Learn it → Apply it → Save as Skill
If someone wants you to believe a claim, producing the evidence for it is their job, not yours. A claim offered with no evidence can be set aside with no evidence of your own, until real evidence shows up.
The Dr. Wallace’s Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
When a claim is made without evidence, you are not obligated to disprove it before setting it aside. The burden of proof sits with whoever is making the claim. Confident delivery, repetition and volume are not evidence.
Important: Hitchens’s Razor cuts on the second bar, not the first. Tone, confidence and repetition should never substitute for the second bar moving.
The model is most useful whenever someone, including an AI system, asserts something as fact and you have to decide how much weight to give it.
The razor is a short filter you run before a claim is allowed to change your beliefs or your decisions.
Spot a claim being presented as settled fact, not as a question or a guess.
Work out who owes the evidence. It is whoever is making the claim, never the person hearing it.
Look for checkable evidence: data, sources, a verifiable mechanism. Confidence and repetition are not evidence.
If evidence exists, weigh it on its merits. If it does not, you may set the claim aside provisionally, without having to disprove it.
Open a use case to see what the razor changes in practice.
Ask what specific, checkable result backs the number: whose output, measured how, over what period, compared to what baseline. Until that evidence arrives, treat the figure as marketing language, not a fact to plan around.
Treat an unverified AI claim exactly like an unverified human one. A confident, well-written sentence is not evidence of accuracy. Ask the system for a checkable source, or verify independently, before the claim is allowed into a real decision.
Trace the claim toward its original source rather than the account that amplified it loudest. If no primary source turns up, the claim has not yet earned belief, even if it is widely shared.
Ask for the comparable deals they are referencing. An assertion about market conditions, offered with no comparable data, is a negotiating tactic until proven otherwise.
Ask for the artifact behind the claim: the dashboard, the report, a name who can confirm it. A strong claim with no verifiable trail is a flag to probe further, not a reason to disqualify outright.
Urgency and alarm are persuasion tactics, not evidence. Pause before forwarding it further, and look for the claim's original, checkable source first.
This is an illustrative example designed to teach the method.
Expand for the origin of the phrase and how it relates to a much older principle.
Christopher Hitchens (1949–2011) was a British-American journalist, author and polemicist. The line most associated with his name, “What can be asserted without evidence can also be dismissed without evidence,” appears in his 2007 book God Is Not Great: How Religion Poisons Everything. It was intended there as a tool for evaluating unfalsifiable religious claims, but it generalizes cleanly to claims of any kind.
No, and Hitchens did not claim otherwise. The underlying principle, that the burden of proof falls on the person making a claim rather than on the person doubting it, is an old one in both law and philosophy, sometimes traced to the Latin maxim that the burden of proof lies with the party who asserts, not the party who denies. Hitchens's contribution was a sharp, quotable restatement that made the principle easy to carry into everyday arguments.
All three are “razors”: simple rules for cutting away explanations that do not deserve consideration. Occam's Razor compares competing explanations and favors the simpler one. Hanlon's Razor cautions against assuming malice when incompetence explains the facts just as well. Hitchens's Razor is different from both: it does not compare explanations at all. It asks only whether a single claim has earned enough evidence to be believed in the first place.
The razor lets you withhold belief. It does not let you assert the opposite with the same confidence the original claim used.
Demanding airtight evidence from people you disagree with while accepting friendly claims on faith is motivated skepticism, not the razor.
If a well-evidenced field just wasn't cited on the spot, check its track record before treating the claim as baseless.
The useful move is to ask for evidence and stay open to it, not to declare victory and stop listening.
A person's own first-hand account of what they felt or experienced is not the kind of claim this tool was built to referee.
This applies to people and to AI systems equally. Fluent, assured phrasing is not a substitute for a checkable source.
the claim is routine, low-stakes and made inside a relationship that already runs on trust. Not every remark from a colleague or friend needs an evidentiary audit.
a decision cannot wait for full evidence and the downside of inaction is severe. The razor governs belief; it is not a substitute for deciding under real uncertainty.
someone is describing their own subjective experience, especially around harm or distress. Treating that as an unproven assertion to be picked apart can do real damage.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply Hitchens's Razor, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its answer.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
No. Most everyday claims do not need laboratory-grade proof, only evidence proportioned to the stakes. A low-stakes, plausible claim can clear the bar with a quick, checkable source. A claim that will drive a major decision deserves correspondingly stronger evidence before you act on it.
Both make the same point from a different angle. Bertrand Russell's teapot analogy argues that the burden of disproof should never fall on the skeptic for an unfalsifiable claim. Carl Sagan's phrase, “extraordinary claims require extraordinary evidence,” scales the required evidence to how much the claim would overturn if true. Hitchens's Razor is the short, general-purpose version of the same family of ideas.
Yes, if it is misapplied to a person's first-hand account of their own experience, particularly around harm, discrimination or distress. The razor was built for third-party factual assertions, not as a tool for cross-examining someone's lived experience. Using it that way is a misuse of the model, not a correct application of it.