, measures, inclusion rules and reasonable methodological detail allows others to locate errors. Preregistration records intended hypotheses and analyses before examining results, making it harder to quietly pass exploration off as prediction. But no badge, registration link or open repository compensates for a poor research question or a weak design.
When a result fails to repeat.
Non-replication can signal error, low precision, an unstable effect, different populations, changed conditions or a theory that needs boundaries. It is not automatically fraud; neither should it be waved away with an improvised rescue story every time a cherished result falters.
Independent criticism is demanding precisely because the investigator owns neither the evidence nor its interpretation. Other people may discover what their excitement concealed. The mature response is not automatic surrender to every critic; it is a serious attempt to distinguish a good objection from a poor one.
Research grounding: The definitions above follow the National Academies' 2019 report. OSF guidance describes preregistration as a time-stamped record of study plans before collection or analysis. See verified sources for scope and detail.
06 / THE INVESTIGATOR'S BLIND SPOTS
A laboratory also has a mirror
Doubt can be
self-serving.
Scepticism feels admirable when it is directed at a rival's claim. The relevant test is whether the same standards survive contact with our own preferred explanation.
The confirmatory search
You ask questions designed to yield agreement, count every supportive anecdote and treat disconfirming evidence as an exception requiring no explanation.
Counterpractice: Before searching, write the strongest credible way your view could be wrong.
The endless sceptic
Every study has limitations, so you demand impossible certainty before accepting conclusions you dislike, while casually accepting claims you want to be true.
Counterpractice: Apply proportional standards to all sides. Action can be justified without omniscience.
The elegant rescue
A theory meets awkward evidence. Instead of revising, you add an invisible condition that explains exactly why the prediction failed—and predicts nothing new.
Counterpractice: Make the rescue assumption testable, or state that the model is now less secure.
Motivated reasoning has good manners.
People need not deliberately lie to themselves. Desire can affect what feels plausible, what sources receive attention, how strictly objections are interrogated and when inquiry is declared finished. The point is not that every strong belief reflects bias, or that emotion makes thinking invalid. Feeling something intensely may tell us what matters; it does not, by itself, settle what happened.
The opposite error is to turn a vocabulary of bias into a weapon against other people. “You are biased” is often just a decorated form of “I disagree.” A good investigation names the specific inferential move: missing comparison, selective sampling, confounding factor, overgeneralisation or untested assumption. It offers a route toward a better answer.
The investigator's courage is not in saying “I doubt.” It is in saying “Here is what would change my mind.”
07 / THE LIFE AROUND THE LAB
Counterarguments and identity tensions
Not every moment
needs a microscope.
The honest investigator is not an ideal of permanent interrogation. Every method has costs; every life contains decisions that cannot wait for perfect knowledge.
Investigator × Creator
Original work often begins in an imaginative leap, not a controlled test. The artist can make something before knowing its significance. The boundary comes when a personal metaphor is presented as a discovered law of human behaviour.
Investigator × Advocate
Advocacy names values and seeks change; inquiry tries to learn what is so. They can inform one another, but incentives diverge. A campaign's effectiveness does not establish the truth of every claim it makes.
Investigator × Friend
A friend in pain does not always need cross-examination. Listening, solidarity and practical help may come first. Respecting a person's experience need not mean endorsing every causal interpretation of it.
Investigator × Decision-maker
Leaders often act under deadlines with incomplete information. Better judgment may mean choosing a reversible course, stating assumptions and preparing to update—not delaying forever in the name of rigour.
Three serious objections to the ideal
“If everything is provisional, why believe anything?”
Provisional does not mean arbitrary. Some conclusions rest on extensive converging evidence; others barely survive first contact. The task is to match confidence to support, not to flatten everything into equal doubt.
“Won't constant checking make us indecisive?”
Yes, when the cost of inquiry exceeds its expected usefulness. Time, reversibility, stakes and available evidence matter. An inexpensive choice and a high-impact public claim deserve different levels of scrutiny.
“Can the powerful define what counts as evidence?”
Institutions can exclude voices, mismeasure experiences and reward convenient questions. Objectivity is not achieved by ignoring power; it improves when methods, assumptions, access and criticism become more transparent.
“Is humility just another performance?”
It can be. Repeatedly announcing uncertainty may earn admiration while leaving behaviour unchanged. Look for the less photogenic evidence: corrections issued, methods shared, rival ideas represented fairly and decisions reconsidered.
08 / A FIELD RUBRIC
Ideal → practice → evidence
The aspiration is not
the observation.
A person may admire a principle long before reliably enacting it. This rubric is a set of behavioural prompts, not a personality scale, credential or system for grading human worth.
| The ideal | A practice to attempt | Observable evidence, not a declaration |
|---|---|---|
| Testable thinking | Rewrite a cherished broad claim with conditions and a disconfirming observation. | The dated original claim and a prediction that could actually be wrong. |
| Rival-mindedness | Write two plausible alternative explanations before searching for support. | A notebook where alternatives were recorded before evidence was reviewed. |
| Proportional confidence | Separate observation, inference and uncertainty in a conversation or report. | A conclusion with explicit limitations and a narrower scope than its evidence cannot bear. |
| Fair source treatment | Inspect methodology and sample for supporting and inconvenient findings. | The same source-quality questions asked of both sides. |
| Revisability | Document one changed judgment, including why the old explanation was insufficient. | A visible correction, not merely a claim to be open-minded. |
| Humane restraint | Notice when the person in front of you needs care rather than a forensic interview. | A respectful choice to listen, alongside accurate boundaries about what is known. |
One good day does not establish a durable identity, and one mistake does not disqualify someone from pursuing it. The rubric asks for traces of conduct over time while recognising that opportunity, stakes, roles and constraints differ.
An ungated field notebook
Questions worth carrying away
- Which belief would I be most tempted to protect from a fair test?
- What competing explanation could a reasonable, informed person hold?
- Have I mistaken how often I noticed something for how often it happens?
- What is the cheapest reliable observation that might distinguish the alternatives?
- What would responsible action look like if the uncertainty remained?
- When have I actually revised a view—and what changed in my behaviour afterward?
09 / THE READING ROOM
Research context & intellectual debts
Follow the
paper trail.
The books and research below inform the concepts; they do not validate the fictional cases, this site's own framework or any claim about the author's personal scientific skill. Links were checked against public source pages during construction.
A broad survey of methods, demarcation debates and methodological pluralism.
Popper's falsifiability proposal and the limitations of simple boundaries.
Normal science, revolutions, scientific communities and progress.
Degrees of belief and normative approaches to updating with evidence.
Definitions, research reliability, complexity of replication and institutional responsibilities.
Practical documentation of time-stamped study plans and transparency measures.
RELATED / SAHILVERSE
Nearby questions, not supporting citations.
These public projects were inspected for topical relevance. Their arguments and fictional interactions are independent explorations, not experimental verification of this page.
The field experiment is not over
Keep one question
open to the world.
Choose one belief that matters but need not define you. Write what supports it, what rivals it and what could change it. Give the world a genuine opportunity to disagree. If your view survives, hold it proportionately. If it changes, you have not lost a self; you have learned something.
There is no promised transformation waiting at the bottom of this page. Just an invitation to investigate more honestly on an ordinary Tuesday.
The monkey has requested a control group.