Download to read offline
More Related Content
Similar to Conferencia de Abertura_Virgilio Almeida.pdf
Conferencia de Abertura_Virgilio Almeida.pdf
- 1. Inteligência Artificial, Éticae Universidade: Por que precisamos de regras? Virgílio Almeida Departamento de Ciência da Computaão Universidade Federal de Minas Gerais Berkman Klein Center, Harvard University Belo Horizonte, 10 de novembro de 2025
- 2. Edward Wilson –Harvard professor, American biologist, naturalist, and ecologist • We will have to evolve a better world order than the one we have now, which I like to call our Star Wars Civilization. I mean we have stone-age emotions, medieval institutions and godlike technology. • In the case of our institutions, especially within religions and ideology, we are in constant conflict. And in the case of our technology, we are seeing things going almost beyond the control of our imagination (2011). Virgilio Almeida, UFMG, 2025
- 3. Digital Future inthe Age of Uncertainties • Technological advances bring enormous possibilities but also deepen inequalities. • Technological advances bring enormous possibilities but pose profound risks to democracy. • Technological advances bring enormous possibilities but demand ethical and political responsibility. • Technological advances bring enormous possibilities but may outpace our capacity to govern them. 3 Virgilio Almeida
- 4.
- 5. Social and EthicalImpacts of AI Virgilio Almeida 5
- 6.
- 7. • AI andthe Future of Work Virgilio Almeida 7
- 8. Building Brazil’s Desired DigitalFuture Science, technology, publicy polic, governance and education
- 9. O Brazil hasbeen making progress in AI through the PBIA, PL 2338, Lei EDCA… but we need to move further and faster… Virgilio Almeida 2025 9
- 10. Virgílio Almeida, 2025 VirgilioAlmeida 10
- 11. Virgílio Almeida, 2025 VirgilioAlmeida 11
- 12. We will livein a hybrid society where humans must interact with artificial agents, comply with automated decisions, and develop new forms of sociality. A vision of the future As AI systems are deployed even more widely, the lives of people and adaptive algorithms will become increasingly intertwined. 12 Virgilio Almeida
- 13. Emerging forms of socialinteraction Virgilio Almeida 13
- 14. Technological advances often happenrapidly, technology adoption occurs gradually, and labor market transformation takes years. Productivity growth driven by General Purpose Technologies leads to the creation of new occupational categories and makes other professions more productive. Virgilio Almeida 14
- 15. A framework for responsibleAI to foster sustainable and inclusive growth. Virgilio Almeida 15
- 16. 10 Priorities forAI Research & Policy • 3. Regulation - Devise best practices, audits, and laws and regulations to incentivize reporting on, modifying, and addressing new capabilities and emergent phenomena to ensure that fielded models and systems are safe and responsibly deployed. • These efforts can lead to appropriate oversight, feedback, and governance mechanisms aimed at aligning AI systems with human values and interests. 16 Virgilio Almeida
- 17. Digital Governance … Spaceopen to new ideas….
- 18. Three fundamental quesCons: •Why is AI regulation and governance a hard problem? • Why is AI regulation and governance an even harder problem in the Global South? • What are promising areas for research? 18 Virgilio Almeida
- 19. Widely Distributed Harms Partof the governance challenge is that at the micro level these ills are subtle and widely dispersed; an algorithmically amplified piece of political content is not acutely harmful to an individual user, for instance. It is only at the macro societal level that the scope and consequences of these local interactions come into view. . 19 Virgilio Almeida
- 20. AI Governance challenges 1.Pace Gap ()mescale mismatch): AI evolves fast, policy lags behind 2. Uncertainty: AI’s impacts remain unpredictable 3. Monopoly Power: AI dominated by few big firms 4. Rising Inequality: AI deepens global digital divides 5. Widespread Harms: While AI offers benefits, it may worsen issues like mental health crises, societal trust erosion, and poli)cal polariza)on. 20 Virgilio Almeida LaForge, Gordon, et al. Complexity and the Global Governance of AI. Washington, DC: New America, 2023.
- 21. AI Governance inthe Global South additional challenges 1. Limited digital infrastructure and tech readiness 2. Weak ins)tu)ons and regulatory capacity 3. Tensions between foreign investment and local policy goals 4. Global imbalances in tech and governance power 5. Shortage of skilled STEM professionals 6. AI governance challenged by deep social and economic inequali)es Virgilio Almeida 21
- 22. Virgilio Almeida 22 Collingridge Dilemmafor AI regulation The Pacing problem • The Collingridge dilemma describes the difficulty of controlling or influencing a technology’s development. • Collingridge’s key insight is that regulation is easier early on, when a technology is new and its unintended consequences are not yet clear. • Waiting until consequences appear risks losing control over how the technology is regulated.
- 23. The Multidisciplinary Foundationsof AI Governance • AI governance à technical, legal, ethical, and societal dimensions: 1. Computer Science – System design, algorithmic transparency, safety, and security 2. Law – RegulaLon, compliance, and legal accountability 3. Public Policy – Governance frameworks and insLtuLonal decision-making 4. Ethics & Philosophy – NormaLve frameworks, fairness, and moral responsibility 5. Psychology & CogniLve Science – Human-AI interacLon and behavioral impacts 6. PoliLcal Science & HumaniLes – Power, equity, linguisLcs and societal implicaLons 7. Economics – IncenLves, market impacts, and resource allocaLon. 23 Virgilio Almeida
- 24.
- 25. Algorithmic Societies § Wedepend on algorithms for a growing number of activities – decision-making ü Urban Mobility ü Relationships ü Definition on what you see in social platforms ü Access to social policies and benefits ü Medical Decision ü Policing and Criminal Dosimetry 25 Virgilio Almeida
- 26. Algorithmic systems have alreadyshaped economies, politics, geopolitical power, culture, and communities 26 q Algorithmic systems are made of mul@ple algorithms, models, data, applica@ons, hardware, goals, people, and networks. q● Algorithmic systems are sociotechnical systems, where technology and social actors are deeply intertwined. q Examples of global algorithmic services: • ● Uber Dynamic Pricing System • ● Waze Route Op@miza@on • ● Recommenda@on Algorithms – Instagram, TikTok For You Page, Spo@fy… • ● Social Welfare Eligibility Systems Virgilio Almeida
- 27. New ideas forseeking an understanding of the profound social and political transformation brought about by the algorithmizing of society. Thinking of algorithms as institutions Virgilio Almeida 27
- 28. Thinking of Algorithmsas Institutions Virgilio Almeida 28
- 29.
- 30. Política dos algoritmos UBUEditora Ricardo Mendonça Fernando Filgueiras Virgilio Almeida Com a difusão da inteligência artificial, sistemas algorítmicos passam a atuar como instituições que moldam normas e rotinas sociais e políticas Outubro 2025 30 Virgilio Almeida
- 31. What are institutions? •Our first thoughts, at least among non political scientists, would be of large government buildings with brass plaques that proclaim that they are the parliament, or a ministry. • A second thought might be of social institutions such as marriage, or the Church, or even law. • We might even think of even more amorphous institutions based on norms or discourses—even a social club or a sports team—that exert influence over individual behaviour because the individuals have learned and accepted the values. • But can a few lines of computer code be an institution? 31 Virgilio Almeida
- 32. • Like socialand political institutions, algorithms are devices that structure our decision-making and problem-solving capacity. As any other device, algorithms must be evaluated in terms of their social and political foundations, as well as in terms of their legitimacy and functionality. • What should be the forms of democratic governance of algorithms? • How to think about public forms of regulation and management not only of algorithmic production but also of its consequences? • How to democratize these institutions that govern a large part of our collective experience. Algorithmic Institutionalism 32 Virgilio Almeida
- 33. What if algorithmic recommenderswere viewed as institutions? 33 Virgilio Almeida
- 34. A nonideological example: Asocial network I use learned I like videos of big cats acting like house cats. At first, the videos were real, but later AI-generated fake videos appeared, many gaining thousands or even millions of likes. For the platform, it didn’t matter if the videos were real—what mattered was keeping me engaged, so I’d see more ads or share more content, capturing others’ attention too. Business model of social media platforms Virgilio Almeida 34
- 35. Viewing recommendation algorithmsas institution • Algorithmic recommenders are key in the aRen@on economy, widely used in industry, markets, and governments, classifying informa@on, crea@ng biases, and changing how we communicate publicly. • Recommenda@on algorithms, like ins@tu@ons, use rules to shape what users believe and value. • Da@ng algorithms shape social and cultural specifici@es through informal rules that establish what is allowed or prohibited in a social rela@onship. e.g.: on Tinder, everyone can make the first move, while, on Bumble, only women can ini@ate communica@on. • The mix of desire, convenience, and efficiency makes algorithmic recommenders powerful in shaping what gets no@ced. They act as ins@tu@ons, seVng limits on individual choices. 35 Virgilio Almeida
- 36. Governing the digitalspace: new ins@tu@ons • Platforms as Institutions • Platforms function like institutions, organizing contexts where people, businesses, and governments interact. • As intermediaries, they reshape how we work, communicate, consume, and relate—driving the re-institutionalization of existing structures. • Streaming Platforms and the Music Industry • Apple, Spotify, and YouTube have redefined the music industry—transforming artist revenue models, shaping listening habits through algorithms, and altering relationships among artists, labels, and listeners by acting as institutions. • Media Platforms and the Public Sphere • Algorithmic platforms, like TikTok, Instagram, Facebook, and X, function as institutions, reshaping how media structures the public sphere. Virgilio Almeida 36
- 37. Why Multidisciplinarity Matters forthe Advancement of AI • Understanding compu@ng concepts in real-world contexts requires; • Ethical impact of AI; • Social impacts of AI; • Poli@cal impact of AI; • Economic impact of Ai; • Knowledge disciplines: the founda@on for informed engagement with AI; • The role of the humani@es as specialists in the crea@on of meaning. • New heuris@cs inspired by theore@cal aspects of social sciences. • Math and physics: Interpre@ng complex systems; 37 Virgilio Almeida
- 38. AI R&D inthe Global South should support the public interest. Challenges: low educational level, huge inequalities, language barriers, poor public services, impact of automation (potential for AI to be labor-saving), unskilled labor , … Virgilio Almeida, 2023
- 39. Digital transformation inthe Global South: A Call for Research • Studying collective human behavior across disciplines is difficult but essential, especially as technology increasingly shapes how societies function. • Algorithmic systems and tech companies already influence global behavior for profit, making inaction risky. • Large language models (LLMs) act as powerful drivers of information flow, shaping behavior on a global scale. 39 Virgilio Almeida
- 40. The Role ofTechnology in Addressing Problems of the Global South Virgilio Almeida 2025 40 Logical appeal, statistical evidence, and authority reduced user engagement compared to other strategies.
- 41. How do illiterate peopleinteract with Intelligent Voice Assistants? Helen T., Furtado F., Almeida V. et al, ``How do illiterate people interact with an Intelligent Voice Assistant?, International Journal of Human–Computer Interaction, 2024. 584-602 41 AI for the underrepresented Virgilio Almeida, 2024
- 42. AI Advances: WhatLies Ahead? • Research priorities • Foster multidisciplinary research linking CS, social, and political sciences. • Examine AI’s impact in the Global South on work, security, environment, and inequality. • Expand equitable data access for Global South researchers. • Deepen understanding of algorithmic institutions and their role in governing social life in the Global South. • Contextual challenges • Digital platforms are challenging democracy. • The global race for technologies intensifies power struggles. • AI is weakening the sovereignty of countries (e.g., through social networks). 42 Virgilio Almeida
- 43. Responsible AI Governance 43 Regulation Principles Rulesfor Research ins3tu3ons Virgilio Almeida, UFMG - 2024 Researchers need clear rules to guide the development and use of AI technologies
- 44. Closing thoughts • Promotemultidisciplinary research that bridges computer science with the social, political, and natural sciences to address the complex challenges of AI in society. • Promote AI and digital technologies in the Global South, with focus on work, security, environment, and inequality. • Expand equitable access to research data, particularly for Global South scholars who face barriers to platform datasets. • Study algorithmic institutions and how they shape governance and social life in the Global South. • Assess the global impact of AI technologies (e.g., LLMs), focusing on challenges that disproportionately affect the Global South, such as environmental degradation and rising inequality. 44 Virgilio Almeida 2025
- 45. Virgílio Almeida, 2025 Obrigado! virgilio@dcc.ufmg.br X:@virgilioalmeida 45 Virgilio Almeida