False Causality: How the Brain Invents Causes That Aren't There

B.F. Skinner's pigeons developed rituals. Given a box where food arrived at random intervals, completely independently of the pigeon's behavior, the pigeons began to elaborate their own quasi-rituals: repeated head-bobs, counter-clockwise rotations, specific pecks at corners of the box. Each pigeon developed a different set.

The rituals were the result of a simple mechanism: the brain connected the last observed behavior to the subsequent positive outcome. Food arrived; the pigeon happened to be turning left; turning left gets associated with food. The association is spurious. The food was always going to arrive. The turning had no causal role. But the pattern-matching machinery didn't care about causation — it cared about conjunction.

Nassim Taleb's point in Fooled by Randomness: traders develop identical rituals. And so does everyone operating in random-adjacent environments.

Skinner's Pigeons in the Trading Pit

Taleb confesses his own version. On a profitable morning, he took a specific taxi route that happened to go through a particularly pleasant corner of the city. The trade worked. The next day he felt a pull to take the same route. Not a strong pull — he knew intellectually it was spurious. But the pull was there.

This is the mechanism: pair a random positive outcome with any salient antecedent, and the brain automatically forms an associative link. The pull to repeat the antecedent follows automatically. The pull is experienced as a preference, sometimes as a belief ("this route has better energy"), rarely as what it actually is — a Skinnerian association formed by one conjunction.

Professional traders name these rituals explicitly: the tie, the breakfast, the sequence of songs on the commute, the side of the chair. Survey trading desks and the diversity of apparently unrelated superstitions is remarkable. Different rituals per person, but the mechanism is universal.

The academic term is post hoc ergo propter hoc — after, therefore because of. It's the default causal inference engine in the human brain. You observe A followed by B; the brain concludes A caused B. The inference isn't warranted by the observation alone. But it's the inference that fires automatically, before deliberate reasoning can intervene.

Why the Bias Persists Under Knowledge

The unsettling thing about false causality is how little it responds to knowing about it. Taleb, the expert on this exact phenomenon, can't stop the pull to take the profitable route. The knowledge doesn't suppress the automatic association.

This is because the bias is System 1 — the fast, automatic, emotional system — while the knowledge is System 2. The two systems operate in parallel, with System 1 generating its response before System 2 can intervene. By the time the rational brain says "this is a spurious association," the emotional brain has already formed the preference.

The practical consequence: expert knowledge about cognitive biases doesn't reliably protect you from cognitive biases. Traders who study behavioral economics in depth still develop superstitions. Psychologists who teach about availability bias still over-weight vivid recent examples. The knowledge is in System 2; the bias is in System 1; the two rarely meet.

The Media Version: Finding Causes for Noise

The same mechanism operates at scale in financial media. Every day's market movement gets a cause: "stocks fell on interest rate fears," "markets rose on strong jobs data," "tech sold off on regulatory concerns."

These explanations are almost always false causality. Daily market movements are dominated by noise — random fluctuation around the underlying signal. Explaining the noise as caused by the day's news is post hoc ergo propter hoc at industrial scale. The news story happened; the market moved; the brain's narrative machinery connects them.

The explanation isn't just useless — it's actively misleading. It trains readers to believe that daily market movements have identifiable causes and that following the news gives predictive information about market direction. Neither is true. The noise is noise. The explanation is invented. And the trained belief in the explanation causes investors to react to noise as if it were signal.

What You Can Do About It

The false causality bias cannot be eliminated. The pattern-matching machinery is too automatic and too deeply wired. But its influence on actual decisions can be mitigated through structural approaches:

Pre-commit to decision rules. If the decision rule is stated before the outcome is observed, the post hoc association can't influence it. You can't change the rule after the fact based on a spurious association if the rule was documented before the fact.

Require minimum sample sizes. A single observation — even a profitable one, even a memorable one — is insufficient to establish a causal rule. Personal operating rules should require enough confirming observations (and controls for alternative explanations) before being adopted. "This worked once" is not a causal inference; "this worked in thirty independent trials across different conditions" is weaker than proof but much stronger than once.

Be especially suspicious of memorable associations. The brain is most likely to form spurious associations when the outcome is emotionally significant — a big win, a painful loss. The more memorable the outcome, the more likely the associated antecedent will form a false causal link. Which means the superstitions formed around the biggest wins and losses are exactly the ones to scrutinize most carefully.

Separate the ritual from the evaluation. If you find yourself developing rituals — specific sequences, specific items, specific behaviors before decisions — treat this as information about the bias, not information about causation. The ritual tells you your pattern-matching system fired. It doesn't tell you the antecedent caused the outcome.

Taleb's personal approach: allow the ticks to run (trying to suppress them directly is counterproductive) but explicitly don't act on them and don't let them influence the decision protocol. The Skinnerian pull is there; the behavior follows the pre-committed rule anyway.

For the full framework, read Living With Randomness.