In a historic global agreement on 25 November 2021, all 193 members of the United Nations Educational, Science and Cultural Organisation adopted the “Recommendation on the ethics of artificial intelligence”. However, the recommendations are not a binding agreement, rather guidelines that nations can follow voluntarily.
Overview of the Document
The document acknowledges the “profound and dynamic” impact of AI on human lives and contemporary societies and how it affects education, social and natural sciences, culture, etc.
Moreover, UNESCO recognises the potential that AI possesses in terms of benefitting ecosystems and the environment.
Sign up for your weekly dose of what's up in emerging technology.
On the other hand, UNESCO, in its document, highlights the dangers of AI. It raises ethical concerns surrounding the gender and ethnic biases that are “embedded” in algorithms which may “exacerbate” inequality, exclusion and threats to cultural, social and ecological diversity. The document further raises AI red flags by stating that AI technology may deepen a divide between countries.
In this regard, UNESCO Chief Audrey Azoulay says, “The world needs rules for artificial intelligence to benefit humanity. The Recommendation on the ethics of AI is a major answer. It sets the first global normative framework while giving States the responsibility to apply it at their level.”
Ethics of AI: Values and Principles
The document defines the scope and provides a universal framework of values and principles of the ethics of AI within its scope. The definition of AI within the document includes and is not limited to machine learning, machine reasoning, cyber-physical systems, IoT, and all stages of AI systems life cycle (process, person, enterprise).
The objective of the document is to guide the actions of the stakeholders in embedding the ethics of the AI, promote human rights and fundamental freedom in all stages of the AI life cycle, and promote equitable access to development and knowledge in the field of AI between the global north and south nations. The recommendations explicitly ban the use of AI social scoring and mass surveillance and advocate stronger data protection.
The text principally calls for:
- Fairness and Non-Discrimination- Fairness implies sharing the benefits of AI technologies at local, national and international levels while considering the specific needs of different marginalised and vulnerable populations.
- Sustainability- AI technologies must be continually assessed for their social, cultural, economic and environmental impact.
- Privacy– Human autonomy and agency must be respected, and data must be collected, shared, archived and deleted ethically. The document also calls for data protection through frameworks and government mechanisms set up by regulatory bodies. In addition, algorithmic systems must undergo privacy impact assessment, which includes societal and ethical considerations.
- Safety and Security– Unwanted harm and vulnerability to attack should be avoided by developing privacy-protective data access frameworks that foster better training AI models.
- Transparency and Explainability- Transparency ensures respect for human rights principles. Explainability refers to making intelligible and insightful outcomes to the AI system.
- Responsibility and Accountability- Similar to transparency and explainability, the principle ensures human rights and freedom.
- Awareness and Literacy- Public awareness and understanding of AI technologies and the value of data must be promoted through open and accessible education to all.
Interestingly, nations like China and Russia who have a comprehensive mass surveillance system, have voted to adopt AI Ethics Recommendations. The USA, a country that houses the most number of AI companies, is not a part of UNESCO, thus, not a signatory to the document. However, in an interview with Politico, UNESCO’s Assistant Director-General for Social and Human Sciences, Gabriela Ramos, argues that the nation’s policies can be pressured by global unity on the subject.