[Paper Review] Responsible Artificial Intelligence -- from Principles to Practice

Virginia Dignum|arXiv (Cornell University)|May 22, 2022

Ethics and Social Impacts of AI4 citations

TL;DR

This paper proposes a comprehensive framework for responsible AI that integrates ethical principles into technical design through value-sensitive engineering, emphasizing transparency, accountability, and stakeholder inclusion. It introduces the Glass Box framework to operationalize trustworthiness by linking societal values to system requirements, shifting focus from performance alone to value-aligned, explainable AI systems.

ABSTRACT

The impact of Artificial Intelligence does not depend only on fundamental research and technological developments, but for a large part on how these systems are introduced into society and used in everyday situations. AI is changing the way we work, live and solve challenges but concerns about fairness, transparency or privacy are also growing. Ensuring responsible, ethical AI is more than designing systems whose result can be trusted. It is about the way we design them, why we design them, and who is involved in designing them. In order to develop and use AI responsibly, we need to work towards technical, societal, institutional and legal methods and tools which provide concrete support to AI practitioners, as well as awareness and training to enable participation of all, to ensure the alignment of AI systems with our societies' principles and values.

Motivation & Objective

• Address the growing gap between ethical AI principles and their practical implementation in real-world systems.
• Counter the misconception that AI is inherently intelligent or magical, by clarifying its nature as a data-driven, algorithmic tool.
• Promote inclusive, value-sensitive design by involving diverse stakeholders in defining ethical priorities and system requirements.
• Develop technical and socio-legal methods to ensure accountability, transparency, and fairness in AI systems.
• Reframe responsibility and regulation not as barriers to innovation, but as essential drivers of trustworthy and sustainable AI development.

Proposed method

• Adopt a Design for Values approach to translate abstract ethical principles (e.g., fairness, privacy) into concrete system requirements.
• Introduce the Glass Box framework as a modular, verification-integrated methodology for building transparent and explainable AI systems.
• Emphasize stakeholder involvement in value elicitation, prioritization, and documentation of design decisions.
• Integrate socio-technical considerations into the AI development lifecycle, including data governance, power dynamics, and societal impact.
• Shift focus from performance-centric AI development to value-centric design, incorporating causality, reasoning, and prior knowledge.
• Use structured moral deliberation methods (e.g., algorithmic, user-controlled, or regulatory) to operationalize ethical choices in system behavior.

Experimental results

Research questions

• RQ1 How can ethical principles such as fairness, transparency, and privacy be systematically translated into technical system requirements?
• RQ2 What methods enable inclusive and transparent value elicitation across diverse stakeholders in AI design?
• RQ3 How can AI systems be built to be explainable and verifiable without sacrificing performance or scalability?
• RQ4 In what ways does the current narrative around AI—framed as magic or inevitable progress—hinder responsible development?
• RQ5 How can responsibility and regulation be reframed as enablers of innovation rather than obstacles in AI development?

Key findings

• There is a global convergence on five core ethical principles: Transparency, Justice and Fairness, Non-Maleficence, Responsibility, and Privacy.
• Despite widespread agreement on ethical principles, their practical interpretation and implementation vary significantly across organizations and guidelines.
• AI is not intelligent or magical; it is a data-driven, statistical tool reliant on patterns, input data quality, and design choices—making data and algorithmic decisions deeply political.
• The Glass Box framework enables modular, verifiable, and transparent AI systems by integrating value-based design with formal verification methods.
• Shifting from performance-driven to value-driven AI development requires rethinking AI innovation to include causality, abstraction, and prior knowledge.
• Responsible AI is not just about mitigating bias or ensuring fairness—it is a holistic, systemic transformation of how AI is conceived, built, and governed.