Contextusban

Contextusban is a policy concept in data governance and AI ethics that refers to a ban or restriction on using contextual information in automated decision-making. It aims to limit the influence of situational data such as user environment, timing, location, device characteristics, or inferred context, unless explicitly provided as input features.

The term is a neologism formed from contextus (Latin for context) and ban, and has appeared in scholarly discussions in the early 2020s. It is not widely standardized and definitions vary by authors and jurisdiction.

In practice, contextusban can be implemented through governance rules, model constraints, and privacy-preserving techniques. Examples include feature-level restrictions, data minimization, removing context-related features from training and inference, or enforcing strict access controls to prevent context leakage.

Proponents argue it improves fairness, reduces bias, and protects sensitive attributes by preventing models from leveraging contextual cues that correlate with protected classes. Critics warn that over-broad bans can degrade performance or hide legitimate context needed for safety or accuracy, and that enforcement can be challenging given evolving definitions of context.

The concept sits alongside related ideas such as contextual integrity, data minimization, and differential privacy. It is often discussed in the context of advertising, credit scoring, and automated risk assessment, where contextual cues may otherwise influence outcomes. Historically, contextusban remains a debated tool rather than a universally adopted standard.