Avoiding Bias and Discrimination

Artificial intelligence learns from vast amounts of data and can therefore unintentionally reproduce human prejudices (bias) or stereotypical patterns. To guarantee fair, unbiased, and objective case processing at the dispute resolution office, the following principles must be strictly observed when using the AI:

• Checking for equal treatment: The results generated by the AI – particularly for translations, email drafts, or text simplification – must always be checked to ensure they do not unduly disadvantage either party (applicant or respondent). No discrimination based on gender, ethnic origin, language, religion, age, or social origin may occur during case processing.
• Balanced summaries: When using the AI summary of a case file, case handlers must ensure that the arguments and facts of both sides are represented objectively and equally. The AI must not cause one party's position to be disproportionately weighted or the context to be distorted.
• Neutral inputs (prompts): When manually entering information or asking questions to the chatbot, pay attention to using factual and neutral language. Avoid judgmental or emotionally charged descriptions of the parties involved, so as not to steer the AI toward a biased direction.
• Human responsibility: The AI has no ethical or moral authority. Sole responsibility for discrimination-free, neutral, and legally compliant communication throughout the entire dispute resolution procedure always lies with the person handling the case.