For most of human history, authority was visible. Guards. Walls. Laws with named penalties. You knew when power touched you.

What is emerging now operates differently. Behavioural scoring — converting a person's observable actions, associations, and data trails into numerical rankings that determine access to resources, services, and freedoms — replaces the visible instrument with an invisible one. No officer inspects your papers. A model updates silently, drawing from thousands of data inputs, consulted by systems you will never directly encounter.

Three forces have arrived at the same moment. Near-total digital legibility: most human behaviour now leaves a data trace. Unprecedented computational capacity to analyse that trace in real time. And political and commercial incentives to act on what the analysis reveals. Together, they have created infrastructure for a kind of power that does not simply respond to what you have done. It attempts to predict and pre-emptively shape what you will do next.

The past offers warnings. Panopticon prisons were designed so inmates could never tell whether they were being watched. Internal passport systems in the Soviet Union controlled movement through categorisation. Apartheid-era pass books sorted people by perceived social worth. Each was a technology of ranking and restriction. The difference today is scale, subtlety, and speed. A pass book required a human officer. A behavioural score requires no one.

What do you actually know about China's social credit system — or what you think you know?

Almost nothing about it is as simple as Western coverage has suggested. The phrase implies a single, unified, nationally integrated apparatus producing one master score for every Chinese citizen. What exists, as of the mid-2020s, is considerably messier.

Multiple overlapping systems operate under different levels of government and private commercial entities. Different data inputs. Different purposes. Inconsistent coverage. Some are sectoral — targeting businesses or financial institutions, not individuals. Some are city-level experiments. Some focus on court-ordered debt obligations. The nationally integrated individual-scoring system that Western articles routinely describe remains, according to researchers who have studied it closely, a partial and fragmented reality rather than a completed fact.

What is documented: blacklists and whitelists that restrict air and rail travel for people found by courts to have violated financial judgments or legal obligations. Municipal systems in cities like Rongcheng that assign points based on traffic violations, business compliance, and community recognition. Corporate social credit systems evaluating business regulatory compliance. These are meaningful instruments of state power. But the image of an omniscient score governing every Chinese person's every decision is a significant simplification.

This matters for a reason that should make Western readers uncomfortable. The oversimplification serves a rhetorical function. It frames behavioural scoring as a foreign and exotic threat — which makes it easier not to look at what is being built at home.

In the United States, the FICO score has governed access to housing, credit, and economic opportunity since the 1980s. It is, in its basic structure, exactly what critics fear in Chinese social credit: a numerical representation of a person's behaviour, used by institutions to make access decisions, updated continuously, and largely opaque to those it affects. The fact that a private company operates it rather than a government does not make it less powerful. In many cases it makes it less accountable, because private companies are not subject to the constitutional constraints that theoretically limit state power.

Beyond FICO, an expanding ecosystem of behavioural data products now operates in everyday American life. Tenant screening algorithms draw on eviction records, credit data, and criminal history to determine whether a person may rent an apartment. Insurance telematics programs use GPS tracking and driving behaviour data to adjust premiums in real time. Workplace monitoring software tracks keystrokes, mouse movements, and calendar behaviour to produce productivity scores. Fraud detection systems at banks flag unusual transaction patterns and can freeze accounts without notice or appeal. Content engagement scores determine whose speech propagates across social media platforms and whose is quietly suppressed.

None of these are called social credit. Each is presented in the language of its own domain — financial prudence, risk management, productivity, security, relevance. But together they form a distributed social credit infrastructure: a web of overlapping behavioural scoring systems that, in aggregate, determine who gets housing, work, credit, audience, and access to public and private services.

Three claims usually separate Western systems from Chinese ones: that Western systems are private-sector rather than state-run; that they are voluntary, because participation in the services that generate data is optional; and that they are subject to democratic oversight. Each claim is weaker than it appears.

Private-sector systems produce the same behavioural control effects as state systems. Opting out of digital life is not optional in any meaningful sense. And democratic oversight of algorithmic systems remains, in most jurisdictions, rudimentary at best.

Why are behavioural scoring systems proliferating so rapidly? The answer is economic before it is political.

Shoshana Zuboff, in her analysis of what she calls surveillance capitalism, identified a new economic logic that emerged from Silicon Valley in the early 2000s and has since spread globally. Human behavioural data is extracted as a raw material. It is processed into predictive products. Those products are sold to advertisers and institutional customers who want to influence human behaviour.

The critical alarm in this framework: the goal is not merely to observe behaviour but to modify it. Predictive products are more valuable when they are more accurate. They are more accurate when the people being predicted can be nudged toward the behaviours the model anticipated. This creates a structural incentive for platforms to engineer the environments in which behaviour occurs — to shape choice architecture, emotional states, and social contexts so that human behaviour becomes more predictable and therefore more profitable.

The score is not just a measurement. It is part of a feedback loop. You are scored on your behaviour. Your score affects what you see, what you are offered, and what options are available to you. The options available shape your subsequent behaviour. Your subsequent behaviour updates your score. At every stage, the system learns which interventions most reliably produce the behaviours its operators prefer.

Zuboff called this the behavioural futures market: the buying and selling of predictions about what people will do. Social credit systems — state-run or corporate — are not a departure from this logic. They are one of its more advanced expressions.

When a score produces an unjust outcome, who is accountable?

A person is denied housing because a tenant screening algorithm flagged them. The landlord may not have understood how the score was generated. The company that produced the score may claim it is simply reporting historical data. That data may come from public court records containing errors, or reflecting historically biased enforcement patterns. The algorithm may have amplified those patterns in ways no individual engineer deliberately intended. At every link in the chain, responsibility disperses.

This dispersal is not accidental. Algorithmic decision-making transfers authority from human agents — who can be named, questioned, sued — to systems described as neutral and objective. The language of objectivity is powerful. A score feels more legitimate than a judgment because it appears to transcend individual bias. But algorithms are not neutral. They encode the assumptions of their designers, reproduce the patterns in their training data, and optimise for the outcomes their operators have decided to value.

The appearance of objectivity can be a more effective way to launder bias than overt discrimination, precisely because it is harder to challenge.

The European Union's General Data Protection Regulation (GDPR) includes provisions requiring that individuals have the right not to be subject to decisions made solely by automated processing in certain circumstances, and that they receive meaningful explanations when decisions affecting them are made. How meaningful these protections are in practice remains contested. The EU's AI Act, which came into force in 2024, establishes risk categories for AI systems and imposes additional requirements on those classified as high-risk — including systems used in credit scoring, employment, and access to essential services. Whether this regulatory framework can match the pace of technological development and the power of economic interests pushing in the opposite direction is an open question. A genuinely open one.

The punitive dimensions of social credit get most of the attention. The positive side is equally worth examining.

Most social credit systems — including many documented in China — are not primarily systems of punishment. They are systems of incentivised compliance: they offer rewards for rated-good behaviour as much as they impose penalties for rated-bad behaviour.

This logic is not unfamiliar in liberal societies. Tax credits reward charitable giving. Reduced premiums reward safe driving. Vaccination incentives reward public health participation. The structure is the same. What changes at scale, and with algorithmic precision, is the comprehensiveness of the incentive structure and the granularity of the behaviours it targets.

When every micro-behaviour potentially affects a score that affects meaningful life outcomes, the pressure to perform compliant behaviour extends into every corner of life. This is the gamification of virtue: the conversion of ethical and social conduct into a scored game with real stakes. It creates powerful pressures toward conformity — not because non-conformity is directly prohibited, but because it is expensive. And because the rules of the game are set by whoever controls the scoring system, the definition of virtuous behaviour becomes, in practice, whatever behaviour those controllers prefer.

This creates a precise political problem. Democratic societies are premised on the idea that citizens can hold minority views, engage in legal dissent, and live unconventionally without losing their standing. Scoring systems that attach costs to behaviours associated with dissent — attending certain protests, associating with certain people, expressing certain opinions online — can penalise political heterodoxy without criminalising it. No law is broken. The score simply moves.

This concern is documented, not theoretical. Discriminatory credit scoring linked to zip codes — which correlate with race — has been extensively studied. Social media background screening by employers has penalised association with activist groups. Predictive policing algorithms have intensified surveillance of communities based on historical arrest patterns rather than actual crime rates. In each case, social and political identity is converted into a risk signal through apparently neutral technical means.

Behavioural scoring systems become qualitatively more powerful when they can draw on biometric data — data derived from the body itself rather than from the traces behaviour leaves in digital systems.

Facial recognition technology is the most widely discussed example. When surveillance cameras equipped with facial recognition are combined with behavioural databases, the result is a system capable of tracking individuals through physical space in near-real time, connecting their offline presence to their digital records. China has deployed facial recognition extensively in public spaces, transit systems, and entry points to facilities. It has been used to identify people crossing roads against traffic signals. It has been used to screen mosque-goers in Xinjiang — part of a system of mass surveillance targeting Uyghur Muslims that researchers, journalists, and multiple governments have characterised as ethnic persecution.

Facial recognition is also in active deployment in democratic countries. Police departments in the United States and United Kingdom use it. Private venues including sports arenas and music halls use it. Border control agencies in multiple countries use it. Retailers use it for loss prevention. Studies have consistently found that facial recognition accuracy varies significantly by race, with substantially higher error rates for darker-skinned faces, women, and older individuals. This is not a minor technical problem. Several documented cases in the United States involve people arrested based on incorrect facial recognition matches.

Beyond facial recognition, emerging biosignal monitoring technologies — wearables tracking heart rate, galvanic skin response, and other physiological indicators — are being deployed in workplace settings to measure stress, engagement, and emotional state. The logic of behavioural scoring is being extended from what you do to what you feel.

That is a meaningful escalation.

The picture so far is demanding. Stopping there would also be dishonest.

Legal resistance is active and has produced real outcomes. San Francisco, Boston, and Portland have banned or sharply restricted police use of facial recognition. The EU's AI Act creates enforceable requirements around high-risk AI systems, including transparency and human oversight provisions. Brazil's Lei Geral de Proteção de Dados (LGPD) extends meaningful data rights to Brazilian citizens. India's data protection legislation is being watched as a potential model for large democracies navigating these questions. These are not solutions. They are evidence that political and legal responses to algorithmic power are possible.

Algorithmic auditing — independently examining how scoring systems work and what effects they produce — is emerging as both a technical discipline and a regulatory requirement. Researchers at academic institutions and civil liberties organisations have used auditing methods to document racial bias in recidivism scoring tools used in US criminal sentencing, discriminatory patterns in healthcare allocation algorithms, and unequal treatment by lending algorithms. These findings have, in several cases, prompted regulatory attention and changes to the systems involved.

Data minimisation — the principle that systems should collect only what they need for their stated purpose — is central to European data protection law and is increasingly advocated as a design principle rather than just a legal requirement. A scoring system can only incorporate the behavioural signals it has access to. Limit the data. Limit the reach.

Collective bargaining and labour law are being tested as tools for addressing workplace monitoring and productivity scoring. Unions in multiple countries are pushing for consent requirements, transparency, and limits on how monitoring data can be used in employment decisions. This frames the problem not only as a technical or regulatory issue but as a question of power relations — one in which workers have standing to negotiate.

No single mechanism is sufficient. Behavioural scoring systems are driven by economic incentives powerful enough to outpace most regulatory responses, by technical capabilities genuinely difficult to govern without understanding them, and by political dynamics in which the institutions most eager to deploy these systems are also among the most influential voices in shaping policy. Effective resistance requires legal frameworks, technical accountability tools, organised civil society pressure, and a public that understands what is at stake — all operating simultaneously.

The infrastructure of obedience being constructed around us is not being built by people who want to control humanity for its own sake. It is being built, one plausible-seeming decision at a time, by engineers solving real problems, executives responding to real commercial incentives, policymakers managing real risks, and institutions attempting to scale their operations efficiently. This is precisely what makes it so difficult to resist and so important to understand.

The question is not whether behavioural data will be used to make decisions about people. It already is, everywhere, at scale. The question is whether those decisions will be made accountably or invisibly. In ways that can be contested, or in ways that are structurally insulated from challenge. With human dignity as a design constraint, or as an afterthought.

Those are choices. They will be made. The only variable is whether they are made consciously — with the full force of democratic deliberation — or by default, in the expanding silence between data collection and its consequences.