Social Scoring

In social scoring, an authority or a company collects data about people’s behavior. From this, a score is calculated that is meant to indicate how trustworthy or reliable someone is. All sorts of things often go into the calculation: bills, traffic violations, entries in social networks, sometimes even one’s association with certain people. Anyone with a low score then faces disadvantages, for example when applying for a loan, a job, or booking a trip. What matters is that the score does not apply to just a single matter. It follows the person across many areas of life and evaluates them as a whole.

Why the EU banned it

Social scoring is considered one of the few techniques in Europe that lawmakers wanted not merely to regulate, but to prohibit outright. The reason is its effect on people’s freedom. Anyone who knows that every statement feeds into a permanent score behaves differently. People say less, take fewer risks, and avoid inconvenient contacts. Experts call this effect self-censorship.

A second problem lies in the data itself. Scoring systems inherit the biases embedded in their data. Someone living in a poorer neighborhood is statistically more likely to receive a poor rating, even without any personal misconduct. The score then appears objective, but in fact reinforces existing inequality. And because it applies everywhere, there is hardly any way to escape it.

The EU’s AI legislation, the so-called AI Act, therefore explicitly prohibits such systems. The ban has been in effect since February 2025 and applies to both authorities and private companies. Violations can be punished with fines of up to 35 million euros or seven percent of global annual revenue. This puts social scoring on the same level as other prohibited practices, such as the deliberate exploitation of vulnerabilities of people in need of protection.

From data point to score

Technically, social scoring is not a particularly new invention. First, data from various sources is combined: administrative records, payment data, movement data, online entries. Then a program searches this data for patterns associated with desired or undesired behavior. In the end, this produces a single number or a grade such as A through D.

The sensitive step is the combination of data. Individual pieces of data are harmless on their own; only the linkage produces a complete profile. A program that knows a person’s payment behavior, place of residence, and circle of friends at the same time knows more about that person than any single authority. This is precisely what distinguishes social scoring from an ordinary database.

It is important to distinguish this from creditworthiness checks, such as those performed by Schufa. There, too, a score is generated, but it is only allowed to answer one question: Is this loan likely to be repaid? What is prohibited is the leap beyond that. If payment behavior suddenly determines access to a daycare spot, the line has been crossed. Experts refer to this as an evaluation outside its original context.

From China to insurance tariffs

The best-known example is the Chinese Social Credit System. In public debate, it is often portrayed as a nationwide score for every citizen. In fact, it consists of many regional projects with very different rules, some of which have since been discontinued. What is documented, however, is that people placed on blacklists were unable to purchase flight or train tickets.

In Europe, the term mostly appears in news about laws and government agencies. Discussions include, for example, whether insurance tariffs based on driving behavior or fitness data are still permissible. Rating systems for drivers at delivery services are also under criticism, because a poor rating can cost them orders.

A common misconception is that any rating on the internet constitutes social scoring. Star ratings for restaurants or sellers are not. They relate to a specific service and have no consequences outside the platform. It only becomes social scoring when an assessment determines a person’s access to society at large.