Learning · Article

Appeal to Ignorance Fallacy: Examples and When No Evidence Counts

What the appeal to ignorance fallacy is, its two forms, everyday examples from work and health headlines, and two simple questions that tell you when a lack of evidence really does mean something.

Appeal to Ignorance Fallacy: Examples and When No Evidence Counts

Farida manages a small accounting office with nine people. Every night, a backup program copies the client files to an external drive and to the cloud. At the yearly IT review, she asks Oskar, who looks after the computers, whether they should test restoring a backup.

"No need," Oskar says. "It's been running for four years and there has never been a single sign of a problem. No errors, no warnings, nothing. If it were broken, we'd know."

Farida nods, then stops. In four years, nobody has ever tried to restore anything. The program could be copying empty folders every night, and the office would see exactly what it sees now: no errors, no warnings, nothing.

Oskar isn't being lazy. He's making one of the most natural mistakes in reasoning: treating "we have no evidence of a problem" as "we have evidence there is no problem." That move has a name, the appeal to ignorance. This guide covers what it is, its two forms, where you'll meet it, and the two questions that tell you when a lack of evidence really does mean something.

What is the appeal to ignorance fallacy?

The appeal to ignorance (also called the argument from ignorance, or in Latin argumentum ad ignorantiam) is concluding that a claim is true because it hasn't been proven false, or false because it hasn't been proven true. The missing evidence is treated as if it were evidence.

The Internet Encyclopedia of Philosophy puts it in two forms:

  1. "Not proven false, so it's true." "Nobody has shown this supplement is harmful, so it's safe."
  2. "Not proven true, so it's false." "Nobody has proven the new process saves time, so it doesn't."

The same entry adds the most useful line about it: the fallacy happens "in cases where absence of evidence is not good enough evidence of absence." In other words, a lack of evidence sometimes does tell you something. The skill is knowing when.

It's entry 23 in our guide to logical fallacies. This article goes deeper, because the hard part is the quiet version, like Oskar's, that sounds like good sense.

Where the name and the famous saying come from

The Latin name goes back to the philosopher John Locke. In An Essay Concerning Human Understanding (1690), Book IV, Chapter XVII, section 20, he described a debating move he called argumentum ad ignorantiam: demanding that your opponent either accept what you offer as proof or come up with a better one. Locke was describing a pressure tactic in argument, and the modern meaning has drifted from his. Today the name refers to the mistake of reading proof into the absence of proof.

The saying everyone quotes in reply, "absence of evidence is not evidence of absence," is often credited to the astronomer Carl Sagan, who used it in his 1977 book The Dragons of Eden, or to the cosmologist Martin Rees. According to Quote Investigator, it's much older. A near match appears in a talk by William Wright given in 1887, and the exact wording was in print by 1895, in The Glacialists' Magazine, attributed to a Mr. Dugald Bell. Sagan and Rees helped make it famous decades later.

Appeal to ignorance examples

You've heard most of these. In each one, a gap in knowledge is quietly promoted to a conclusion.

Where What they say Form What's actually known
Office IT "The backups have never shown an error, so they work." Not proven false, so true Nobody has tested a restore
Health "There's no proof this herbal remedy is unsafe." Not proven false, so true It may never have been properly studied
Work "Nobody has proven remote Fridays help, so they don't." Not proven true, so false Nobody measured it either way
Home "No one complained about the new curfew, so the kids are fine with it." Not proven false, so true Teenagers may not say so to their parents
Online "Scientists can't explain it, so it must be aliens." Not explained, so my explanation wins There's a gap, not an answer
Software "The tests all pass, so there are no bugs." Not proven false, so true The tests only check what someone thought to check

"Can't explain it, so it must be X" is a close cousin: a gap in one explanation is treated as support for another. A mystery is a reason to keep looking, not a vote for your favorite answer.

Debate it on Debatly“Is "we found no evidence of harm" a good enough reason to approve a new product, or should it have to prove it's safe?”

When absence of evidence IS evidence of absence

Here's the part most explanations skip, and it's the most practical. "Absence of evidence is not evidence of absence" is a great slogan, but taken literally it's wrong. Sometimes not finding something is excellent evidence that it isn't there.

You check the fridge for milk and find none. You conclude there's no milk. That isn't a fallacy, because you looked properly, and milk in a fridge is easy to see.

The Internet Encyclopedia of Philosophy gives the rule in one sentence: reasoning from a lack of proof is proper "only if the proof attempts were quite thorough, and it were the case that, if the being or object were to exist, then there would be a discoverable proof of this." That boils down to two questions you can ask about any "there's no evidence" claim.

Question 1: Did anyone actually look?

"No evidence" can mean two very different things: someone searched carefully and found nothing, or nobody searched at all. Oskar's backups fall in the second group. Four years of silence came from four years of not checking.

Question 2: If it were there, would we have found it?

Even a real search can miss things it wasn't built to catch. A smoke alarm that stays quiet is good evidence there's no fire in the kitchen, because it was designed to go off. Backup software that reports "no errors" only tells you it didn't notice an error. It can't tell you whether the copies are usable.

Put the two questions together and you get a simple grid.

We'd have found it if it were there We might have missed it
We looked carefully Strong evidence of absence (the empty fridge) Weak evidence (a small study that found nothing)
Nobody really looked No evidence either way (an alarm nobody switched on) No evidence at all (Oskar's backups)

Only the top left corner supports "it isn't there." Everything else supports, at most, "we don't know yet." And "we don't know yet" is a perfectly respectable answer.

"No evidence that..." in health and science headlines

The appeal to ignorance does its biggest damage in news about health. "Study finds no link between X and Y" can mean that a large, careful study ruled out any meaningful effect. Or it can mean that a small study couldn't detect one either way.

In 1995, the medical statisticians Douglas Altman and Martin Bland wrote a short note in the BMJ with the title "Absence of evidence is not evidence of absence." They pointed out that clinical trials which don't find a significant difference between treatments are often called "negative," and that this "wrongly implies that the study has shown that there is no difference, whereas usually all that has been shown is an absence of evidence of a difference." Part of the problem, they wrote, is that many trials are too small to detect real, worthwhile differences in the first place.

Their advice for readers is simple: when you're told "there is no evidence that A causes B," first ask whether that just means there is no information at all.

A few follow-up questions help:

  • How big was the study? A study of 40 people can easily miss an effect that a study of 40,000 would catch.
  • Was it designed to find this? A study about sleep may not have measured headaches carefully.
  • What's the range? Good reporting gives an estimate with a range of uncertainty. A wide range means "we don't know," not "there's nothing there."
Debate it on Debatly“Should news outlets be banned from writing "no link found" when a study was too small to detect one?”

Is "innocent until proven guilty" an appeal to ignorance?

No, and it's a helpful contrast. In a criminal trial, the defendant is presumed innocent, and the prosecution has to prove guilt beyond a reasonable doubt. If it fails, the defendant is acquitted.

That looks like "not proven guilty, so innocent," but it isn't a claim about what happened. It's a rule about what to do when the evidence falls short. Society has decided that convicting an innocent person is worse than letting a guilty one go, so the tie goes to the defendant. That's why the verdict is "not guilty" rather than "innocent": it says the burden wasn't met.

Everyday life has the same kind of rule. "We don't switch suppliers until someone shows the new one is better" is fine as a decision rule. It becomes a fallacy when it turns into "the new one isn't better." For who should carry the load, see our guide to the burden of proof.

The appeal to ignorance often travels with other errors. Telling them apart helps you pick the right question.

Mistake What happens The question that catches it
Appeal to ignorance Lack of proof is treated as proof of the opposite "Did anyone look, and would they have found it?"
Burden of proof shift Someone makes a claim and demands you disprove it "Whose claim is it?"
Survivorship bias Only the cases that made it are visible "Who isn't in this picture?"
Confirmation bias You search only where agreeing evidence would be "Where would I look if I wanted to prove myself wrong?"
Moving the goalposts Evidence arrives, so the standard is raised "What would change your mind? Let's agree now."

The burden shift and the appeal to ignorance are a familiar pair. First someone says "prove it isn't true," and then, when you can't, they treat that as a win.

How to respond without starting a fight

"That's an appeal to ignorance" is accurate and rarely helps. People hear "you're being illogical" and defend their position harder. Questions work better, because they invite the other person to check the gap themselves. (That's the idea behind Socratic questioning.)

A simple pattern: agree on what's known, ask how it was checked, offer a test.

What you hear What you can say
"There's never been a problem." "Good. Have we ever checked in a way that would show a problem?"
"There's no evidence it works." "Has anyone tested it? If not, we just don't know yet."
"Nobody has proven it's dangerous." "Has anyone looked carefully? What did they test?"
"Scientists can't explain it, so it must be..." "Maybe. What would we expect to see if that were true?"
"No one complained." "Would people have told us if they were unhappy?"

The third step matters most with someone you'll see tomorrow. "Let's find out" turns a standoff into a shared project.

Worked example: Farida's backups

Farida didn't tell Oskar he'd committed a fallacy. She told him he might well be right, and that she'd like to be able to say so with confidence if a client ever asked.

They ran the two questions together:

Question Their answer
Did anyone actually look? No. Nobody had ever restored a file from either backup.
If the backups were broken, would we have noticed? Probably not. The program reports copy errors, not whether the copies open.
What would a real check look like? Restore one client's full folder from each backup to a spare laptop and open ten files.
What would count as a pass? Every file opens and matches the original.
How often should we check? Once a quarter, with a different client folder each time.

The cloud backup passed on the first try. The external drive didn't. Its settings still pointed to a folder the office had stopped using two years earlier, so it had been faithfully backing up an old, almost empty directory every night. No errors, because nothing had gone wrong from the program's point of view.

Oskar fixed it in ten minutes and put a quarterly restore test in the shared calendar.

How to catch it in your own thinking

Everyone does this, especially about things they set up themselves or hope are fine.

  • Listen for "never" and "no sign." "It's never failed" and "there's no sign of trouble" are the moments to ask whether anyone has looked.
  • Name your detector. For any "nothing's wrong," ask what would have told you if something were. If the answer is "nothing," you have no evidence either way.
  • Say "we don't know yet." It feels weaker than a confident answer, but it's often the honest one, and it opens the door to a test.
  • Check both directions. It's easy to say "no proof it's harmful" about something you like and "no proof it works" about something you don't. Use the same standard for both.

How to practice this with AI

The appeal to ignorance is hard to catch in yourself because silence feels like reassurance. A debate makes the gap visible: one side has to say what the evidence actually is, and the other side asks how it was gathered. On Debatly, several AI models and characters argue different sides of your question over several rounds, you can step in at any point, and a judge gives a verdict with reasoning.

  • Put your "no sign of trouble" on trial. Type a belief into the debate room as a question, like "Are our office backups reliable, given they've never reported an error?" Ask one side to defend it and the other to ask how anyone would know.
  • Ask a skeptic to run the two questions. Have Viktor, the skeptic, go through "did anyone look?" and "would we have found it?" for your claim. It's a quick way to see whether you have evidence or just silence.
  • Try it on a headline. Paste a "no link found" headline and ask the debaters what the study would have needed to find a link if one existed.
  • Read the verdict. Judge Alex explains which arguments carried the round. Check whether the winning side had positive evidence or simply pointed at the other side's lack of it.

When Farida ran Oskar's backup question as a debate, the questioning side asked in its first reply when a file had last been restored. That was the question she brought to the next IT meeting.

Debate it on Debatly“Can two AI models agree on whether "we found nothing" means "there's nothing there"?”

An honest limit

AI models can't see your systems, run your restore test or read the full study behind a headline. They also make this very mistake: ask a chatbot whether something is safe, and it may answer "there is no evidence of harm" as if that settled it. And many models lean toward what you seem to believe already (see why AI agrees with you). Use the debate to find the gap and design the check. The answer still has to come from actually looking.

Ask how anyone would know

The appeal to ignorance turns "we haven't found it" into "it isn't there," or "we haven't proven it" into "it's false." The fix is two questions: did anyone actually look, and would they have found it if it were there? If both answers are yes, the silence means something. If not, the honest answer is "we don't know yet."

This week, pick one thing you're sure is fine because there's never been a problem, and ask how you'd know if there were. If it matters, run it as a debate on Debatly and let a skeptic look for the gap first. Give it a try. Plans are on the pricing page.

Debate it on Debatly

“Appeal to Ignorance Fallacy: Examples and When No Evidence Counts”

FAQ

It's claiming something is true because nobody has proven it false, or false because nobody has proven it true. The lack of evidence is treated as if it were evidence. "Nobody has shown it's unsafe, so it's safe" is a typical example.
"Our backups have never shown an error, so they must work," said about backups nobody has ever tried to restore. Another is "There's no proof the new process saves time, so it doesn't," when nobody has measured it. In both, the missing evidence comes from nobody checking.
Sometimes. If someone searched carefully, and the thing would have been found if it existed, then not finding it is good evidence it isn't there, like finding no milk in the fridge. If nobody looked, or the search couldn't have detected it, the absence tells you very little.
Shifting the burden of proof is demanding that someone else disprove your claim. Appeal to ignorance is treating the failure to disprove it, or to prove it, as if it settled the question. They often come together: first the burden is shifted, then the lack of disproof is counted as a win.
No. It's a rule about what a court should do when the evidence falls short, not a claim that the defendant definitely didn't do it. The prosecution must prove guilt, and if it can't, the defendant is acquitted. That's why the verdict is "not guilty" rather than "innocent."
It's often credited to Carl Sagan, who used it in his 1977 book The Dragons of Eden, or to Martin Rees. Quote Investigator traced it much earlier: a near match from 1887, and the exact wording in 1895, attributed to Dugald Bell.
#appeal-to-ignorance #logical-fallacies #critical-thinking #evidence #decision-making