Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
Ethics by Design
Methodological approach that systematically integrates ethical considerations from the initial design phase of AI systems, rather than as a post-hoc correction.
Algorithmic Justice
Fundamental principle aimed at ensuring that AI systems produce fair and non-discriminatory outcomes for all concerned groups.
Predictive Fairness
Quantitative measure evaluating whether a model's prediction errors are distributed fairly across different demographic groups.
Explainability
Quality of an AI model that allows human users to understand the specific reasons behind each individual decision or prediction.
Value-Centered Design
Development methodology that places human and ethical values at the center of technical and decision-making design processes.
Algorithmic Impact Assessment
Structured process of prospective analysis of the potential effects of an AI system on individuals, communities, and society.
Data Governance
Organizational framework defining policies, procedures, and responsibilities for the ethical and secure management of data used in AI.
Data Diversity
Principle ensuring balanced and complete representation of all relevant groups in training datasets.
Continuous Ethical Monitoring
Process of continuous monitoring of ethical performance of an AI system in production to detect and correct drifts.
Bias Testing
Set of statistical and analytical techniques aimed at quantitatively identifying systematic discriminations in AI models.
Bias Mitigation
Set of proactive techniques applied to data, algorithms or results to reduce or eliminate identified discriminations.