Introduction
As Artificial Intelligence (AI) rapidly evolves and integrates into various aspects of daily life, ethical considerations related to AI have become increasingly important. Addressing ethical considerations like fairness, transparency, and accountability helps promote the responsible use, development and governance of AI technologies.
The 41st session of The General Conference of the United Nations Educational, Scientific and Cultural Organisation (UNESCO) in Paris in November 2021 addressed the pressing ethical issues surrounding AI. The Recommendations on the Ethics of Artificial Intelligence (The Recommendations), a comprehensive set of recommendations on the moral problems related to AI in education, culture, communication and information, was discussed and officially adopted on 23rd November 2021.
Scope
The Recommendation addresses ethical issues related to AI to the extent that it falls within the mandate of UNESCO. This means it addresses the broader ethical implications of AI systems about education, culture, communication and information.
This Recommendation applies to member states, both as AI actors and as authorities responsible for developing regulations and overseeing AI governance. Furthermore, it also provides ethical guidance to all other AI-related actors by providing a basis for analyzing or examining the ethical Impact of AI systems.
Principles
The recommendations aim to respect, protect, and promote human rights, fundamental freedoms, and human dignity while ensuring flourishing environments and ecosystems. They also emphasize ensuring diversity, inclusiveness, and interconnected societies. The ethical principles delved into by the Recommendations address these values.
- Proportionality and not harm: AI processes must be proportionate and context-appropriate, with risk assessments and preventive measures to protect humans, society, and the environment. AI should not be used for social scoring or mass surveillance.
- Fairness and non-discrimination: AI actors must promote fairness, social justice and non-discrimination, ensuring inclusive access, tackling digital divides and minimizing bias. Advanced nations should share AI benefits globally.
- Safety and security: Unwanted harms from safety and security risks should be eliminated with sustainable and privacy-protection-centric frameworks ensuring quality AI training and validation data.
- Sustainability: Continuous assessment of the impact of AI on sustainability, as reflected in the UN’s Sustainable Development Goals, must be conducted and tailored to the development needs of each country.
- Right to privacy and data protection: It is essential for data processing practices to comply with international law and UNESCO recommendations, with strong data protection frameworks and accountability for safeguarding personal information.
- Human oversight: It must be ensured that ethical and legal responsibility AI systems have human accountability and oversight, with ultimate decision-making responsibility remaining with humans.
- Transparency and explainability: Transparency and Explainability are crucial for ensuring the protection of human rights, freedoms, ethical principles and public trust. Transparency and meaningful explanations must be provided for high-risk AI decisions.
- Responsibility and accountability: AI actors and states must respect human rights and the environment, with clear responsibility and accountability, including oversight, audits and impact assessments.
- Awareness and literacy: Public understanding of AI must be promoted through accessible education, digital skills training and civic engagements led by various stakeholders.
- Multi-stakeholder collaboration and adaptive Governance: It is essential to engage diverse stakeholders in AI governance, adapt to technological changes, and respect data self-governance.
Policy Areas
The following policy action areas discussed in the Recommendations operationalize the principles laid out above:
Ethical Impact Assessment
Frameworks for impact assessments should be introduced by member states to identify and assess any risks, benefits and concerns related to AI systems, along with an analysis of appropriate risk prevention, mitigation and monitoring measures. These impact assessments must address human rights, freedoms and environmental concerns while promoting citizen participation and transparency. Governments and businesses must implement due diligence and oversight mechanisms to monitor AI’s socio-economic effects. Furthermore, it is recommended that regulatory frameworks mandate ethical impact assessments.
Ethical Governance and Stewardship
The Recommendations suggest that Member States should ensure inclusive, transparent, and multilateral AI governance with mechanisms for effective protection, monitoring, and redress of harms, including auditability and traceability of AI systems. It is essential to ensure that AI strategies balance innovation with the protection of rights, ethical oversight and multi-stakeholder involvement while considering the impact on marginalized groups. It is crucial for key ethical considerations to be promoted by all AI regulations, with ultimate responsibility placed on natural or legal persons rather than the AI systems themselves.
Data Policy
Member states should develop data governance strategies that enable continuous evaluation of the quality of AI training data, promote protection of privacy and encourage privacy impact assessments. These frameworks or strategies must ensure that individuals retain control over their personal data through transparent frameworks with safeguards for sensitive data while also balancing open data policies and secure sharing of information.
Development and International Cooperation
AI ethics should be integrated into discussions at relevant international or intergovernmental forums. International cooperation on AI for development must be promoted through platforms that provide funding, expertise and collaboration to tackle development challenges related to Low-Middle Income Countries. International AI research and innovation, as well as research in various cultural contexts and technological exchanges to bridge geo-technological gaps while respecting international law, must be promoted.
Environment and Ecosystems
Environmental impacts of AI systems, including carbon footprint, energy consumption, and raw material extraction, must be efficiently evaluated and mitigated. In doing so, compliance with environmental laws is crucial. It is also recommended that incentives be introduced to promote AI solutions for disaster resilience, environmental protection, and sustainability. AI systems should involve local and indigenous community participation and support sustainable practices. When selecting AI methods, preference should be given to data, energy, and resource-efficient options. The Recommendation calls for the application of the precautionary principle and emphasizes that an AI that may disproportionately harm the environment must not be used.
Gender
AI’s potential to contribute towards the promotion of gender equality must be maximized while ensuring the protection of the rights, safety, and integrity of women. For this purpose, Ethical Impact Assessments should incorporate a gender perspective. Member States are further urged to allocate public funds towards gender-responsive initiatives. AI technologies must help reduce gender gaps and prevent gender stereotyping and bias in AI systems while also addressing issues like harassment, bullying, or trafficking. Furthermore, economic and regulatory initiatives must encourage female entrepreneurship and participation in AI research and management.
Culture
Member states are encouraged to incorporate AI in preserving, promoting and managing cultural heritage. The cultural impact of AI must be assessed, especially in natural language processing (NLP) applications, to prevent the disappearance of endangered languages and address the cultural nuances. The promotion of AI education and training for creative professionals is further recommended. Awareness of AI tools must be spread to local cultural industries and SMEs. Furthermore, research at the intersection between Intellectual Property and AI must be promoted.
Education and Research
Collaboration between various governments, international organizations, and educational institutions is encouraged to provide comprehensive AI literacy. Research on responsible and ethical use of AI must be supported. Furthermore, regular monitoring, impact and quality assessment, diverse and inclusive participation in the use of AI in education and widely accessible use of data per data protection standards are encouraged in the development and use of AI systems.
Communication and Information
The use of AI to enhance access to information is encouraged. It must be ensured that AI systems respect freedom of expression and access to information, including transparent online communication, diverse viewpoints, and processes for content removal. The recommendations urge for increased investment in digital and media literacy to strengthen critical thinking, help individuals understand the implications of using AI systems, and combat misinformation. Additionally, states must also endeavour to create conditions for media to effectively report on AI’s benefits and harms and encourage the ethical use of AI in journalism.
Economy and Labour
AI’s impact on labour markets and the economy must be thoroughly assessed, and collaborative efforts between various stakeholders, including the government, educational institutions and industry, must be promoted to align training with future job requirements. Encouraging interdisciplinary research on AI’s labour impact, promoting competitive markets while preventing monopolies, and respecting international standards are also crucial.
Health and Social Well-being
The use of AI to enhance healthcare services while ensuring compliance with international standards is highly encouraged. To ensure safety and efficient practices, robust regulations for AI in healthcare are required. Due to the involvement of sensitive and industry-specific data, it is essential to minimize bias, include domain experts, ensure privacy, obtain informed consent, and maintain human oversight. The recommendations also address the use of AI in relation to mental health and regulation of the relationship between patients and AI while also emphasizing various areas of research and awareness requirements at the intersection of AI and healthcare.
Implementation, Monitoring and Evaluations
States should transparently monitor and evaluate AI ethics policies using both quantitative and qualitative methods. Furthermore, UNESCO aims to assist by developing methodologies for ethical impact assessment, readiness assessment, and policy evaluation while also strengthening research and dissemination of best practices. Monitoring should include broad stakeholder participation and ensure compliance with international data protection standards.
Conclusion
UNESCO’s Recommendation on AI Ethics provides a comprehensive guiding framework for the ethical development and governance of AI. It sets guidelines for member states and other AI actors to foster an environment of responsible AI governance, emphasizing on fairness, human rights, freedoms and inclusivity. Implementing these recommendations requires collaborative efforts from governments, international organizations, AI actors and the general public to ensure that AI technology benefits all.
