How behavioural science explains the decisions you regret: Kahneman, Cialdini, Thaler and the documented research

Most people rate themselves as more rational than the average person, which is itself one of the biases this collection documents. The researchers here spent careers running the experiments that show where reasoning actually breaks, not to embarrass anyone but because the gap between how people think they decide and how they actually decide has consequences: bad hires, bad pension defaults, bad product decisions, bad verdicts. Daniel Kahneman and Amos Tversky built the experimental foundation. Robert Cialdini mapped how persuasion actually operates on the people receiving it. Richard Thaler took the research into policy. This guide works through the signature findings, with the study or mechanism behind each one, before looking at where the field's various strands genuinely disagree about what to do with the evidence.

None of what follows requires taking the researchers' word for it. Every finding below comes from a published experiment or a documented field study with a specific method, which is precisely why the field has held up better under scrutiny than most of applied psychology: the claims are checkable, and several have been checked repeatedly by independent labs since.

System 1 and System 2

Kahneman and Tversky's decades of joint research, later collected in Kahneman's Thinking, Fast and Slow, established that human judgement runs on two distinct systems. System 1 operates automatically: reading a simple sentence, recoiling from a loud noise, sensing that a stranger is angry. System 2 operates deliberately and requires effort: working through a multi-step calculation, comparing two mortgage offers on the actual terms rather than the headline rate. The finding that reshaped the field is not that System 1 exists, but that System 2 is lazy: it defers to System 1's fast answer far more often than people believe, and rarely audits it.

A worked example: asked whether a bat and ball together cost $1.10, with the bat costing $1 more than the ball, most people answer 10 cents almost instantly. The correct answer is 5 cents. This specific question comes from Shane Frederick's cognitive reflection test, not from Kahneman and Tversky's own experiments, but Kahneman discusses it at length in Thinking, Fast and Slow precisely because it demonstrates the same mechanism their work established: the 10-cent answer is System 1 pattern-matching to a familiar shape ("$1 and 10 cents") without System 2 ever checking the arithmetic, and the error is not stupidity; it is System 2 declining to intervene when System 1 feels confident.

Where it fails as a framework: knowing that System 1 and System 2 exist does not, by itself, tell you which one is running at any given moment, and Kahneman himself was candid in later interviews that awareness of the biases barely reduces his own susceptibility to them. The framework describes the mechanism accurately; it does not hand you a reliable switch to turn System 2 on when it matters most.

The six principles of influence

Cialdini's research, built on years of field study inside sales and fundraising organisations that had no idea they were being studied, identified six recurring levers: reciprocity, commitment and consistency, social proof, authority, liking, and scarcity. Each works largely below conscious deliberation, which is why the same six levers show up across cultures and contexts that have never read his book.

A worked example: a charity that sends a small unsolicited gift (a personalised address label) before asking for a donation reliably increases response rates, because reciprocity creates a felt obligation disproportionate to the gift's actual value. Cialdini documented this pattern directly in fundraising campaigns, and the mechanism, an automatic sense of owing something back, operates whether or not the recipient consciously notices the gift arrived first.

Where it fails: the principles describe what moves behaviour, not what should. Cialdini's own later writing draws an explicit ethical line between applying a principle to a genuine claim (real scarcity, real social proof) and manufacturing a false signal to trigger the same automatic response (fake countdown timers, fabricated reviews). The framework is silent on which side of that line a given use falls; that judgement has to come from outside the model.

Choice architecture

Thaler, building on the same foundational biases Kahneman and Tversky documented, argued with Cass Sunstein that no choice is ever presented without some structure: an order, a default, a framing. Since architecture is unavoidable, the question is not whether to design it but whether to design it in the chooser's interest. Their signature policy example is retirement saving: switching employee enrolment from opt-in to opt-out, with automatic escalation as pay rises, produced dramatically higher participation without removing anyone's freedom to leave the scheme.

A worked example: a cafeteria that places fruit at eye level and puts dessert in a less visible location changes what people choose without banning anything, because the default option in low-attention decisions carries outsized weight. Thaler and Sunstein's term for this, libertarian paternalism, is meant to capture precisely that tension: nudging toward a better default while leaving the exit fully available.

Where it fails: choice architecture assumes the designer has a reasonably clear and legitimate view of what serves the chooser's actual interest, which is a much easier case to make for a pension default than for many commercial contexts, where the designer's interest and the chooser's interest diverge. The same technique that increases retirement saving can be redirected to increase impulse spending, and the framework alone doesn't distinguish the two cases; only the stated intent and the transparency test do.

Growth mindset and effort-based praise

Dweck's research, running from childhood praise studies through decades of follow-up work, found that praising a child for being smart produces more fragile performance under difficulty than praising the specific effort or strategy that produced a result. Children praised for intelligence were more likely to avoid harder subsequent tasks, apparently protecting the identity the praise had assigned them; children praised for effort took on harder tasks and persisted longer when they struggled.

A worked example: two students get the same test score. One is told "you're clearly talented at this." The other is told "that strategy of checking your working paid off." Dweck's studies found the second framing produced more resilient behaviour on the next, harder test, because the story the student tells about the result, talent versus applied effort, changes what they think a poor result on the next attempt would mean about them.

Where it fails: the framework has been widely diluted in classroom and corporate implementation into slogans ("adopt a growth mindset!") detached from the specific mechanism, effort-focused feedback tied to concrete strategy, that produced the original results. Dweck herself has pushed back publicly against shallow applications that treat mindset as a motivational poster rather than a specific feedback discipline.

Tiny habits and the ability lever

Fogg's research reframes behaviour change around a simple equation: behaviour happens when motivation, ability and a prompt converge at the same moment. His central practical claim, developed over years of lab and field testing at Stanford, is that ability is the more reliable lever to pull, because motivation fluctuates day to day and cannot be manufactured on demand, while a sufficiently small behaviour requires almost none.

A worked example: rather than committing to a 30-minute daily workout, which depends on motivation that will predictably fail on a bad day, Fogg's method starts with two push-ups attached to an existing habit ("after I brush my teeth") and an immediate small celebration on completion. The behaviour is deliberately trivial to sustain the ability side of the equation; expansion happens naturally once the habit is anchored, rather than being demanded up front.

Where it fails: shrinking a behaviour to near-zero effort works well for genuinely new habits but is a poor match for behaviours that are inherently effortful regardless of framing, complex negotiations, difficult conversations, sustained deep work, where no amount of habit-stacking removes the real cognitive load the task requires.

Where the field disagrees with itself

The apparent unity of "behavioural science" hides a real methodological split. Kahneman and Tversky's tradition is descriptive: document how judgement actually deviates from a rational baseline, without necessarily prescribing what to do about it. Thaler's tradition is prescriptive: take the documented biases and use them to design better defaults, accepting the ethical responsibility that comes with deliberately shaping someone's environment. Cialdini's tradition sits closer to the applied and commercial end, describing levers that can be deployed for outcomes the researcher does not control once the technique is published.

This is not a contradiction to resolve so much as three different professional postures toward the same underlying findings. A researcher publishing a bias is not thereby endorsing every use of it, which is precisely why Thaler's explicit ethical framing (the transparency test) and Cialdini's explicit ethical framing (genuine versus manufactured signals) both exist: without them, purely descriptive research becomes a toolkit with no attached judgement about use.

---

The Behavioural Science category collects these frameworks, and others including Angela Duckworth, BJ Fogg, Viktor Frankl and Jonathan Haidt, as downloadable .md files for Claude, ChatGPT or any LLM. If you are applying this thinking to an actual high-stakes call rather than studying the research, Decision Brief is a $79 tool built to structure exactly that decision against your own likely blind spots.