This page is still being polished. If you have thoughts, please share them via the feedback form.
Data on this page is preliminary and may change. Please do not share or cite these figures publicly.
Design-time architectural choices affecting safety, interpretability, and modularity.
Also in Model
For complex, ‘black-box’ models like deep neural networks, achieving explainability requires the use of post-hoc interpretation techniques.
Reasoning
Foundational research investigating model interpretability and decision-making mechanisms post-training.
Fairness Metrics
robust fairness metrics, such as demographic parity and equalized odds, to rigorously evaluate and quantify a model's performance across different populations.
2.2.2 Testing & EvaluationSystematic Bias Auditing
The systematic auditing for and mitigation of these biases are not merely corrective measures but are fundamental to the system's legitimacy and social acceptance.
2.2.3 Auditing & ComplianceTransparency
Transparency refers to the degree to which the inner workings of an AI system: its data, algorithms, and models are accessible and comprehensible.
2.4.2 Design StandardsExplainability
Explainability, a related but distinct concept, pertains to the ability to furnish a clear, human-understandable rationale for a specific decision or prediction made by the system.
1.1.4 Model ArchitectureAccountability Structures
“Establishing accountability requires the creation of clear, pre-defined structures that assign responsibility for the system's behavior to specific human actors or organizational entities.”
2.1.2 Roles & AccountabilityLogging and Audit Trails
Mechanisms such as detailed logging, immutable audit trails, and designated ethics officers are essential for creating a framework where the actions of AI can be traced back, and responsible parties can be held to account.
2.1 Oversight & AccountabilityEthical Imperatives in AI Design: A Comprehensive Framework for Risk Mitigation and Responsible Innovation
Tariq, Bilal; Ashraf, Muhammad Rehan; Rashid, Umar (2025)
As artificial intelligence (AI) becomes increasingly integral to critical sectors, the gap between abstract ethical principles and their concrete technical implementation presents a significant barrier to responsible innovation. This paper addresses this challenge by introducing a comprehensive framework designed to embed ethical considerations directly into the AI development lifecycle.
Operate and Monitor
Running, maintaining, and monitoring the AI system post-deployment
Developer
Entity that creates, trains, or modifies the AI system
Manage
Prioritising, responding to, and mitigating AI risks
