4  How People Actually Think

How do real humans make decisions, and how can that be exploited?

NoteLearning Objectives
  • Distinguish fast/automatic from slow/deliberate processing.
  • Name several biases and give an example of each in an online setting.
  • Explain how identity and group membership shape belief, and what that implies for “just show people the facts”.

Why did you click the last thing you clicked? The honest answer is usually not “because I weighed my options and chose rationally.” It is something closer to “because it was there, it looked familiar, and everyone else seemed to be clicking.” Most of what people do online is not the product of careful deliberation; it is the product of fast, automatic cognition, and fast, automatic cognition is predictable. Predictability is exploitability.

4.1 Two Systems, Many Biases

The psychologist Daniel Kahneman described two ways of thinking (Kahneman 2011). System 1 is fast, automatic, and associative: it is what lets you read a familiar face, flinch at a loud noise, or scroll without deciding to scroll. System 2 is slow, effortful, and deliberative: it is what you use to check a calculation, weigh a tradeoff, or resist the scroll. Most online behavior is System 1, which is why it is so predictable, and why it is so cheap to exploit at scale.

The heuristics System 1 uses are usually adaptive. Nearby numbers are usually informative, so anchoring—the tendency to lean on the first number you see— usually helps. In one experiment, people spun a wheel rigged to land on 10 or 65, then estimated the percentage of African countries in the United Nations; those who had seen 65 guessed about 45 percent higher than those who had seen 10. They knew the wheel was random, but the number shaped their thinking anyway. This is why prosecutors set high anchor charges, why retailers display high “original” prices, and why the person who names the first number in a salary negotiation usually wins (Galinsky and Mussweiler 2001).

The availability heuristic says that people estimate how likely something is by how easily examples come to mind. After a plane crash dominates the news, people overestimate the risk of flying, even though the statistics have not changed. Confirmation bias pushes people toward evidence that supports what they already believe, and the representativeness heuristic makes them judge by resemblance rather than base rates. None of these are bugs; they are features of a mind tuned for speed. But a feature is an exploit waiting to be found, and the attention economy has found them all.

Two more findings matter for understanding online behavior. The first is loss aversion: a loss produces roughly twice the emotional impact of an equivalent gain. Taxi drivers in New York and Singapore work longer hours on bad days, when earnings are below their daily target, and knock off early on good days— managing losses against a reference point rather than maximizing income. The second is hyperbolic discounting: an extremely steep preference for the present over any future point. This is why someone can genuinely plan to quit smoking next year while lighting a cigarette, and why gym memberships are purchased with full intention and then rarely used. Our future selves are strangers, and we are generous to ourselves and stingy with strangers.

4.2 Belonging and Obedience

People do not hold beliefs only to be right; they hold them partly to belong. The psychologist Henri Tajfel found that merely assigning people to groups at random— by a coin flip, by an arbitrary label— was enough to produce in-group favoritism (Tajfel and Turner 1979). Once a belief is tied to membership in a group, changing it is not just an intellectual update; it is a social risk. Motivated reasoning is what happens next: the mind becomes a lawyer hired to defend a conclusion rather than a judge weighing evidence.

Social context can overpower private judgment outright. In Solomon Asch’s line-length studies, subjects who could plainly see that two lines were the same length still agreed with a roomful of confederates who said otherwise (Asch 1956). In Stanley Milgram’s obedience experiments, sixty-five percent of subjects administered what they had been told was the maximum voltage to an unseen “learner,” because a person in a lab coat told them to (Milgram 1974). Milgram’s interpretation remains contested—his subjects were also following cues that this was a legitimate scientific procedure— but the core finding is robust: ordinary people will do harmful things when the social setting demands it. And the setting can be broken by a single dissenter. When one confederate refused to continue, compliance collapsed, because the social proof of refusal is as powerful as the social proof of compliance.

This is where moral differences become legible. Jonathan Haidt’s moral-foundations theory holds that different audiences weight different values—care, fairness, loyalty, authority, purity—differently (Haidt 2012). Two reasonable people can talk past each other about content moderation because one is reasoning from care and fairness while the other is reasoning from loyalty and authority. Persuasion that ignores which foundation the other person is standing on is not persuasion at all.

4.3 Status and the Social Self

What people want is not only things; it is position. In 2001, Norway made its tax records publicly searchable online, and Ricardo Perez-Truglia used the moment as a natural experiment (Perez-Truglia 2020). Absolute incomes did not change, but the well-being gap between higher- and lower-income Norwegians widened by 29 percent, because people could now see exactly where they stood relative to everyone else. Robert Frank found the same preference in the workplace: workers systematically accept lower total pay in exchange for higher rank within their peer group (Frank 1985). A programmer who is the best-paid person on a small team may prefer that to being a lower-ranked member of a higher-paying team.

Thorstein Veblen saw this a century earlier. In The Theory of the Leisure Class, he described conspicuous consumption: spending whose primary purpose is to signal social rank, and whose signal only works because it is costly (Veblen 1899). Fred Hirsch later pointed out that some goods are inherently positional— their value depends on how few other people have them, and they cannot be democratized. An ocean view, a title, a degree from a selective school: these lose their value the moment everyone has one. Digital platforms have turned status into a continuous, quantified feed, where follower counts and reaction metrics do the work Veblen described, at a scale and refresh rate he could not have imagined.

Preferences themselves are not as given as economics pretends. Edward Bernays, Sigmund Freud’s nephew, was hired by American Tobacco in 1929 to make women smoke. He staged a group of women lighting cigarettes on Fifth Avenue as “torches of freedom,” framing a taboo-breaking act as feminist defiance, and it worked (Bernays 2024). Recommendation systems are the digital successor to this technique. They do not merely show you content that matches what you already want; they build a model of your pauses, clicks, and watch time, then serve content designed to maximize engagement. A recommendation system is also a preference-construction system, and the customer is never meant to arrive, because arrival would end the business model.

4.4 People Believe Stories, Not Spreadsheets

A colleague once observed that people want data but believe stories. This is why the most effective public-health campaigns present specific, named people in specific situations rather than statistics. It is also why industries that want to prevent belief change fund think tanks that produce reports full of data: the point is not to persuade on the merits, but to borrow the appearance of rigor. The fossil-fuel industry has run this operation for thirty years, and tech companies have learned the technique.

The deeper problem is that beliefs contested in public function primarily as signals of group identity (Lubrano 2025). Holding the wrong belief does not just mean being misinformed; it means being outside the group, so the social cost of updating is often higher than the mental cost of staying wrong. The Facebook Papers, leaked by Frances Haugen in 2021, showed that the company’s own researchers knew Instagram was worsening mental health among teenage girls, and that its algorithms amplified outrage, and that the company chose not to act because acting would reduce engagement (Frenkel and Kang 2021). The people inside those organizations were not failing to understand the evidence. They were responding to incentives that made understanding expensive.

4.5 Nudges and the Attention Economy

If small changes in how choices are presented can have large effects, then choice architecture is a policy tool. When workers must actively opt in to a pension plan, participation runs around 50 to 60 percent; when they are enrolled unless they opt out, it rises to 80 to 90 percent, with no change to the financial terms. Britain’s automatic-enrollment reform of 2012 brought over ten million additional workers into workplace pensions, and the Behavioural Insights Team found that adding “nine out of ten people in your area pay their taxes on time” to a tax letter recovered tens of millions of pounds at essentially no cost (Thaler and Sunstein 2009).

The same science runs the attention economy in the other direction. Variable rewards, infinite scroll, manufactured urgency, disguised ads, social proof, and default nudges are all engineered to exploit known psychology. The distinction between a nudge and a shove is not in the mechanism but in whose interests the design serves. Automatic pension enrollment serves the worker; automatic subscription renewal that is hard to cancel serves the company. Every notification badge and “people who liked this also liked” recommendation is a behavioral intervention. A design choice is a psychological intervention, whether the designer acknowledges it or not.

4.6 Moral Disengagement

The last piece of the puzzle is how good people do harmful work without feeling like bad people. Albert Bandura called the mechanism moral disengagement (Bandura 1999). It operates through euphemism, displacement of responsibility, diffusion of responsibility, and treating the harmed as abstractions. The tech industry’s vocabulary is a catalogue of these moves: user data described as “exhaust,” manipulative design described as “engagement optimization,” surveillance described as “connecting people with relevant products.”

None of this requires individual villainy. When Robert Hare, the psychologist behind the Psychopathy Checklist, was asked to evaluate a publicly traded corporation as if it were a person, his conclusion was that it fit the profile (Bakan 2005). And the people who rise to the top of such organizations are not a random sample: one study of senior UK managers found they scored higher than the patients of a high-security psychiatric hospital on three personality-disorder traits (Board and Fritzon 2005). The system selects for the traits it rewards, and the traits it rewards are not always the ones it says it wants. Design that respects how people actually think is possible; design that exploits it is the business model of the attention economy.

NoteTakeaway

People are not broken computers: they are social animals with fast, fallible, identity-driven cognition. Design that respects this is possible; design that exploits it is the business model of the attention economy.

4.7 Exercises

Bias autopsy.
Take a recent online argument or viral post. Identify at least two specific biases at work (confirmation, availability, in-group signaling, anchoring) and say how the platform’s design amplified them.
Dark-pattern hunt.
In a product you use, find one dark pattern (forced continuity, disguised ads, manufactured urgency, infinite scroll, confirshaming). Name the psychological mechanism it exploits.
Identity check.
Think of a belief you hold partly because of group membership. Describe, honestly, what would happen to your social life if you changed it. Discuss what this implies for “if we just showed people the facts.”
Moral-foundations translation.
Pick a policy debate (e.g., whether to moderate misinformation). Argue it once from a care/fairness frame and once from a loyalty/authority/purity frame. Which audiences respond to which?
Design a nudge.
Choose a desirable online behavior (e.g., reading before sharing). Design one System 1 intervention (default, friction, social proof) and one System 2 intervention (education, disclosure). Predict which works better and why, then discuss the ethics of each.
Status check.
Identify one place where you compare yourself to others online (follower counts, salary, shipping speed). What does the comparison do to your satisfaction, and who benefits from keeping you in that comparison?
Nudge autopsy.
Find a default in a product you use (auto-renew, pre-checked boxes, a default privacy setting). Trace who set it, who benefits, and what the opt-out rate likely is.