The following special sessions will be carried out at CAI 2026:
- (Multi-)Agentic AI in Action: Real-World Applications of Contemporary Multi-Agent Systems
- Emerging Trends in Data, Web, and Social Media Analysis and Generation with LLMs
- AI for Space Exploration
- Foundation and Emerging AI Models for Applications in Finance, Cyber Security, and Life Sciences
- From the Streets to the Systems: A Journey Through Ethical AI, Workforce Justice, and Radical Inclusion
- FIRE-AI: Federated Intelligence for Responsible, Evolving AI
(Multi-)Agentic AI in Action: Real-World Applications of Contemporary Multi-Agent Systems
Proposers
COSTIN BĂDICĂ, MARIA GANZHA, MIRJANA IVANOVIĆ and MARCIN PAPRZYCKI
Rationale and Vision
As highlighted in recent works ([1], [2]), the field of intelligent agents and multi-agent systems is at a fascinating inflection point. Here, two powerful paradigms can be distinguished: the structured, verifiable world of classic Multi-Agent Systems (MAS) and the fluid, semantically rich world of LLM-driven generative agents (Agentic Systems). The most exciting breakthroughs are now happening at their intersection, where these paradigms are not just compared, but actively combined to solve real-world problems.
This special session moves beyond theoretical debates to showcase concrete, impactful use cases and applications. We aim to gather researchers and practitioners who are deploying the next generation of multi-agent systems to tackle complex challenges. The focus is on empirical results, novel use-cases, and the practical lessons learned from building and deploying agentic AI “in the wild.” We are particularly interested in systems that strategically blend the strengths of classic MAS (e.g., explicit coordination, formal negotiation) with the advanced reasoning and communication capabilities of foundation models.
List of Topics
This session invites submissions demonstrating tangible applications of (multi-)agentic AI. The emphasis
is on systems, results, and real-world impact.
- Hybrid Agent Systems in Practice: Case studies of systems successfully integrating classic agent architectures (e.g., BDI, auctions, argumentation) with LLM-based components for enhanced performance.
- Agents for Science and Engineering: Applications in automated scientific discovery, lab automation, complex system design, healthcare & life sciences, and AI-assisted software development.
- Autonomous Business and Economic Systems: Agent-based models for supply chain optimization, automated trading, and decentralized autonomous organizations (DAOs).
- Human-Agent Collectives: Systems and interfaces for effective human-AI teaming in areas like creative design, strategic planning, and complex data analysis.
- Generative Agents for Social Simulation: Using LLM-powered agents to create high-fidelity simulations of social, cultural, and economic dynamics for research and policy-making.
- Agents in Interactive Entertainment: Applications in gaming, interactive narratives, education, and the creation of dynamic, believable non-player characters (NPCs).
- Evaluation of Deployed Agent Systems: Methodologies and metrics for evaluating agent performance,robustness, and safety in real-world, uncontrolled environments.
[1] Costin Bădică, Amelia Bădică, Maria Ganzha, Mirjana Ivanović, Marcin Paprzycki, Dan Selişteanu and Zofia Wrona, Contemporary Agent Technology: LLM-Driven Advancements vs Classic Multi-Agent Systems, arXiv, 2509.02515 [cs.MA], 2025. https://arxiv.org/abs/2509.02515
[2] Zofia Wrona, Maria Ganzha, Marcin Paprzycki, Wiesław Pawłowski, Angelo Ferrando, Giacomo Cabri, and Costin Bădică. Comparison of Multi-Agent Platform Usability for Industrial-Grade Applications. Applied Sciences 14 (22): 10124, 2024. https://doi.org/10.3390/app142210124
Emerging Trends in Data, Web, and Social Media Analysis and Generation with LLMs
Proposers
M. Dolores Ruiz and Maria J. Martin-Bautista
Session description
This session focuses on the transformative role of large language models (LLMs) in enabling advanced analysis and generation of data from the web and social media. It aims to bring together researchers, practitioners, and industry experts interested in how LLMs are reshaping our ability to interpret, summarize, retrieve, and generate content across large-scale, unstructured, and dynamic data environments.
Contributions are welcome from both theoretical and applied perspectives, covering novel methodologies, system architectures, and real-world applications. The session also encourages submissions that address critical issues related to transparency, ethics, and responsible use of LLMs in open and high-impact domains.
Topics of Interest
- LLMs for Social Media Analysis. Investigating the use of LLMs to extract insights from social media platforms, including sentiment and stance detection, topic modelling, misinformation identification, and audience analysis.
- Text Understanding and Knowledge Extraction with LLMs. Approaches leveraging LLMs for summarization, classification, information extraction, question answering, and retrieval-augmented generation (RAG) over complex and noisy text sources.
- Multimodal and Hybrid LLM Applications. Integration of LLMs with other data modalities (e.g., visual, audio, structured data) or symbolic systems to enhance performance in cross-domain analysis and interactive AI systems.
- Real-Time and Context-Aware Processing. Using LLMs for streaming data analysis, adaptive content generation, and real-time decision support, particularly in fast-changing web and social media environments.
- Ethics, Safety, and Responsible Deployment of LLMs. Exploring issues of fairness, bias, explainability, and user privacy in LLM-based applications, with emphasis on accountability and trustworthy AI in high-impact contexts.
Target Audience
This session is intended for researchers, practitioners, data scientists, and industry experts working in LLMs, generative AI, and web/data analysis, with a particular focus on applications to social media and text data analysis. It also welcomes professionals interested in the ethical, methodological, and practical implications of applying LLMs and generative AI to data-driven decision-making.
AI for Space Exploration
Proposers
Ignacio G. López-Francos, Andrés Martinez and Federico Lozano-Cuadra
Abstract
Artificial Intelligence has long played a role in space exploration. NASA has developed and flown AI technologies for over four decades, from early demonstrations in automated planning and scheduling, to Earth observation systems capable of mapping and responding to geographical events from orbit, to autonomous landers and rovers conducting science on Mars. Today, space exploration stands at a critical inflection point where advances in AI (particularly with the recent generative AI wave) have the potential to dramatically expand scientific discovery and mission operational capability.
Several converging drivers underscore the timeliness of this topic. The cost of launch continues to decline, enabling a surge of high-capitalized commercial entrants and accelerating national programs such as NASA’s Artemis campaign, which aims to return humans to the Moon and prepare for Mars exploration. Moreover, NASA JPL’s upcoming CADRE (Cooperative Autonomous Distributed Robotic Exploration) mission will deploy a swarm of rovers on the lunar surface, demonstrating how multi-agent autonomy is becoming central to future space missions. At the same time, intensifying geopolitical competition has catalyzed a renewed space race, reminiscent of the Cold War, but now occurring in a far more complex technological ecosystem. These developments demand unprecedented levels of autonomy, adaptability, and decision support — areas where AI has become indispensable.
This Special Session will address the technical challenges and opportunities of deploying advanced AI in space systems. Key issues include:
- Scalability of science and mission operations: Leveraging AI to analyze massive volumes of multisensor data collected from spacecraft, orbiters, and planetary missions.
- Autonomous exploration: Developing AI-driven agents capable of “single-shot” or minimally supervised missions to remote worlds where real-time ground control is infeasible.
- Validation, verification, and certification: Ensuring trust, safety, and robustness of AI algorithms when human lives and mission-critical objectives are at stake.
- Generalization to novel environments: Designing algorithms that can adapt to out-of-distribution conditions characteristic of extraterrestrial terrains and atmospheres.
- Human-AI teaming: Enabling effective collaboration between astronauts, mission controllers, and AI systems in dynamic and uncertain operational contexts.
The timeliness of this session stems from the convergence of two forces: the maturation of AI into powerful foundation technologies, and the emergence of a new space economy driven by both government agencies and private industry. The synergy between these trends makes AI not only relevant but essential for realizing ambitious goals such as sustained lunar presence, Mars exploration, and long-duration autonomous missions in deep space.
By bringing together experts in AI, robotics, aerospace engineering, and planetary science, this Special Session will provide a focused forum to assess progress, highlight open challenges, and shape the research agenda for the coming decade. In doing so, it will underscore how breakthroughs in AI can unlock new frontiers of human and robotic exploration beyond Earth.
Foundation and Emerging AI Models for Applications in Finance, Cyber Security, and Life Sciences
Proposers
Giuseppe Agapito, Francesco Cauteruccio, Enrico Corradini and Andrea Tundis
Abstract
Artificial Intelligence (AI) is reshaping modern societies, driving innovation in finance, cybersecurity, logistics, and life sciences. With the rise of foundation models and emerging AI architectures, there is both opportunity and urgency to explore how these methods can ensure resilience, trust, and efficiency in critical infrastructures. This special session brings together researchers from diverse domains to discuss methodologies and applications that leverage AI for robustness, security, and sustainability.
The growing complexity of digital ecosystems raises common challenges: how to design infrastructures that remain secure, adaptive, and resilient under uncertainty. In finance, the emergence of digital and central bank digital currencies (CBDCs) demands intelligent mechanisms for stability and systemic risk protection. Similar concerns arise in cybersecurity and networked systems, where AI-driven analysis of complex networks and IoT security are key to anticipating threats and ensuring robustness. In life sciences, large-scale biomedical data, genomics, and proteomics require advanced AI and graph mining methods, supported by high-performance and energy-aware computing to enable scalable, sustainable healthcare analytics.
Bridging these domains highlights a shared reliance on AI to analyze and manage complexity—from spatiotemporal graph transformers for logistics, to algorithms for resilience in digital infrastructures, to biomedical network mining. The integration of network science, intelligent systems, and AI optimization underscores both the interdisciplinary nature of today’s challenges and the opportunities for impactful solutions. This special session timely addresses these converging needs, advancing AI models and applications for critical infrastructures.
Topics of interest include, but not limited to:
- Intelligent systems
- Resilience in Critical Infrastructures
- Artificial Intelligence
- Payment systems
- Digital Currency
- CBDC systems
- Spatiotemporal Graph Transformers
- Last-Mile Delivery Optimization
- Demand Forecasting & OD Estimation
- Resilience & Robustness of Logistics Networks
- Digital Twins for Urban Logistics
- Energy-/Emission-Aware Routing
- Network Science and Analysis
- Complex Networks
- Internet of Things
- Algorithms and Games
- Data Science
- Cybersecurity
- Data Mining for Biomedical Big Data
- Genomics and Proteomics Analytics
- Graph Mining in Life Sciences
- Parallel and Distributed Algorithms for Healthcare
- Machine Learning and AI for Health
- Energy-Aware Computing in Healthcare HPC
- High-Performance Computing (HPC) in Life Sciences
- Biological Network Analysis
From the Streets to the Systems: A Journey Through Ethical AI, Workforce Justice, and Radical Inclusion
Proposers
Chris Hope and Javier Valls Prieto
Abstract
Artificial Intelligence is reshaping economies, governance, and culture, but the communities
most affected by inequity are often the least prepared to navigate this transformation. From
Roma populations in Europe to Black and Latino learners in the United States, systemic barriers
in education, digital access, and workforce opportunity continue to widen the AI literacy gap.
This special session brings together two perspectives grounded in lived experience and legal
scholarship to propose a transatlantic framework for ethical AI inclusion.
Together, they will present three outcomes:
- Fairness metrics to measure and reduce algorithmic bias.
- Culturally grounded, bilingual curricula that expand digital capacity.
- Policy and legal levers bridging U.S. and EU protections.
This session models radical inclusion while challenging IEEE stakeholders to move beyond bias
awareness toward measurable equity and structural justice.
FIRE-AI: Federated Intelligence for Responsible, Evolving AI
Proposers
Dr. Dipanwita Thakur, Prof. Giancarlo Fortino, Prof. Sajal K. Das and Prof. Dr. Torsten Braun
Techinical Description
Federated Learning (FL) has become a foundational approach to privacy-preserving, distributed machine learning across diverse domains such as healthcare, finance, IoT, and edge computing. However, classical FL methods—largely focused on optimization and scalability—face growing limitations as artificial intelligence enters a new era driven by large foundation models (e.g., LLMs and multi-modal architectures). The next generation of federated systems must move beyond efficiency alone, embracing responsibility, autonomy, personalization, and ethical alignment with societal needs.
The FIRE-AI Special Session addresses these challenges by exploring the emerging concept of federated intelligence: distributed AI systems that are not only technically robust but also socially aware, ethically grounded, and capable of adapting foundation models in decentralized environments.
Key issues include:
- Federated fine-tuning of foundation models: techniques for personalization, efficiency, and privacy-aware adaptation.
- Decentralized architectures: robust, agent-based, and zero-trust systems for scalable federated intelligence.
- Responsible and ethical FL: embedding fairness, accountability, regulatory compliance, and sustainability by design.
- Cross-domain applications: federated AI for healthcare, finance, smart environments, and IoT.
- Green and efficient FL: optimization techniques for reducing energy footprint and communication overhead.
The timeliness of this session stems from two converging trends: (1) the deployment of foundation models across sectors where privacy and distributed data are critical, and (2) the growing societal demand for AI that is trustworthy, transparent, and sustainable. Recent workshops (e.g., FLBD2024, FL@FM-NeurIPS 2024, CFAgentic@ICML 2025, FLUID@AAAI 2025) have laid groundwork in federated optimization and privacy, but few have explicitly tackled the intersection of FL, foundation models, and responsible AI. This Special Session aims to fill that gap by fostering a dialogue between federated learning researchers, distributed systems experts, and the responsible AI community.
Through this session, we expect to create a multidisciplinary forum that advances the theory and practice of federated intelligence, identifies open problems at the intersection of technical and ethical dimensions, and outlines a roadmap for future research and deployment.