Humanity in the Loop- Balancing Innovation and Ethics in the Age of AI

Detailed Summary

• Welcomed the audience and introduced the panelists, emphasizing UNESCO’s commitment to “human‑centred, ethical AI” that also drives innovation, especially for the Global South.
• Stated the session’s aim: to move from high‑level principles to concrete practices that embed values throughout the AI lifecycle.

2. UNESCO’s Position on Innovation vs. Ethics

• Key claim: Innovation does not have to be at odds with ethics. UNESCO’s 2021 Recommendation on the Ethics of Artificial Intelligence, adopted by 193 member states, provides a universal framework that can be operationalised.
• Three core pillars identified for ethical AI: human rights, human dignity, and fundamental freedoms.
• Emphasised that ethics must be integrated ex‑ante (by design) rather than retro‑active “post‑mortem” fixes.

3. From Principles to Practice – ADG Dr. Tawfik Jelassi

• Highlighted the principle‑practice gap: many countries have signed the UNESCO recommendation but struggle to translate it into on‑the‑ground actions.
• Identified three major gaps:
  1. Lack of operational guidelines for AI developers.
  2. Insufficient monitoring mechanisms for compliance.
  3. Limited capacity within national institutions to assess AI impacts.
• Stressed that UNESCO is supporting readiness‑assessment missions and developing toolkits to help governments create context‑specific governance models.

4. Operationalising Ethics – Debjani Ghosh (NITI Aayog)

• Argued that the real choice is how AI is used – to solve existential problems (e.g., cancer, food security) or to amplify conflict.
• Warned that a single, universal ethical code is unattainable; instead, accountability must stay with humans.
• Proposed a “human‑in‑the‑loop” governance model that embeds oversight at every stage: design, data‑curation, testing, deployment, and post‑deployment monitoring.
• Called for sandbox environments where new AI systems can be evaluated against ethical checklists before market release.

5. Regulation and the EU AI Act – Mr. Brando Benifei (European Parliament)

• Described the risk‑based approach of the EU AI Act:
  • High‑risk AI (e.g., biometric surveillance, predictive policing) is subject to strict conformity assessments, transparency obligations, and in some cases outright bans.
  • Limited‑risk AI can be deployed with lighter documentation and self‑assessment.
• Highlighted four pillars the Act demands: quality training data, robust cybersecurity, clear data‑governance, and human oversight.
• Stressed that trust is a prerequisite for AI uptake, especially in the Global South where skepticism remains high.

6. Education, “AI as a Hammer”, and Collective Intelligence – Prof. Virginia Dignum

• Critiqued the metaphor of AI as a single hammer; innovation should be seen as a toolbox that includes diverse cultural perspectives (e.g., African Ubuntu vs. Cartesian individualism).
• Advocated for curriculum reform that pairs technical skills with humanities and social‑science questions: Why is a problem worth solving? Who benefits? Who may be harmed?
• Emphasised collective intelligence: the true “AGI” is the combined intelligence of people plus AI, not a monolithic super‑intelligence.

7. Private‑Sector Practices – Paula Goldman (Salesforce)

• Explained Salesforce’s Ethical‑by‑Design framework:
  1. Transparency dashboards that show model outputs and confidence scores.
  2. Human‑in‑the‑loop escalation mechanisms (AI → human → AI).
  3. Sandbox testing of new features before release.
• Noted that customers repeatedly ask the same three questions: What results am I getting? How can I detect failures? What is my responsibility vs. the vendor’s?
• Reported that inclusive design – supporting multiple English dialects, accessibility features for disabled users – leads to higher accuracy and commercial success.

8. Multi‑Stakeholder Collaboration & Global Cooperation – Summary from Panel

• AI Impact Commons (launched during the summit) aggregates impact stories from >30 countries, showcasing AI applications that address malnutrition, farmer suicides, flood early‑warning, etc.
• Panelists agreed that global standards (UNESCO recommendation, EU AI Act) must be complemented by regional cooperation for issues that transcend borders (e.g., military AI, existential risks).
• The need for continuous capacity‑building (training policymakers, civil‑society actors, and developers) was reiterated.

9. Audience Q&A

Questioner	Main Query	Respondent(s)	Highlights of Answer
Rajan (Business Club TV)	“What is AI policy?”	Prof. Virginia Dignum	AI policy focuses on impact assessment, governance, and societal goals, not just technology design.
Rita Soni	“How can we involve developers who experience power‑cuts/poor infrastructure in AI design?”	Debjani Ghosh & others	Emphasised democratising design through capacity‑building programs (e.g., Startup India), fostering tier‑2/3 participation, and creating local innovation ecosystems.

10. Closing Remarks

• Moderator thanked participants and highlighted the collective intelligence demonstrated in the panel’s dynamic exchange.
• The session ended with a group photo and a reminder to continue the dialogue through the AI Impact Commons platform.

Key Takeaways

• Ethics and innovation are complementary; embedding ethical reflection from the outset yields more trustworthy, adoptable AI.
• UNESCO’s 2021 AI Ethics Recommendation (adopted by 193 states) provides a global baseline, but operational tools and capacity‑building are needed to bridge the principle‑practice gap.
• The EU AI Act exemplifies a risk‑based regulatory model that can prohibit harmful AI use‑cases while enabling responsible deployment.
• Human‑in‑the‑loop governance must be continuous—design, testing, deployment, and post‑deployment monitoring require explicit oversight checkpoints.
• Inclusive design (language dialects, accessibility, diverse cultural values) produces better performance and market success.
• Education reform is critical: engineers must be trained to ask “why” and “who” alongside “how”.
• Collective intelligence—the synergy of many people plus AI—is the true pathway to achieving AGI‑level problem solving.
• AI Impact Commons shows that AI can deliver concrete social benefits in low‑resource settings; scaling such stories depends on sharing best practices.
• Global cooperation is essential for addressing cross‑border challenges (e.g., military AI, existential risk); multilateral fora like UNESCO can broker common standards.
• Accountability rests with humans; no current AI system can be fully autonomous in ethical decision‑making.

---

Prepared from the verbatim transcript of the “Humanity in the Loop – Balancing Innovation and Ethics in the Age of AI” panel at the AI Conference, Delhi.

See Also:

• shaping-secure-ethical-and-accountable-ai-systems-for-a-shared-future
• ai-for-democracy-reimagining-governance-in-the-age-of-intelligence
• ai-capacity-building-scaling-knowledge-driving-innovation
• enterprise-adoption-of-responsible-ai-challenges-frameworks-and-solutions
• thriving-with-ai-human-potential-skills-and-opportunity
• governing-safe-and-responsible-ai-within-digital-public-infrastructure
• catalyzing-global-investment-for-equitable-and-responsible-ai-in-health
• welfare-for-all-ensuring-equitable-ai-growth-across-the-worlds-largest-and-oldest-democracies
• harnessing-ai-for-health-equity-building-inclusive-human-capital-and-strengthening-researchindustry-collaboration