IEC and ISO have published a new Technical Report (TR) that provides advice on addressing ethical and societal concerns about artificial intelligence. Intended for technologists, regulators and society at large, it takes an objective and practical approach.
"Our committee took the novel approach of concurrently working on enabling broad adoption by looking at the entire AI ecosystem and simultaneously addressing ethical and societal concerns," said Wael William Diab who chairs the IEC and ISO joint committee on AI. "Ethical considerations are tightly integrated across the entire work program of the committee. This project provides the blueprint of how it all comes together to enable responsible AI adoption and its benefits."
Indeed, the growing program of work within the committee not only has dedicated projects such as this one on ethics and societal concerns but enables and supports responsible adoption across the portfolio of work. For instance, use cases, governance, foundational, trustworthiness, the management systems standards and other projects directly and indirectly address this topic. This novel approach means that the suite of standards that the committee works on is ready address this topic from day one, highlighting the importance of responsible adoption. "This publication is for all AI stakeholders," says the project convenor, David Filip. "It gives them practical guidance on ethical frameworks, risk mitigation and best practices."
"Handled properly, AI can radically transform the world for the better," says the project leader, Viveka Bonde. "ISO/IEC TR 24368 will help regulators and decision-makers who often struggle to keep up with the rapid evolution of the technology."
"This is an essential and practical guide for anyone interested in acquiring a greater understanding about the responsible adoption of AI technologies. It is non-partisan and objective," said ISO/IEC expert Lenora Zimmerman.
The TR leverages use-cases to apply the principles and themes discussed to real world situations and technical work. One of the examples concerns semi-autonomous vehicle platooning.
One of the use-cases describes a convoy of trucks travelling along a motorway in close formation. The lead driver grips the wheel as she stares out at the road in front, but the vehicles training behind are self-driving, freeing the other drivers to sleep, read social media posts or pursue other interests.
The authors of ISO/IEC TR 24368 use the road train example to highlight a number of issues, ranging from stress and boredom for the driver to trust over system reliability when driving next to a computer-controlled platoon. They identify a number of ethical and societal considerations related to the use-case, such as accountability, professional responsibility, privacy, safety and security, community involvement and development, respect for the rule of law and labour practices.
The TR provides a general introduction to the broad categories of ethical theory - virtue ethics, utilitarianism and deontology - and international human rights practices. It includes practical examples of practices for building and using ethical and socially acceptable AI.
ISO/IEC TR 24368 features a handy glossary of terms and definitions, as well as a bibliography for further reading. Two annexes provide additional context.
Annex A presents a selection of principles for AI that have emerged out of the international community. Annex B offers practical use-cases linked to the themes and principles identified in the TR.
The ISO/IEC AI committee, known as SC 42, brings together engineers, ethicists, lawyers and others to work on issues such as trustworthiness and biases in algorithms, as well as foundational concepts, data standards, reference architecture, use-cases and application guidelines. Recent SC 42 publications include guidelines on algorithms and the governance of AI in organizations.
(Source: IEC)
