EDAI 2024: Evolutionary Dynamics in social, cooperative and hybrid AI
19.10, 2024, Santiago de Compostela, Spain
This is the second edition of the EDAI workshop, after its success in Kraków (Poland) 2024. Our mission remains the same: Social, cooperative and hybrid AI have increasingly gained attention. Researchers imagine ecosystems wherein (artificial) intelligent agents and humans interact and make decisions, addressing simultaneously, either individually or collectively, a range of heterogeneous problems. To handle the complex dynamics inherent to such systems, and to ensure that results are globally beneficial, research into AI design and analysis methods is needed. This workshop aims to connect the traditions of single- and multi-agent AI research (reasoning, learning, ...) to the areas of (eco-)evolutionary dynamics, typically investigated in the context of complex systems. We hope to foster collaboration and cross-fertilisation of ideas among these communities to advance the areas of social, cooperative and hybrid AI.
News
- 7 Sept 2024: Tentative schedule is published!
- EDAI will take place on Saturday 19th of Oct 2024!
- 19 or 20 Oct 2024: EDAI Workshop in Santiago de Compostela!
- 15 Mar 2024: EDAI website goes live!
EDAI - Workshop at ECAI 2024
Research agendas have been announced around social, cooperative and hybrid AI
wherein ecosystems are
imagined where intelligent agents and humans interact and make decisions, addressing simultaneously
a collection of heterogeneous problems. Such an imagined world implicitly assumes that the personal
beliefs, observed behaviours and potentially the mechanisms for decision-making themselves evolve
over time. Depending on the type of encounters and the frequency of different types of participants,
different outcomes can be imagined and these outcomes may turn out to be beneficial or detrimental
to society. As developers of AI systems, one thus needs to be aware of the complex macro
evolutionary dynamics potentially generated in such ecosystems, even when the design is happening at
the micro-scale.
We see here an opportunity to connect the traditions of single- and multi-agent AI research
(symbolic, probabilistic, …) to the area of evolutionary dynamics and eco-evolutionary dynamics. The
purpose of
this workshop is therefore to promote collaboration and cross-fertilisation of ideas between AI
researchers and researchers typically working on questions related to behavioural dynamics based on
Evolutionary Game Theory and Complex Systems.
The workshop invites high-quality abstracts and papers covering all aspects at the intersection
of evolutionary dynamics with social, cooperative and hybrid AI. The following list is a
non-exhaustive selection of topics:
- Behavioural evolution and learning
- Belief evolution and learning
- Learning in evolving environments
- Evolution, cognition, and reasoning
- Learning to learn from whom
- Complex dynamics in learning ecosystems
- Delegation and principal-agent problem in AI
- Evolution and communication
- Behavioural experiments on social and cooperative AI
- Norm dynamics in hybrid ecosystems
- Trust in evolving AI eco-systems
- Decision-making and emotions
- Dynamics in AI development ecosystems
- Networks in learning and evolution
- Evolution of cooperation with and through AI
Important Dates
Submission Details
You can make your submission through Microsoft CMT. We will accept the following submission types:
Full Submissions (max. 7 pages not considering refs):
- Original full papers: This must be unpublished work
- Transfers from ECAI: papers submitted to ECAI which have not been accepted. Please attach to the PDF of your submission the reviews you received.
Short Submissions (max. 5 pages not considering refs):
- Preliminary results: this format is intended for work in progress which you would like to discuss during the workshop and receive feedback.
- Visionary outlook articles (blue-sky).
Extended Abstracts (1-2 pages not considering refs):
- Extended abstract for articles already published in a journal in the last two years.
- PhD/Early career researchers Highlights: this format is intended for PhD students and early post-doctoral researchers to present ideas for discussion.
Poster to highlight papers accepted in ECAI (1-2 pages not considering refs):
You can submit an extended abstract of a paper accepted in ECAI 2024 and present it as a poster. This will help discuss and promote your work within our community.
In function of the number of submissions you might be selected for a talk or a poster presentation. Except for the PhD/Early career researchers Highlights for which we will set up a round table and each participant will have 5-10 minutes to present their work, and the table will have another 5-10 minutes to discuss.
Accepted Papers
- Understanding the emergence of pro-social behaviour in optional games: a reinforcement learning approach. Authors: Apurva Shah, Paolo Turrini
- Comparing RLHF Approaches for Different Goals. Authors: Oleksandr Skorupskyy, Ivana Dusparic
- AI-Generated Compromises for Coalition Formation. Authors: Eyal Briman, Ehud Shapiro
- Dealing with Expert Bias in Collective Decision-Making. Authors: Axel Abels, Tom Lenaerts, Ann Nowé
- Promoting social welfare or promoting cooperation?. Authors: The Anh Han, Manh Hong Duong, Matjaz Perc
- Bootstrapping the Evolution of Communication via Mimicry. Authors: Dylan R. Cope, Peter McBurney
- Cooperation in Hybrid AI Systems with Private Reputations. Authors: Alexandre S. Pires, Fernando P. Santos
- Social norms dynamics in a behavioral epidemic model on multiplex networks. Authors: Christos Charalambous
- Emergent Communication Under Misinformation. Authors: Heeyoung Lee
- An experimental study on an AI Race. Authors: Elias Fernández Domingos, The Anh Han
Tentative Program
The workshop will take place on Saturday 19th of October 2024!
| Time Slot | Activity |
|---|---|
| 8:15 - 9:00 | Registration |
| 9:00 - 9:30 | Welcome: EDAI Organizers (30 min) |
| 9:30 -10:30 | Session 1 |
| 1. Bootstrapping the Evolution of Communication via Mimicry (30 min) | |
| 2. Emergent Communication Under Misinformation (30 min) | |
| 10:30 - 11:00 | Coffee Break |
| 11:00 - 12:30 | Session 2 |
| 1. Dealing with Expert Bias in Collective Decision-Making (30 min) | |
| 2. AI-Generated Compromises for Coalition Formation (30 min) | |
| 3. Understanding the emergence of pro-social behaviour in optional games: a reinforcement learning approach (15 min) | |
| 4. Comparing RLHF Approaches for Different Goals (15 min) | |
| 12:30 - 14:00 | Lunch Break |
| 14:00 -15:00 | Session 3 |
| 1. Promoting social welfare or promoting cooperation? (30 min) | |
| 2. Cooperation in Hybrid AI Systems with Private Reputations (30 min) | |
| 15:00 - 15:30 | Keynote 2: Fernando P. Santos |
| 15:30 - 16:00 | Coffee Break |
| 16:00 - 17:00 | Session 4 |
| 1. Social norms dynamics in a behavioral epidemic model on multiplex networks (30 min) | |
| 2. An experimental study on an AI Race (30 min) | |
| 17:00 - 17:30 | Closing remarks and open discussion about the future of EDAI and Complex systems in AI |
Organization
The EDAI 2024 workshop is organized by:
- Elias Fernández Domingos (ULB & VUB, Belgium)
- Chaitanya S. Gokhale (University of Würzburg & MPI for Evolutionary Biology, Germany)
- The Anh Han (Teesside University, UK)
- Tom Lenaerts (ULB & VUB, Belgium)
- Ana Paiva (IST & University of Lisbon, Portugal)
- Paolo Turrini (University of Warwick, UK)
Program committee:
Will be announced soon!Contact
If you have any questions about the EDAI workshop, please contact the organizers at:
edai.workshop AT gmail.com