Schedule - Fellowship Program
Session 1: SATURDAY, 06.05.2023. - AI FOR THE COMMON GOOD
2:00 - 4:00 CEST
2:00 WELCOME!
Opening by the Goethe-Institut and the host Mpho Raborife from the University of Johannesburg
2:10 INPUT & DISCUSSION
Introduction to AI and ethics / Big Data and Human Rights:
Talk by Alex Tsado/ Alliance4AI (Nigeria/USA): Steps to building ethical and practical AI tools
As artificial intelligence (AI) becomes increasingly prevalent in our daily lives, it's important to ensure that the AI tools we build are not only effective and efficient but also ethical and practical. The development of ethical and practical AI tools requires a structured approach that considers the potential social impact of AI systems and mitigates any potential harm.
Talk by Laura Alonso Alemany/ Via Libre Foundation (Argentina): Now that you are concerned about AI, what can you do about it?
Serious concerns about AI have been known lately, as more powerful systems are applied to more aspects of our lives, possibly affecting fundamental rights. In this talk we will briefly show some illustrative cases where AI systems have harmed people, by errors that have led to unjust criminalization, privation of freedom, or discrimination. Then, we will show how very frequently these problems go unchallenged, mainly because most of the people feel that they are not prepared to challenge such harmful behaviors from automatic procedures.
The central part of the talk will be devoted to showing how the behavior of AI applications can be inspected. We will also show how a more nuanced inspection of the innards of AI systems can be carried out with tools like inspection of biases in language models. We will demonstrate how these inspections are more valuable when experts in discrimination can translate their expertise to hypotheses about biases and have them checked within AI systems. We will show EDIA, a tool to carry out audits of AI systems without technical barriers, and how technical people can facilitate the implementation of these tools in a multidisciplinary team.
Finally, we will conclude showing how AI systems are not objective, but they are shaped by the values of the people who design them, and how the concentration of these systems in a handful of huge corporations amplifies the impact of those values.
DISCUSSION AND EXCHANGE IN GROUPS
Session 2: SATURDAY, 13.05.2023 - GLOBAL NORTH AND GLOBAL SOUTH
1:00 - 3:00 CEST
1:00 - 1:10 WELCOME
1:10 - 2:00 INPUT & DISCUSSION
Talk by Paola Ricaurte Quijano (Tecnologico de Monterrey/ Mexico; Berkman Klein Centre, Harvard, USA/ Tierra Comun, Interventions for Data Decolonialization): Ethics for the majority world: AI and the question of violence at scale
Hegemonic AI is becoming a more powerful force capable of perpetrating global violence through three epistemic processes: datafication, algorithmization, and automation. These articulated epistemic mechanisms lead to global classification orders and epistemic, economic, social, cultural and environmental inequality.
Talk by Pelonomi Moiloa (Lelapa AI): Protecting Machines from Us - Lessons from the majority world
There is a lot of talk about AI taking over the world. Pelonomi Moiloa will talk about this concept from a global majority perspective to highlight how we might think differently about addressing this topic.
DISCUSSION AND EXCHANGE IN GROUPS
Session 3: FRIDAY, 19.05.2023 - INCLUSION
1:00 - 3:00 CEST
INPUT & DISCUSSION
Talk by Nathan-Ross Adams (South Africa): Building Inclusive AI for Sustainable Development in Africa
In this talk, we’ll explore the critical importance of inclusive AI in achieving sustainable development goals from an African perspective through:
- Examining the various ways in which AI can perpetuate systemic bias and discrimination.
- Discussing how to build more inclusive and diverse AI systems to address these challenges.
- Delving into the challenges of designing and developing AI systems that account for different African languages, dialects, and cultural nuances.
- Considering the ethical implications of AI communication and language in Africa and the need to ensure that AI systems do not cause harm to marginalised communities.
We can create a more just and equitable future for Africa by aligning with the United Nations’ Sustainable Development Goals.
Outcomes:
- Understand the role of AI in achieving sustainable development goals in Africa.
- Identify the challenges of building inclusive and diverse AI systems that address systemic bias and discrimination in Africa.
- Learn the importance of accounting for different African languages, dialects, and cultural nuances in AI.
- Discuss the ethical implications of AI communication and language in Africa and the need to ensure that AI systems do not cause harm to marginalised communities.
- Develop strategies for creating more inclusive and diverse AI systems in Africa to achieve sustainable development goals.
DISCUSSION AND EXCHANGE
Reflection time to identify a relevant topic in the participants' location
Session 4: MONDAY, 22.05.2023 - BIAS
2:00 - 4:00 CEST
INPUT & DISCUSSION
Talk by Jose Eduardo (Researcher, Europeana, Europe): AI at the Europeana. Strategies for mitigating bias.
The talk covers some of the recent Machine Learning work at Europeana. Special focus on strategies to improve data quality and mitigate bias, such as documentation, dataset curation, evaluation and model interpretability.
WORKSHOP
by the project "Fair Diffusion": Instructing text-to-image models on fairness (Felix Friedrich, Patrick Schramowski)
As artificial intelligence continues to shape our world, it is important to confront the biases that are inherent in these systems and to develop effective strategies for mitigating their negative impacts. This is particularly important for text-conditioned image generation AI models, which are increasingly being employed in a wide range of applications.
In this workshop, we will explore the biases that are present in popular diffusion models and examine the sources from which they originate. We will also present practical solutions for combatting these biases and provide strategies for attenuating them after generative text-to-image models have been deployed. Specifically, we will explore methods for shifting biases based on human instructions, which can be used to produce a wider range of outcomes that more fairly represent different identity groups.
Session 5: MONDAY, 05.06.2023 - GENDER EQUALITY
1:00 - 3:00 CEST
INPUT & DISCUSSION
Talk by Marcelo Torres Llamas (Laboratoria (Latin America): A Brief Atlas of Gender and AI
During this talk, the participants will learn about three aspects in which AI is impacting gender equality.
1. How is artificial intelligence manufactured? We will learn how women are impacted in distinct ways by the hardware manufacturing methods, as well as the training and tagging procedures, and the individuals responsible for performing the technical work behind the scenes.
2. How does the technology operate and what roles do models play? We'll examine black box thinking and gender biases and whether there's a way to prevent ongoing discrimination from the ground up. This is not only a matter of data, but also of building gender-balanced teams and aligning goals.
3. What outcomes does it generate? The social repercussions of the technology and its consequences may have distinct effects on men and women, so we need to reflect on the context in which these outcomes are adjusted and/or regulated.
DISCUSSION AND EXCHANGE
Reflection time to identify a relevant topic in the participants' location and start of the idea creation for the hackathon
Session 6: SATURDAY 10.06.2023 - SUSTAINABILITY
1:00 - 3:00 CEST
INPUT & DISCUSSION
Talk by Tijs Vandemeulebroucke (Sustainable AI Lab (Europe)): AI and Sustainability – A Difficult Ethical Conundrum
Tijs Vandemeulebroucke will present the relation between artificial intelligence and sustainability as a difficult ethical conundrum, namely that the physical and digital development of artificial intelligence, which can be used for sustainable natural end goals, heavily impacts natural sustainability itself. After setting the natural context in which artificial intelligence is developed and used, by sketching the current climate and biodiversity crises, a deeper analysis of the concept “Sustainable AI” will be carried out. This is done by zooming in on the difference between the concepts “AI for Sustainability” and the “Sustainability of AI”. While the first concept entails artificial intelligence as an isolated tool to be used in natural sustainability practices, the latter concept entails artificial intelligence as a “world object”, impacting the world as a whole. By describing artificial intelligence as a “world object” a broader account of the natural impact of artificial intelligence will be given. Finally, based on the sketched concepts and reflections, some rudimentary future actions regarding “Sustainable AI” will be sketched.
DISCUSSION AND EXCHANGE
Discussion and idea creation for the hackathon
Session 7: SATURDAY, 17.06.2023 - HEALTH DATA AND ETHICS
2:00 - 4:00 CEST
INPUT & DISCUSSION
Talk by Effy Vayena, Module Head and Principal Investigator of Health Data Governance and Value Creation, ETH Zürich: AI in Health- opportunities and ethical risks
DISCUSSION AND EXCHANGE
Reflection time to identify a relevant topic in the participants' location
Session 8: SATURDAY, 24.06.2023 - MARKETPLACE OF IDEAS
2:00 - 4:00 CEST
PRESENTATION OF PROJECTS/ IDEAS
TEAM FORMATION FOR THE HACKATHON
EXCHANGE IN TEAMS
***********
The program is not public; only participants of the fellowship-program can take part.