Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
Pre-deployment Ethical Audit
In-depth examination of algorithmic components, data, and decision-making processes to detect biases and discriminations before production deployment.
Ethical Impact Matrix
Structured assessment tool for quantifying and visualizing potential ethical consequences across multiple dimensions and stakeholders.
Bias Risk Score
Quantitative indicator measuring the probability and severity of algorithmic discriminations based on protected characteristics.
Procedural Fairness Test
Assessment checking whether AI decision-making processes treat all individuals fairly regardless of their personal characteristics.
Ethical Sensitivity Analysis
Method for examining variations in the system's ethical outcomes in response to changes in input data or model parameters.
Stakeholder Mapping
Process of identifying and analyzing all actors affected by the AI system to assess potential differentiated impacts.
Societal Impact Framework
Holistic assessment structure measuring collective and long-term effects of AI deployment on social cohesion and institutions.
Algorithmic Transparency Metric
Set of indicators quantifying the ability of users and regulators to understand and interrogate AI decision-making mechanisms.
Ethical Validation Protocol
Standardized series of tests and verifications ensuring system compliance with established ethical principles before deployment.
Distributive Justice Indicator
Measure evaluating the equity of benefit and burden distribution of the AI system among different demographic groups.
Ethical Robustness Assessment
Test of the system's ability to maintain its ethical guarantees in the face of manipulation attempts or adverse conditions.
Non-Discrimination Test
Systematic verification of the absence of unjustified differential treatment based on protected characteristics in algorithmic decisions.
Accountability Framework
Organizational structure defining responsibilities, oversight mechanisms, and recourse pathways for AI decisions and impacts.
Human Rights Impact Assessment
Systematic analysis of the potential effects of the AI system on the exercise of internationally recognized fundamental rights.
Consequentialist Analysis Method
Evaluation approach focusing on the actual outcomes and consequences of the system rather than its intentions or intrinsic principles.
Systemic Bias Assessment Protocol
Comprehensive methodology to identify emergent discriminations from the complex interaction between data, algorithms, and usage contexts.
Algorithmic Vulnerability Matrix
A mapping tool identifying populations and situations at high risk of negative impact from system automated decisions.
Predictive Fairness Test
Statistical evaluation of the model's prediction consistency across different groups to ensure equitable performance.
Regulatory Impact Assessment Framework
An analytical structure ensuring the AI system's compliance with legal and regulatory requirements applicable in its deployment domain.