
Privacy & Surveillance · 9 min read
The Most Important Technology–Psychology Finding Is That Behaviour Can Now Be Designed at Scale
The consequential change is not simply that more of life happens through screens, nor that artificial intelligence is becoming more capable. It is that digital systems can increasingly observe patterns, infer context, an
The consequential change is not simply that more of life happens through screens, nor that artificial intelligence is becoming more capable. It is that digital systems can increasingly observe patterns, infer context, and alter the choices presented to people in the moment those choices are made. Behavioural science supplies a model of how people respond to defaults, prompts, reminders, friction and framing; data-rich technology supplies reach, speed and feedback. Together, they turn influence from a one-off message into an adjustable environment.[S2]
That is the strongest current Technology × Psychology × Society finding: the same infrastructure that can help someone follow through on an intention can also make their attention, data and habits easier to steer. Its social consequence is that autonomy is no longer protected merely by being informed. It depends increasingly on whether the systems around us are designed to leave room for a considered decision.
For ordinary people, this is practical rather than abstract. It appears in the choice that is already selected, the reminder that arrives at a vulnerable moment, the privacy setting hidden behind several screens, the recommendation that makes the next action feel natural, and the service that learns which prompt gets a response. The question is not whether technology affects behaviour. The more useful question is: who has designed the prompt, what outcome does it serve, and can the person meaningfully refuse it?
From Tools to Choice Environments
A tool can be used or put down. A choice environment does more: it arranges the conditions in which a decision is made. Digital products can set defaults, order options, time notifications, simplify one route and complicate another. None of those moves needs to command anyone. Their power lies in working with familiar limits of attention, time and memory.
The City Bar report describes behavioural nudging as subtle guidance through design choices such as prompts, reminders and default settings. In its benign form, a nudge can encourage a person to complete training, take part in a sustainability programme or make a decision they already endorse. The report also makes the boundary clear: ethical approaches require transparency, an ability to opt out and consideration of psychological effects.[S2]
This matters because digital environments do not merely contain occasional nudges. They can become a continuous sequence of them. A paper form may have one default. A connected service can test an interface, observe behaviour, change the next prompt and repeat. Digital technologies expand the capacity to collect and analyse data; behavioural science offers a framework for interpreting that data and designing interventions aligned with human psychology.[S2]
The finding is therefore about a change in capability. Influence can be personalised, responsive and persistent. Society has long known that advertising, institutions and peers shape choices. What is different is the ability to make that shaping operational at scale, inside the ordinary services through which people communicate, shop, work and seek help.
The Helpful and Harmful Versions Share a Mechanism
It would be a mistake to treat behavioural technology as inherently harmful. The same design mechanism can reduce needless effort. A clear reminder can support a decision someone has already made. Remote mental-health services can extend access beyond a clinician’s office, including to people facing geographic or logistical barriers.[S5]
Psychology also has a constructive role in technology design. The American Psychological Association argues that psychological research can help people engage safely with innovations from AI to social media and gaming. Its discussion of AI emphasises that technology needs more than a technical solution: values and ethics matter too.[S1]
The problem is not that a system makes a choice easier. The problem begins when ease is assigned selectively: when the path that serves an organisation is effortless while the path that serves the user is obscure, delayed or exhausting. A helpful reminder is legible and proportionate. A manipulative prompt is designed to be hard to notice as influence, difficult to decline, or costly to escape.
Intentions matter, but outcomes and power matter as well. A person may welcome a prompt to protect an account or attend an appointment. They may not welcome a pattern that turns a moment of distraction into more disclosure, more spending or more time than intended. The interaction can look trivial on its own. Repeated across many decisions, it can reshape the practical conditions of daily agency.
Why Informed Consent Is No Longer Enough
For years, the standard consumer response to digital risk has been to read the policy, adjust the settings and make an informed choice. That approach assumes that a person can reasonably manage each application’s permissions and foresee the future consequences of sharing data. The evidence supplied by current practice points in the opposite direction: privacy permissions form a complex patchwork, and specialists argue that privacy should be designed so consumers are not expected to manage each individual application.[S1]
This is not an argument for helplessness. It is an argument for realism. A person can make a thoughtful decision only when the decision is comprehensible, timely and reversible enough to be meaningful. A lengthy notice at sign-up cannot neutralise a stream of later prompts. A buried setting cannot balance a default that has already directed behaviour. A nominal opt-out is weak if it requires patience, specialised knowledge or repeated resistance.
The issue becomes sharper with data. Smartphones and other ubiquitous technologies can gather information across many settings, and privacy can become a commodity more available to people with greater resources.[S1] When protection depends on time, literacy, money or access to expert advice, the burden falls unevenly. A society that treats privacy as an individual consumer skill risks making unequal protection appear like individual failure.
There is also a durability problem. The APA discussion notes that even where consumers have a right to be forgotten, removing data from the influence and output of machine-learning algorithms remains an unresolved implementation challenge.[S1] Ordinary people should read that as a reason for caution before disclosure, not as a reason to panic. It means that prevention often provides more control than later cleanup.
The Social Stakes Are Distributional
Technology–psychology systems do not affect everyone on equal terms. A model that fails to represent the populations it serves can amplify bias and discrimination. The APA points to early police facial-recognition systems that were not trained on darker skin tones and produced errors leading to wrongful arrests.[S1]
That example is important because it moves the conversation beyond personal preference. A poorly designed recommendation or prompt may be annoying. A poorly designed decision system used in employment, public services, policing or healthcare may change opportunities, scrutiny and consequences. The same principle applies: systems encode judgments about what to notice, predict and optimise. Those judgments can be distributed unevenly.
Representation in design is therefore not a public-relations extra. It is part of whether a technology can be trusted. The APA records a concern that powerful AI systems are being built by a tiny minority of the world’s population, and that broader participation would improve their design.[S1] For the public, trust has two dimensions: whether a system is safe, and whether the people and institutions using it deserve trust with the system and its data.[S1]
The social consequence is that digital self-defence cannot be the whole answer. Individual habits are useful, but they cannot correct a system that offers one group less privacy, makes errors about another group, or turns essential access into an obstacle course. Public rules, organisational accountability and inclusive design are not alternatives to personal agency; they are conditions that make personal agency possible.
What Ordinary People Can Do Without Becoming Experts
The goal is not to inspect every line of a privacy policy or abandon technology. It is to restore a small pause between a system’s prompt and a person’s response. A useful first habit is to identify the action a screen is trying to make automatic. Is it asking for more data, a renewal, a purchase, a notification permission or another minute of attention? Naming the intended action makes the design visible.
Next, use friction deliberately. Turn off non-essential notifications. Avoid completing consequential choices from a lock-screen prompt. Revisit permissions when an app asks for access that does not obviously support its purpose. Treat a default as a starting point chosen by someone else, not as evidence that it is right for you. These are modest actions, but they change the pace at which a system can turn a cue into a decision.
Be especially deliberate where a decision is hard to reverse. The current difficulty of removing data’s influence from machine-learning systems means it is sensible to distinguish between information needed to use a service and information that merely improves its targeting, profiling or convenience.[S1] The relevant question is not whether sharing is always wrong. It is whether the benefit is clear enough to justify a potentially durable loss of control.
Finally, make ethics a selection criterion. Prefer services that explain their prompts and choices plainly, make privacy controls usable and let people opt out without punishment. The City Bar report identifies transparency, opt-out and respect for autonomy as central to ethical digital nudging.[S2] Consumers cannot solve structural problems alone, but their choices can reward products that do not treat confusion as a business model.
What Organisations Should Be Required to Prove
The burden should not rest only on users. Organisations that deploy behavioural systems should be able to explain the goal of an intervention, the data used to personalise it, the people affected and the route to refusal. They should test whether an apparently helpful feature creates pressure, confusion or unequal effects for particular groups.
This is not anti-innovation. The City Bar report frames the central task as balancing progress with ethical integrity while safeguarding privacy and fairness. It also calls for controls that strengthen accountability, trustworthiness and reliability, including attention to the accuracy and precision of analytics and the quality of data sources.[S2]
In practical terms, that standard changes the question from “Can we improve engagement?” to “What behaviour are we trying to change, for whose benefit, by what means, and with what exit?” It asks product teams to distinguish support from extraction. It asks leaders to accept that a technically effective prompt can still be socially unacceptable.
Psychological expertise should enter early, before an interface or model becomes difficult to change. The APA’s account of technology development makes the case for psychologists as contributors to safer innovation, privacy design and equitable systems rather than as after-the-fact critics.[S1] That is a promising direction only if their role includes the power to challenge what the system is optimising.
A Better Standard for Digital Progress
The tempting story is that technology will either solve human problems or corrupt human life. The stronger finding is less dramatic and more useful: technology increasingly operates through the conditions of human decision-making. That means its effects depend on design, context, incentives and governance.
A better standard for progress is not simply more personalised, predictive or persuasive technology. It is technology that enlarges a person’s capacity to act on their own considered purposes. It makes important choices understandable. It makes refusal real. It does not reserve privacy for the well resourced. It treats errors and unequal effects as design failures to address, not collateral damage to accept.
That standard also clarifies why the debate cannot be left to engineers alone. Psychological research explains how people respond to systems; social institutions determine where those systems are used and what they are permitted to optimise. Neither perspective is sufficient by itself. The central work is to join them without allowing efficiency to swallow autonomy.
Conclusion
The most important current Technology × Psychology × Society finding is that influence has become an infrastructure. Data and digital systems can make behavioural interventions more continuous, tailored and scalable, while psychological knowledge makes those interventions more likely to work.[S2] This capacity can support access, safety and follow-through. It can also exploit attention, weaken privacy and reproduce unequal treatment.
For ordinary people, the response begins with a practical shift: see defaults, prompts and permissions as designed choices rather than neutral background. Pause before irreversible disclosure, add friction where a service adds pressure, and favour products that make control understandable. For institutions, the obligation is larger: build for transparency, meaningful refusal, privacy, fairness and accountability. Technology will keep shaping the conditions in which people decide. The public question is whether it will do so in ways that help people remain authors of their own lives.