Two invisible rules run every room: what people around you actually do, and what they would approve of you doing. They usually agree, but when they split, they pull in opposite directions, and a campaign, a brand, or a parent who confuses the two will reliably produce the opposite of what they intended.
A norm is a shared rule of behaviour, and the single most useful idea here is that there are two kinds, which people constantly confuse. A descriptive norm is what most people actually do ("most guests reuse their towels"). An injunctive norm is what most people approve of ("littering is frowned upon"). Robert Cialdini and colleagues drew the distinction sharply and showed the two work through different channels: descriptive norms guide by information and imitation (if everyone does it, it is probably sensible and safe), while injunctive norms guide by anticipated social approval or sanction (Cialdini, Reno and Kallgren, 1990). Usually they line up, but they often diverge, most people exceed the speed limit (descriptive) while most people disapprove of speeding (injunctive), and it is exactly when they split that the distinction earns its keep.
Norms also steer us far more than we think, and mostly below awareness. When researchers asked people what influenced their energy use, they ranked "what my neighbours do" dead last, behind saving money and protecting the environment. Yet when actual behaviour was measured, the neighbours' norm predicted it best of all. Social influence is not just powerful; it is powerful and underdetected, so we are moved by the crowd while sincerely believing we are not (Nolan et al., 2008).
The finding that makes all this matter for anyone running a campaign is the boomerang. In a now-classic field experiment, households were told how their energy use compared with their neighbourhood average. Heavy users, told they were above average, cut back, exactly as hoped. But light users, told they were below average, increased their consumption, drifting up toward the norm. A descriptive norm, it turns out, is a magnet for the average: it pulls the high down and the low up. The fix was elegant: adding an injunctive cue, a simple smiley face signalling approval, to the low users' feedback kept them low. Approval stopped the boomerang (Schultz et al., 2007). The same team's hotel-towel studies showed the constructive side: telling guests that most previous occupants of their room had reused towels lifted reuse more than a generic environmental appeal (Goldstein, Cialdini and Griskevicius, 2008).
Cristina Bicchieri gave the most rigorous account of when a norm actually governs behaviour. A social norm, she argues, holds when two conditions are met within a reference group: empirical expectations (you believe others conform) and normative expectations (you believe others expect you to conform, and may sanction you if you do not). Change either belief and the norm's grip changes, which is why norms are a conditional preference, not a fixed value (Bicchieri, 2006, 2017).
Two cautions keep norms from being oversold as a lever.
The first is size and reliability. Norm-based interventions work on average, but the effects are modest and highly context-dependent. A meta-analysis of field experiments found real but small average effects, with wide variation by behaviour and setting, and the celebrated boomerang itself, while genuine, was not large (Bergquist, Nilsson and Schultz, 2019). Norms are a cheap and useful tool, not a magic dial, and the honest expectation is a nudge, not a transformation.
The second is over-application. The descriptive-injunctive distinction is sometimes wielded as a checklist ("add an injunctive cue and you are safe") rather than understood as a theory. It matters whose norm you invoke: a norm only moves you if the reference group is one you identify with, so telling people that a group they do not belong to, or actively define themselves against, behaves a certain way can do nothing or even backfire (this is Bicchieri's reference-network point, and it connects to the identity dynamics of L3-07). "Most people" is not a magic phrase; "most people like you" is closer to the real mechanism.
A large share of the strong evidence comes from pro-environmental field experiments, energy, water, hotel towels, littering, so how far the exact effects transfer to health, finance, or politics is a live question. Norm effects also vary across cultures and can fade over time, and much of the work measures short-run behaviour rather than lasting change.
When does a descriptive norm dominate an injunctive one, and vice versa? How durable are norm-based interventions once the message stops? And how do highly visible online norms, where you can see exactly what thousands of people are doing, change dynamics that were worked out for private behaviour and local reference groups?
The usable core: two norms, what people do and what they approve, and confusing them backfires. Because the influence is real, cheap, and mostly invisible to the person being moved, it sits squarely in this series' dual-use territory: the same lever that gently steers people toward a genuine good can fabricate a false consensus or accidentally normalise the very thing you are fighting.
Social proof is descriptive-norm marketing, "bestseller", "most popular", "your neighbours switched", and it works because it is one of the most reliable influence cues there is. Two disciplines make it honest and effective. First, it must be true: a fabricated norm ("everyone's switching") is both manipulation and fragile, because a claimed consensus that turns out to be invented spends trust the moment it is seen through. Second, watch which norm you are actually advertising. "Only a handful of customers have upgraded" tells everyone that upgrading is abnormal, the boomerang in a marketing deck, so if the behaviour you want is still a minority, do not cite the descriptive norm at all; lead with the injunctive one ("customers who care about X choose this") or with a reference group the customer wants to join. And because norms move people who do not notice they are being moved (Nolan et al., 2008), the honesty bar is higher here than it feels, since the customer cannot weigh an influence they cannot see.
This is the classic graveyard, and the mistake is always the same. Campaigns against littering, tax evasion, binge drinking, or vaccine hesitancy that dramatise how common the bad behaviour is ("X% of people cheat on their taxes", "most students binge drink") inadvertently broadcast it as the descriptive norm and can increase it (Schultz et al., 2007). Before any norms-based message, audit the current descriptive norm honestly. If the good behaviour is already the majority, say so. If it is not, never announce the bad norm; lead with injunctive framing ("nine in ten people think you should") or highlight a growing trend rather than a grim majority. And resist manufacturing false consensus, since a bandwagon built on invented numbers is both unethical and brittle.
Norms are among the cheapest and most scalable policy tools going, tax letters noting that "most people in your area have already paid", energy comparisons, health defaults, and they produce real gains. But the meta-analytic reality is modest, heterogeneous effects (Bergquist, Nilsson and Schultz, 2019), so the honest posture is to pilot, measure, and expect a nudge rather than a fix, and to build in the injunctive backstop so that telling the compliant majority what everyone does does not quietly license the low-users to drift upward.
Three habits. First, notice that "what everyone's doing" is steering you more than you would ever credit, especially since the influence is designed to be invisible, so when you feel a pull toward the common choice, ask whether it is actually the right one or merely the popular one. Second, check whose norm is being invoked: "most people" only binds you if they are your people, so an appeal to a group you do not identify with carries no real weight. Third, when you want to encourage a behaviour, in your team, your kids, your community, describe the good behaviour as already normal and approved, and never, ever advertise how common the bad behaviour is, because you will simply be telling everyone it is what people like them do.
.png)