Glosarium AI
Kamus lengkap Kecerdasan Buatan
Procedural Justice
Ethical principle evaluating the fairness of algorithmic decision-making processes regardless of their final outcomes. It focuses on transparency, consistency, and impartiality of methods used by AI systems.
Decision Explainability
Ability of an AI system to provide clear and justifiable interpretations of its individual decisions. Explainability allows users to understand the factors that influenced a specific algorithmic decision.
Equality of Opportunity
Algorithmic fairness principle ensuring that true positive rates are equal across different demographic groups. Equality of opportunity ensures that qualified individuals have the same probability of being selected regardless of their group.
Equalized Odds
Extension of equality of opportunity requiring that both false positive and false negative rates be equal between groups. This stricter metric aims to eliminate all forms of discrimination in algorithmic errors.
Algorithmic Accountability
Mechanism establishing responsibility for algorithmic decisions with identifiable and accountable entities. Accountability includes documentation, traceability, and the possibility of recourse against automated decisions.
Group Fairness
Algorithmic fairness approach evaluating system performance and impacts at the level of entire demographic populations. It contrasts with individual fairness which focuses on specific cases.
Model Calibration
Process of adjusting a model's probability scores to accurately reflect true occurrence frequencies. Proper calibration is essential for ensuring fair and interpretable decisions.
Decision Consistency
Procedural justice principle requiring that similar cases produce consistent decisions over time and between different algorithmic evaluators. Consistency strengthens the predictability and legitimacy of the system.
Contextual Fairness
Adaptation of algorithmic fairness criteria based on specific application contexts and relevant societal standards. Contextual fairness recognizes that definitions of fairness vary according to domains of use.
Procedural Ethics
Normative framework evaluating the morality of development and deployment processes of AI systems. It examines stakeholder participation, governance, and oversight mechanisms throughout the algorithmic life cycle.