Heuristics and biases

Heuristics are mental shortcuts. System 1 uses them constantly to make rapid judgments and decisions without expensive deliberation. They work well in most cases — that's why they evolved. But they fail systematically in specific patterns, producing biases that matter for economic behaviour.

Availability heuristic

When judging the frequency or probability of an event, people use how easily examples come to mind. Easily-recalled events are judged more probable than they actually are; hard-to-recall events less probable.

• Plane crashes feel more dangerous than car crashes (vivid news coverage), even though car crashes kill vastly more people. Causes over-investment in flight safety relative to road safety
• Recent events are over-weighted — recent volatility shapes risk perceptions; recent inflation shapes expected inflation
• Salient examples dominate — if you know someone whose business failed, you over-estimate business-failure rates in your judgements

Representativeness heuristic

When judging whether something belongs to a category, people compare its features to stereotypes of the category. They ignore base rates (how common the category is overall).

Classic example (Kahneman-Tversky): 'Linda is 31, single, outspoken, very bright, majored in philosophy, was deeply concerned with issues of discrimination and social justice. Which is more probable: (a) Linda is a bank teller; (b) Linda is a bank teller and a feminist movement activist?' Most respondents choose (b). But (b) is logically a subset of (a) — it must be LESS probable. The descriptive match to 'feminist activist' triggers the representativeness heuristic, overriding the base-rate logic.

Anchoring and adjustment

When making numerical judgments, people start from an initial value (anchor) and adjust. The adjustment is typically insufficient — the final estimate stays close to the anchor even when the anchor is arbitrary.

• Negotiation — the first number on the table strongly shapes the final outcome
• Suggested-retail-price labels — set the anchor for purchase decisions
• Salary discussions — initial offers heavily influence final agreements
• Tip calculations — % suggestions on restaurant receipts shape tip size

Affect heuristic

Judgments are colored by emotional reactions. If something feels good, we judge it as both more beneficial and less risky than it actually is. If it feels bad, the opposite.

• Investment decisions — investments in 'glamour' sectors (tech, AI, EV) get over-weighted relative to fundamentally similar investments in unglamorous sectors. Same risk profile, different perceived risk
• Charitable giving — appealing-photo appeals raise more than statistical appeals with the same need
• Risk perception — nuclear power is widely perceived as much riskier than coal-power on the same units of comparison, partly because of accident-imagery affect

Confirmation bias

We search for, interpret, and remember information that confirms our prior beliefs. Information that contradicts them is treated more sceptically or forgotten.

• Investment thesis stickiness — once an investor forms an investment thesis, contradictory evidence is rationalised away
• Political views — voters interpret the same fact differently depending on prior political identity
• Health beliefs — once someone forms a view about a supplement or remedy, they remember positive experiences and discount negative ones

Overconfidence and the illusion of control

Most people rate themselves as above-average drivers, above-average earners, above-average decision-makers. The average can't be above-average; most are wrong.

• Entrepreneurship — survey: most aspiring entrepreneurs rate their business-success probability at 50%+; empirical 5-year survival rate is ~30-40%. Beneficial bias for entrepreneurship (without it nobody would start firms) but consequential for financial planning
• Risk assessment — drivers in crashes overestimated their driving skill before the crash. Investors who lose money overestimated their skill before the loss
• Time estimation — people consistently underestimate how long tasks will take (planning fallacy). Project managers across industries; PhD students; book authors. Estimates are 30-50% short of actual completion time on average

Framing effects

The same option presented in different frames produces different choices. Tversky-Kahneman 'Asian disease problem' classic experiment showed strong frame-dependence.

• Gain frame vs loss frame — '90% survival' vs '10% mortality'. Same statistic; different choices
• Per-day vs per-year framing — '$50/year' feels different from '$1/week' (which is more)
• Absolute vs relative — 'this drug works 70% better' is more persuasive than 'this drug improves outcomes by 0.5 percentage points'
• Default opt-in vs opt-out — same underlying choice, different uptake rates

Bias correction is hard

Awareness of biases doesn't eliminate them. Researchers who study biases still display them. The biases are System 1 features; awareness is System 2. Training, careful procedure design, and explicit checks help — but full correction is essentially impossible.

Implications for financial-services design

Behaviorally-informed design recognises that users will use heuristics — and either accommodates them or counteracts them:

• Loan-application UX — clear, salient disclosure of total cost and repayment burden (not just monthly payment) to fight loss-side framing tricks
• Investment disclosure — risk-warning prominence; comparison-class disclosure; historical-performance with appropriate disclaimers
• Savings-product framing — emphasise the future-self benefit; use gain-frame language; show trajectory of accumulated value
• Insurance pricing — calibrate premium menus to leverage screening (covered in Micro module 6) rather than relying on customer self-classification