Scribd
Upload
19 views

Open Science Principles in AI

UNESCO has called for the adoption of open science principles in AI to address issues like the 'black-box' nature of AI and its dominance by private companies. The organization emphasizes the importance of ethical AI governance and has introduced a global standard for AI ethics, highlighting transparency, human rights, and inclusivity. UNESCO also provides tools and platforms to support ethical AI practices and gender equality in AI design and deployment.

Uploaded by

devojit_doley
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
19 views

Open Science Principles in AI

UNESCO has called for the adoption of open science principles in AI to address issues like the 'black-box' nature of AI and its dominance by private companies. The organization emphasizes the importance of ethical AI governance and has introduced a global standard for AI ethics, highlighting transparency, human rights, and inclusivity. UNESCO also provides tools and platforms to support ethical AI practices and gender equality in AI design and deployment.

Uploaded by

devojit_doley
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

UNESCO Calls for Open Science Principles in AI

insightsonindia.com/2024/09/17/unesco-calls-for-open-science-principles-in-ai

September 17, 2024

Syllabus: Applications of Ethics

Source: Unesco

Context: UNESCO urged the application of open science principles to AI, noting that AI’s
“black-box” nature and dominance by private companies limit open science.

What is Open Science?

Open Science is an inclusive approach that aims to make scientific research and data
openly available, accessible, and reusable for everyone. It encompasses sharing
research outputs, methodologies, software, and data to promote transparency, collaboration,
and inclusivity in scientific knowledge.

1/5
Benefits of Open Science in AI:

1. Reproducibility: Enables validation and replication of AI experiments.


2. Innovation: Accelerates new developments through shared research.
3. Transparency: Makes AI systems more understandable by reducing “black-box”
issues.
4. Bias Mitigation: Helps identify and address biases in AI models.
5. Inclusivity: Allows researchers from diverse backgrounds to contribute.
6. Data Quality: Promotes the creation of standardized, high-quality datasets.
7. Ethical Development: Encourages responsible and ethical AI practices.
8. Cost Efficiency: Reduces duplication of research efforts.

Challenges AI Poses to Open Science:

Reproducibility Crisis: Difficulty in replicating AI-based experiments.


Interdisciplinarity: Limited collaboration between AI and other fields.
Data Issues: Challenges in data quality and potential biases.
Changing Incentives: Pressure on researchers to prioritize AI over other scientific
rigour.

UNESCO’s Call for Ethical AI Governance

2/5
UNESCO also emphasizes the need for ethical AI governance and the use of the “Global AI
Ethics and Governance Observatory.” This observatory supports policymakers,
academics, and the private sector in tackling AI’s challenges, promoting ethical and
responsible AI adoption globally.

UNESCO’s Ethics of AI Recommendation:

In 2021, UNESCO introduced the first global standard on AI ethics, emphasizing


transparency, human rights, and fairness. It sets forth four core values for AI:

1. Human Rights and Dignity: Respect and protection of fundamental rights.


2. Peaceful and Inclusive Societies: Ensuring AI fosters justice and interconnectivity.
3. Diversity and Inclusiveness: Promoting varied perspectives and inclusivity in AI.
4. Environmental Sustainability: AI should contribute to environmental well-being.

Ten Core Principles:


The recommendation outlines ten principles, including proportionality, safety, privacy,
transparency, human oversight, sustainability, and fairness. It also suggests actionable
policies in eleven key areas for ethical AI development.

Implementation Tools:
UNESCO developed the Readiness Assessment Methodology (RAM) and Ethical Impact
Assessment (EIA) to help Member States implement these principles effectively.

Women4Ethical AI Platform:

UNESCO’s Women4Ethical AI platform supports gender equality in AI design and


deployment, uniting female experts to advance ethical AI practices and non-discriminatory
algorithms.

AI’s Potential for Ethical and Moral Behavior:

Aspect AI’s Potential for Ethical and Moral Behavior

Views

3/5
Understanding For e.g., AI systems can be trained to identify hate speech and offensive
Ethics and content to maintain a respectful online environment.
Morality

Bias Mitigation AI can be programmed to mitigate biases and avoid unfair


discrimination.

Decision- AI can make ethical decisions based on predefined rules and data.
Making (but lacks true moral understanding)

Counterview

Learning from AI learns from data, which might include biased or unethical information,
Data leading to unintended consequences.

Ethics in AI: Applying Kantian ethics to AI decision-making within governance


Kantian raises concerns. Delegating decisions to algorithms could undermine
Perspective human moral reasoning and responsibility. Isaac Asimov’s ‘Three
Laws of Robotics’ also highlights the challenges in translating ethics
into AI rules.

Programming Programming ethical AI is more challenging than programming AI for


Ethics into AI: tasks like chess due to the intricate nature of ethical considerations.
A Complex
Task

Autonomy and AI lacks consciousness and intent, making its actions neither
Intent inherently moral nor immoral. E.g., A robot that assists the elderly with
daily tasks completes them efficiently but without genuine care or
compassion.

Accountability As AI assumes decision-making roles, accountability questions arise.


and Liability If AI-based decisions turn out to be unethical, who bears responsibility?
Punishing AI is problematic as it lacks emotions. Deciding who is
accountable—AI developer, AI user, or AI itself—poses a significant
challenge.

Unintended E.g., Social media algorithms, while aiming to show relevant content,
Consequences might inadvertently create echo chambers and reinforce biases.

Continuous AI’s ability to learn and adapt can lead to ethical shifts over time,
Learning requiring ongoing evaluation.

Human The ethical behaviour of AI often requires human oversight and


Oversight intervention. E.g., Content moderation platforms use AI to flag
potentially inappropriate content, but human moderators make final
decisions.

Steps Taken for Ethical AI:

Steps Description

4/5
International Global Alliance for Social Entrepreneurship: Launched AI for Social
Innovation initiative at WEF 2024 with Microsoft to promote positive AI
impact and responsible guidelines. Examples: China, Canada, and
Singapore have AI regulations

EU AI Act: Comprehensive regulation for AI risk governance and citizen


protection

California: Bill for AI safety testing to prevent misuse.

UK AI Safety Summit: The 2023 summit focused on AI safety and


international cooperation.

Tech Giants: Microsoft, Meta, Google, Amazon, and Twitter have


responsible AI teams for ethical oversight

National Advisory on AI Models: MeiTY issued guidance on AI models and


deepfakes in 2024.

IndiaAI Mission: Promotes AI innovation through public-private


partnerships, improving data quality and ethical AI.

Responsible AI for Youth: National program launched for youth.

National Strategy on AI: NITI Aayog’s 2018 strategy for safe, inclusive AI
adoption across sectors with the “AI for All” mantra

For Generative AI: What are the potential applications


and ethical concerns? Click Here

Insta Links:

A new global standard for AI ethics

5/5

You might also like