1 of 13

Navigating the Future: AI Ethics & Responsible AI

Presentation by

Shruti Kakade

Hertie school of Governance

2 of 13

Why AI Ethics?

Can we trust AI systems with our lives?
Imagine a world where AI dictates our choices—who gets a job, who gets a loan, who has access to healthcare

3 of 13

What’s at Stake?��

Fairness, Human Rights, Accountability

Example: Bias in hiring algorithms.

4 of 13

AMAZON’S AI HIRING TOOL, DEVELOPED TO AUTOMATE RESUME SCREENING, WAS TRAINED ON 10 YEARS OF BIASED DATA THAT FAVORED MALE CANDIDATES, REFLECTING THE GENDER IMBALANCE IN TECH. THE AI DOWNGRADED RESUMES WITH WORDS LIKE “WOMEN’S” AND PENALIZED GRADUATES OF WOMEN’S COLLEGES. THIS BIAS, LEARNED FROM HISTORICAL DATA, LED AMAZON TO SCRAP THE TOOL IN 2018. IT SERVES AS A CAUTIONARY TALE ABOUT THE IMPORTANCE OF FAIR DATA, HUMAN OVERSIGHT, AND TRANSPARENCY IN AI SYSTEMS.�

5 of 13

Key Ethical Questions

Who is accountable if AI fails?
How do we ensure fairness?
What happens when AI makes a biased decision?

6 of 13

What is Responsible AI?

Aligning technology with human values and ethics

7 of 13

Principals of Ethics

1. Fairness: Robots can’t be jerks!
2. Transparency: No shady stuff, explain your decisions.
3. Accountability: Own up to mistakes.
4. Privacy: Respect our midnight snacks!
5. Safety: No robot uprisings.
6. Inclusivity: AI for everyone, not just geeks.

8 of 13

Building Inclusive AI

Importance of inclusivity in AI development to benefit everyone.

Initiatives for diverse teams in AI development.

9 of 13

Interactive Activity: Ethical Dilemmas

Your AI system recommends different treatments based on race. What do you do?

Investigate the Source of Bias

Ensure Fairness

Explainability

Redesign or Retrain the Model

10 of 13

Technical tools of AI governance

Bias Detection (Aequitas, Fairness Indicators): These tools evaluate AI systems for biases across demographic groups, ensuring fairness in decision-making processes.

Explainability Tools (LIME, SHAP): These provide transparency by explaining how AI models make decisions, making it easier to understand and address potential issues.

Model Auditing (Fairlearn, AI Fairness 360): These tools audit AI models to ensure they comply with fairness and accountability standards, helping mitigate bias before deployment.

Privacy-Preserving Techniques (Differential Privacy, Federated Learning): These methods protect user privacy while allowing AI to learn from data without accessing sensitive information directly.

11 of 13

What governments are doing

Overview of major AI regulations

EU AI Act: Classifies AI systems based on risk (e.g., minimal risk to unacceptable risk)

GDPR: Enforces data protection and privacy regulations in AI applications

OECD AI Principles: Global cooperation on promoting responsible AI

UNESCO AI Ethics: Encourages development aligned with human rights

Citation: https://www.software.com/src/ai-ethics-diverging-global-strategies-open-gaping-regulatory-void

12 of 13

The Future of AI Ethics

Global Impact of AI: AI systems developed in one country often affect users worldwide. This makes international collaboration essential to ensure consistent ethical standards.

Challenges of Fragmented Regulation: Without global consistency, different regions may have conflicting AI regulations, leading to ethical "blind spots" and unequal protection for people across borders.

International Frameworks: Efforts like the OECD AI Principles and UNESCO’s AI Ethics Guidelines are paving the way for unified ethical standards that promote fairness, transparency, and accountability.

Unified Ethical Standards: The future of AI ethics lies in creating a global framework that addresses key challenges like bias, privacy, and transparency, ensuring AI benefits everyone, not just specific regions.

13 of 13

Conclusion: Shaping the Future of AI Ethics

The future of AI is in our hands, and it's our responsibility to ensure it is ethical.
Join the mission to build ethical AI.