Cognitive Biases: The Predictable Bugs in Human Thinking

Quick: a bat and a ball cost ₹110 together. The bat costs ₹100 more than the ball. How much is the ball?

If “₹10” jumped to mind, you just met System 1 — and its answer is wrong (it’s ₹5; check: 5 + 105 = 110). That reflexive answer isn’t stupidity. It’s the signature of a mind built for speed over accuracy, and its errors are not random — they’re systematic, which means they’re predictable, which means you can design around them. That’s the entire field of cognitive biases in one sentence.

Two systems, one mind

Daniel Kahneman’s framing (from Thinking, Fast and Slow, built on decades of work with Amos Tversky):

The key insight: System 1 cannot be turned off. You will feel the wrong answer to the bat-and-ball problem even after you know it’s wrong (the same way an optical illusion persists after it’s explained). Debiasing therefore isn’t about becoming unbiased — it’s about knowing when to summon System 2.

Heuristics: the features behind the bugs

Biases aren’t design flaws; they’re the visible cost of heuristics — shortcuts that trade accuracy for speed and are usually right:

Anchoring

The first number you see drags every subsequent estimate toward it — even when the anchor is obviously irrelevant. In the classic study, spinning a rigged wheel of fortune before estimating “what % of UN countries are African” shifted answers by ~20 points. This is why the first offer in a negotiation matters, why “was ₹4,999, now ₹1,999” works, and why estimate meetings should collect numbers before anyone speaks.

Availability

We judge how likely something is by how easily examples come to mind. Plane crashes feel more dangerous than car trips because crashes are vivid and televised; shark attacks feel likelier than falling coconuts for the same reason. Anything that boosts mental availability — recency, drama, media coverage, personal experience — inflates perceived probability.

Representativeness

We judge probability by resemblance to a stereotype, and it makes us ignore base rates. “Linda is 31, outspoken, majored in philosophy, active in social justice — is she more likely (a) a bank teller or (b) a bank teller active in feminism?” Most people pick (b), which is logically impossible — every feminist bank teller is a bank teller. The description’s fit to the stereotype overwhelms the math.

The big biases worth knowing by name

• Loss aversion. Losses hurt roughly twice as much as equivalent gains feel good (the core of Kahneman & Tversky’s prospect theory — the work the 2002 Nobel was for). Consequences everywhere: we hold losing investments too long, fear cancelling a subscription more than we valued starting it, and respond harder to “don’t lose ₹500” than “gain ₹500.”
• Framing. “90% survival rate” and “10% mortality rate” are the same fact and produce different decisions — even among physicians.
• Confirmation bias. We search for, notice, and remember evidence that fits what we already believe, and interrogate only the evidence that doesn’t. The most consequential bias on this list, because it compounds: it curates your information diet.
• Sunk cost fallacy. Continuing a doomed project because of what’s already spent. The rational move ignores the past entirely — only future costs and benefits count — but abandoning feels like converting a paper loss into a real one (loss aversion again, wearing a different hat).
• Hindsight bias. After the outcome, it feels like you “knew it all along,” which quietly destroys your ability to learn from decisions — every result feels predicted, so no prediction ever gets graded honestly.
• Overconfidence & the planning fallacy. People’s “90% confident” ranges contain the truth ~50% of the time, and projects planned from the inside view (“our steps, our team”) reliably blow past estimates. Fix: the outside view — “how long did similar projects actually take?”
• Dunning–Kruger effect. The least skilled overestimate themselves the most, because the skill needed to perform is the same skill needed to judge performance. (Popular tellings overstate it — it’s a real but modest miscalibration, not “idiots think they’re geniuses.”)
• Survivorship bias. We study the successes because the failures aren’t around to be studied — the WWII insight that you armour where returning planes weren’t shot, because planes hit there never returned.

Debiasing: what actually works

Honest news first: knowing about biases barely helps. Reading this page doesn’t uninstall System 1. What works is changing the process or the environment, not the mind:

1. Change the environment (best ROI). Make the good path the default — opt-out organ donation, auto-enrolled savings. Defaults beat willpower because they don’t fight System 1, they enlist it.
2. Use the outside view. Before estimating, ask what the base rate is for projects/startups/renovations like this one, and start from there.
3. Premortem. Before committing: “It’s one year later and this failed — write the story of why.” Legitimizes doubt that confirmation bias would otherwise silence.
4. Consider the opposite. The one cognitive strategy with decent evidence: explicitly generate reasons your judgement might be wrong.
5. Decide with rules, not vibes, for repeated decisions. Checklists and simple scoring models beat expert intuition in noisy domains (hiring, forecasting) with embarrassing consistency.
6. Slow the trigger. For irreversible calls, insert delay between impulse and action — sleep on it, or have a “24-hour rule” for big purchases. Time is what System 2 needs and System 1 hates.

Why this matters beyond trivia

These aren’t party facts — they’re load-bearing in other domains on this site. The laws of UX are largely biases operationalized (defaults, anchoring in pricing pages, loss-framed copy). User behaviour is System 1 navigating your interface. And if you build AI agents, note the irony: LLMs exhibit anchoring, framing, and sycophantic confirmation too — they learned from us.

Next: the machinery those judgements run on — memory, and how to actually learn things.