Conclusion Shopping: Confirmation Bias, Echo Chambers and “Do Your Own Research”

Last updated: October 22, 2026

The “conclusion shopping” meme is a joke about confirmation bias: you pick the conclusion first and then browse the internet until you find evidence that fits it. Psychologists define confirmation bias, in Raymond Nickerson’s 1998 review in Review of General Psychology, as “the seeking or interpreting of evidence in ways that are partial to existing beliefs”. The classic demonstration is older. In 1960 the psychologist Peter Wason gave 29 students the numbers 2, 4, 6 and asked them to find the rule behind them; only 6 named the correct rule on their first try, because most kept testing number triples that could only confirm their guess. The meme compresses that finding into two whiteboard panels and adds the word “insane”, which Wason was too polite to use.

The meme is a two-panel reaction image. In both panels a man in a light blue office shirt stands beside a whiteboard on an easel, in front of window blinds. In the top panel he points a marker at the board and looks at the camera mid-sentence; the board reads “You are not ‘researching’ your conspiracy theory.” In the bottom panel he sits lower in the frame, hands out of shot, with a tight, knowing smile and a dark tie, and the board reads “You are ‘conclusion shopping’ your insane hypothesis.”

Six Out of Twenty-Nine: The Number Game Behind the Bias

Confirmation bias was first pinned down in a lab with a simple number puzzle, and the puzzle shows that the trouble starts with the questions people ask.

Wason’s set-up in the Quarterly Journal of Experimental Psychology (1960) was almost insultingly easy. Students saw the triple 2, 4, 6, proposed triples of their own, heard whether each one fitted the rule, and announced the rule when they felt sure. The rule was “three numbers in increasing order of magnitude”. A typical student guessed “goes up by two”, tested 8, 10, 12 and 20, 22, 24, heard “yes” every time and announced the guess with confidence. A triple like 1, 2, 3 would have broken the guess at once, because it gets a “yes” that “goes up by two” forbids, but people rarely try the triples they expect to fail. Wason wrote that many subjects “assumed that confirming evidence alone justified their conclusions”, and he read the result as a warning about “the dangers of induction by simple enumeration”.

The strongest objection comes from inside psychology. Joshua Klayman and Young-Won Ha argued in Psychological Review (1987) that much of what gets called confirmation bias is a “positive test strategy”: checking cases where you expect your idea to hold. In many realistic situations, they showed, this is a decent heuristic, and it only produces “systematic errors or inefficiencies” under certain conditions, such as Wason’s task, where the guessed rule sits neatly inside a much broader true rule. Conclusion shopping starts where that heuristic turns exclusive: you only look at cases that cannot prove you wrong.

Where Are the Ones Who Drowned? Bacon and Popper on Rigged Evidence

Francis Bacon described the bias in 1620, three centuries before the lab, and Karl Popper later turned its cure into a test for science.

In Book One of the Novum Organum, aphorism 46, Bacon says the human understanding, once it has adopted an opinion, “draws all things else to support and agree with it” (Fordham’s Internet Modern History Sourcebook, translator not named). His example is a temple wall covered with paintings of sailors who made vows to the gods and survived a shipwreck. A visitor is asked whether this proves the power of the gods, and he replies with a question: “but where are they painted that were drowned, after their vows?” Bacon adds that the mind is “more moved and excited by affirmatives than by negatives”. Wason opened his 1960 paper with Bacon’s phrase “induction by simple enumeration”, and Nickerson quotes aphorism 46 in his review. A conspiracy video works like the temple wall: survivors only.

Karl Popper made Bacon’s question into a criterion. According to the Stanford Encyclopedia of Philosophy, Popper held that confirming evidence is easy to find for nearly any theory, so it only counts when it comes from a risky prediction that could have failed. Einstein’s prediction that light bends near the sun could have been refuted by Eddington’s 1919 measurements. Psychoanalysis, by contrast, struck Popper as untestable: “There was no conceivable human behaviour which could contradict them” (Conjectures and Refutations, 1963, p. 37, as quoted by the SEP). A conspiracy theory that treats every missing piece of evidence as proof of the cover-up is built the same way.

Same Study, Opposite Verdicts: How Motivated Reasoning Works

Motivated reasoning is confirmation bias with a stake in the outcome: people judge evidence by whether it supports what they want to believe, and they usually do it while feeling perfectly fair.

In a well-known experiment published in the Journal of Personality and Social Psychology in 1979, Charles Lord, Lee Ross and Mark Lepper gave 48 students, half for and half against capital punishment, two invented studies on whether the death penalty deters crime. One study supported each side. Both groups found the study that agreed with them better designed and more convincing, picked holes in the other one, and reported that their own views had become more extreme. Later attempts to repeat the effect with larger samples, including a study in the British Journal of Political Science (2020) that adapted the death-penalty design, found people updating, if at all, toward the evidence rather than away from it (Guess and Coppock), so the hardening is the shakier part of the 1979 result.

The psychologist Ziva Kunda explained why this rarely feels like cheating. In “The Case for Motivated Reasoning” (Psychological Bulletin, 1990) she argued that people’s ability to reach the conclusions they want “is constrained by their ability to construct seemingly reasonable justifications”. We need reasons that would look good to a neutral observer, and so we keep an “illusion of objectivity”. Hugo Mercier and Dan Sperber go further in Behavioral and Brain Sciences (2011): reasoning evolved for arguing, so arguers are “not after the truth but after arguments supporting their views”. Their theory has a hopeful side: placed in a proper argumentative setting, people turn out to be “skilled arguers”.

Echo Chambers Are Built From Distrust

An echo chamber, in the philosopher C. Thi Nguyen’s sense, is a community that teaches its members to distrust every outside source, which is why, on Nguyen’s account, more facts may leave them unmoved.

Nguyen’s paper “Echo Chambers and Epistemic Bubbles” (Episteme, 2020) separates two things that public debate tends to lump together. Members of an epistemic bubble have simply missed information; show it to them and the bubble pops. Members of an echo chamber have been “brought to systematically distrust all outside sources”. On Nguyen’s account, a fact-check from a newspaper can then read as more evidence that the mainstream is lying, and leaving may take a “radical rebooting of one’s belief system”. Experiments rarely find corrections actually backfiring in the general public (Wood and Porter, Political Behavior, 52 issues, more than 10,100 subjects), so the effect Nguyen describes is a feature of tight communities more than of people in general. Whether algorithms build such chambers is a separate question, and the research suggests their effect is smaller than assumed, as our post on filter bubbles and the Gettier problem explains.

What fills such chambers travels fast. In a large study published in Science in 2018, Soroush Vosoughi, Deb Roy and Sinan Aral tracked about 126,000 rumour cascades on Twitter from 2006 to 2017 and found that false news was about 70 percent more likely to be retweeted than true news and reached 1,500 people about six times faster. The authors removed bots and found that their main conclusions did not change, so the spreading was mainly done by humans. Their best explanation was novelty: false stories tended to be more surprising.

“Do Your Own Research”, Says Kant (Sort Of)

“Do your own research” sounds like the Enlightenment’s own motto, and Kant would have liked the courage in it, but he thought hardly anyone gets there alone.

Kant’s essay “An Answer to the Question: What is Enlightenment?” (1784) opens with the slogan Sapere aude, which Mary Gregor translates as “Have courage to make use of your own understanding!” He mocks the comfort of having “a book that understands for me”. Read on, though, and the essay turns social. Kant writes that it is difficult for “any single individual to extricate himself from the minority” (Gregor’s word for Kant’s Unmündigkeit), while “that a public should enlighten itself is more possible”, and his remedy is the public use of reason: arguing as a scholar before the whole reading public, where others can check you.

The slogan now does work in medicine, investing and conspiracy theories, as Nathan Ballantyne, Jared Celniker and David Dunning show in “Do Your Own Research” (Social Epistemology, 2024). In 1985 John Hardwig put the uncomfortable point plainly in “Epistemic Dependence” (Journal of Philosophy). Everyone, experts included, believes far more than they could ever check, or in his words, “I believe more than I can become fully informed about”. Because the layperson is the “epistemic inferior of the expert”, he argues, “rationality sometimes consists in refusing to think for oneself” and deferring instead. That can sound like a licence for blind trust, but the evidence on lone searching is sobering. In five experiments reported in Nature (published online December 2023), Kevin Aslett and colleagues asked people to evaluate false news by searching online, and searching “actually increases the probability of believing them”, by about 20 percent. The effect was concentrated among people whose results came from low-quality sources, in “data voids” where little reliable material exists.

What Honest Research Looks Like

Doing your own research well means working out whom to trust and trying to prove yourself wrong, which is harder and less fun than collecting confirmations.

The old formula that knowledge is justified true belief leaves room for a lot of borrowed justification: few of us have tested a vaccine or measured a temperature record, so most of our justification runs through people and institutions. The skill this requires is judging sources. It helps to know what peer review does and does not guarantee, and whether a headline rests on one study or a meta-analysis. A fair test for any source: does it ever report results that cut against its own story?

The rest is Wason’s lesson applied to yourself. Before searching, write down what result would change your mind, then search for that result first, the way a good solver of the 2, 4, 6 puzzle tries 1, 2, 3. Look for the drowned sailors, the predictions that failed. And argue with someone who disagrees, because Mercier and Sperber’s work suggests that is where reasoning does its best work. Socrates built a whole method on that kind of questioning, and in 399 BC Athens put him on trial, a story told in our post on the trial and death of Socrates.

For another take on spotting bad arguments in the misinformation age, read C.M. Black’s post on why it’s a critical time for critical thinking.