Cognitive Biases: The Map of Systematic Errors

Cognitive biases are systematic patterns of deviation from rationality in human judgment. Unlike random errors, they are predictable, repeatable, and universal. They appear across all cultures, age groups, and education levels. A statistics professor makes the same biases as a layperson, though they may recognize them afterward.

The term was coined by Amos Tversky and Daniel Kahneman in the 1970s, during their research on heuristics. They realized that System 1’s mental shortcuts do not merely simplify reality; they distort it in specific, catalogable ways. To this day, more than 200 biases have been identified in the cognitive psychology literature.

The most troubling thing about cognitive biases is that knowing about them doesn’t eliminate them. Knowing that confirmation bias exists doesn’t stop you from seeking information that confirms your beliefs. Awareness of the error improves performance marginally, but doesn’t solve it. It’s like an optical illusion: even after someone explains the trick, you keep seeing the distorted image.

Confirmation bias is the tendency to seek, interpret, favor, and remember information that confirms our preexisting beliefs. It operates at three levels: selective search (we look for confirming evidence), biased interpretation (ambiguous data are read as confirmation), and selective memory (we remember better what confirmed our position).

From a neurological perspective, there's an elegant explanation. When we encounter information that confirms our beliefs, the dopaminergic system releases a small reward. We literally feel pleasure when we're right. When we encounter information that contradicts them, the insular cortex and amygdala generate discomfort, a threat signal. The brain treats informational dissonance as pain.

Marcus Bruzzo argues that recommendation algorithms exponentially amplify confirmation bias. Social media feeds users content that confirms their views, creating feedback loops in which the brain’s natural bias and the algorithm’s bias reinforce each other. The result is a hypertrophied version of confirmation bias that no previous generation has experienced.

The Halo Effect, described by Edward Thorndike in 1920, occurs when one positive characteristic of a person colors your assessment of all their other characteristics. If someone is physically attractive, we tend to judge them as more intelligent, more competent, more honest, and more trustworthy. A single perceived quality radiates through our entire judgment, like a halo of light around a sacred figure.

Research shows the effect with disturbing precision. In a classic experiment, the same academic papers were rated as significantly more intelligent and better written when accompanied by photos of authors considered attractive. The evaluators were unaware that the author’s appearance had influenced their assessment of the content.

In politics, the Halo Effect explains why taller candidates win more elections, why appearance influences election results, and why charismatic leaders maintain support despite objective failures. In the job market, studies show that attractive people earn, on average, 10-15% more than equally qualified colleagues. The bias operates in hiring, promotion, and performance evaluations.

The Dunning-Kruger effect, documented by David Dunning and Justin Kruger in 1999, reveals a disturbing paradox: people with low ability in a domain tend to drastically overestimate their competence, while experts tend to underestimate their own. Incompetence robs people of the ability to recognize their own incompetence.

The effect follows four predictable stages. In unconscious incompetence, a person doesn’t know enough to realize what they don’t know. In conscious incompetence, they begin to understand the vastness of their own ignorance. In conscious competence, they have the skill but need deliberate effort. In unconscious competence, the skill becomes automatic. The peak of unjustified confidence occurs in the first stage.

Bertrand Russell captured the phenomenon before it had a scientific name: "The fundamental problem of the modern world is that fools are full of certainty and intelligent people are full of doubt." Social media dramatically amplifies the first stage: platforms give a megaphone to those who are most certain, and extreme certainty generates more engagement. The result is an information ecosystem where the least qualified speak the loudest.

The availability bias leads us to estimate the probability of an event based on how easily examples come to mind, not on its actual frequency. After watching a news report about a plane crash, we feel that flying is dangerous. But the chance of dying in a plane crash is 1 in 11 million flights. The chance of dying from causes related to poor diet is hundreds of times higher, but we do not feel afraid when we open the refrigerator.

The media is the biggest machine for distorting availability that exists. News about terrorism, shark attacks, and spectacular accidents create the perception that these events are frequent. Meanwhile, cardiovascular disease, diabetes, and depression kill orders of magnitude more people, but receive proportionally little coverage. What's dramatic and memorable replaces what's statistically relevant.

In everyday life, the availability bias shapes concrete decisions: parents overprotect children against kidnapping (extremely rare) while neglecting seat belts (a real cause of child deaths). Investors sell stocks after crisis headlines, when statistically the best choice is to hold. Governments allocate billions to counterterrorism while underfunding the prevention of chronic diseases that kill a thousand times more people.

Recommendation algorithms amplify cognitive biases. They don’t create new biases, but take the ones already in the human brain and scale them to unprecedented proportions. Confirmation bias becomes an information bubble. The availability bias becomes a curated feed. The Halo Effect becomes influencer marketing. Every natural bias gets a technological multiplier.

Epistemic bubbles form when the algorithm delivers only content aligned with the user's beliefs. Unlike echo chambers (where disagreement is actively rejected), bubbles are invisible: the person is simply never exposed to alternative views. They believe they are seeing "the world" when they are seeing a selection curated by the algorithm to maximize engagement.

The feedback loop between cognitive biases and algorithmic biases works like this: your brain has confirmation bias, so you engage more with content that confirms your beliefs. The algorithm detects this pattern and delivers more of the same. Your bias grows stronger. You engage even more. The algorithm optimizes further. Within a few weeks, your informational reality has narrowed without you noticing. It’s a system of mutual reinforcement with no natural brake.