This summary follows the path from an impression to a judgement and sometimes to a decision. It is not legal advice. Legal position and retrieval date: 13 November 2026.
The chain at a glance
The foundation is categorisation. Three transitions build on it: category to stereotype, stereotype to prejudice, prejudice to discrimination. The chain is a model for ordering ideas, not an automatism. Prejudice does not necessarily lead to discrimination, and discrimination can arise without personal prejudice, for example through institutional rules.
Three steps
A stereotype assigns traits to people solely because of group membership. A prejudice, in Gordon Allport's 1954 formulation as reported by Sibley and Barlow (2016), is an antipathy based on a faulty and inflexible generalisation. Both are attitudes. Discrimination is the act: denying people rights or opportunities because they belong to a group.
What German law says
The General Equal Treatment Act (AGG) aims, under section 1, to prevent or eliminate disadvantage on grounds of race (the statutory wording), ethnic origin, sex, religion or belief, disability, age or sexual identity. It has been in force since 18 August 2006 and was last amended by Article 15 of the Act of 22 December 2023.
Direct disadvantage means less favourable treatment than another person in a comparable situation (section 3(1)). Indirect disadvantage means apparently neutral provisions, criteria or procedures that can put persons at a particular disadvantage, unless a legitimate aim justifies them and the means are appropriate and necessary (section 3(2)). Neither mentions intent. Under section 7(1), the ban also applies where the ground is only assumed.
A bill amending the AGG (Bundestag paper 21/6178) had a public hearing on 5 October 2026. No adoption was recorded on 13 November 2026; the existing text applies.
Why knowledge alone does not protect
People tend to see bias in others more readily than in themselves (Pronin, Lin and Ross, 2002). Studies on training are mixed:
- Forscher and colleagues (2019), 492 studies with 87,418 participants: implicit measures changed, but effects were often small and short-term, with no clear transfer to behaviour.
- Paluck and colleagues (2021), 418 experiments: 76 percent were short interventions with unclear long-term effect.
- Bezrukova and colleagues (2016), 260 samples: overall g = .38, weaker for behaviour and attitudes, larger when training was embedded and longer.
Knowing the chain is therefore a precondition, not protection, so attention belongs also to criteria, procedures and decisions. That is an inference, not a proven effect.
This is a summary. The full German article with all details is at the same address under the German site.