Sarit Kraus

Professor of Computer Science
Faculty of Exact Sciences
Department of Computer Science
Bar-Ilan University, Ramat Gan, Israel

 
Keynote address:
Smarter Together? Assisting Humans in a World of Intelligent Agents

The rapid advancement of intelligent systems—ranging from large language models (LLMs) to autonomous drones and robots—has unlocked remarkable capabilities while introducing new challenges for human interaction and coordination. As these systems grow in autonomy and complexity, humans may find it increasingly difficult to collaborate effectively or maintain situational awareness.

To address these challenges, intelligent advising agents can support decision-making and reduce cognitive load by providing guidance and explanations that help users assess when and how to act on AI-generated advice. Our overarching goal is to ensure that AI + Human > max(AI, Human)—that is, the combined intelligence of humans and machines should surpass the performance of either working alone.

This presentation explores the key challenges of human collaboration with intelligent systems and introduces methods that enable LLM-based agents to coordinate more effectively with people. These approaches aim to establish shared reference points for understanding, foster goal alignment, and promote adaptive interaction. Ultimately, we outline strategies for developing assisting agents that enhance human performance, trust, and satisfaction within human–AI–robot teams.

Sarit Kraus is a Professor of Computer Science and Head of the Artificial Intelligence research at Bar-Ilan University. Her research focuses on intelligent agents and multi-agent systems that integrate machine learning (including LLMs) with optimization, logic, and game theory. She develops agents capable of interacting and negotiating effectively with people and robots.

She has received many honors, including the IJCAI Computers and Thought and Research Excellence Awards, ACM SIGART and Athena Lecturer Awards, the EMET Prize, and two IFAAMAS Influential Paper Awards. A Fellow of ACM, AAAI, and EurAI, she also received an ERC Advanced Grant and a commendation from Los Angeles city  for the ARMOR system.

Kraus has published over 400 papers, co-authored five books, was IJCAI-2019 program chair and was elected as IJCAI-2027 conference chair.  She is an elected member of the Israel Academy of Sciences and Humanities.

Miłosz Kadziński

Assoc. Professor of Computer Science
Institute of Computing Science
Faculty of Computing and Telecommunications
Poznan University of Technology, Poznan, Poland

 
Keynote address:
When Rules Meet Plurality: From Rough Sets to Group Decision Aiding

The presentation will address the potential of the Dominance-based Rough Set Approach (DRSA) for group decision aiding. We will build on two real-world case studies from the energy sector that illustrate how the method can be adapted to different kinds of collective decision contexts. This perspective is particularly meaningful in Poland – the venue of this year’s GDN conference, where rough set theory originated and where its dominance-based developments have opened an important line of research in preference learning, explanation, and recommendation under multiple criteria.

The first case study concerns the assessment of energy technologies by a large international panel of experts representing different countries, backgrounds, and areas of competence. In such a setting, the central challenge is not only the multiplicity of criteria, but also the diversity and partiality of expertise: not every expert can credibly evaluate every option, and yet the final recommendation must be transparent and collectively meaningful. The presentation will show how DRSA can be extended to learn from partial judgments provided on a carefully selected reference set, how individual rule-based models can be preserved rather than averaged away, and how these models can be aggregated into an interpretable group recommendation that makes both consensus and disagreement visible.

The second case study concerns the acceptance of wind energy technologies and examines how DRSA can support the analysis of evaluations expressed by a large group of stakeholders. Here, the problem is not to identify a single “best” option, but to understand the structure of support, hesitation, and opposition in a socially sensitive decision context. The presentation will show how DRSA can be adapted to explain collective patterns of acceptance, reveal the main drivers of attitudes, and generate decision-relevant insights without sacrificing interpretability.

Taken together, the two studies suggest that DRSA offers more than a classification tool for ordered decision problems. It provides a coherent framework for handling heterogeneous viewpoints, incomplete and unevenly distributed knowledge, and the need for explanation in collective settings. The broader argument of the presentation will be that DRSA is especially well-suited to group decision aiding whenever legitimacy depends not only on the quality of the recommendation, but also on the transparency of the path leading to it.

Miłosz Kadziński is an Associate Professor at Poznan University of Technology. He is the head of the Institute of Computing Science and the Laboratory of Intelligent Decision Support Systems. He has been awarded scientific prizes by EURO, the International Society on MCDM, the INFORMS MCDM Section, the Polish Academy of Sciences, and the Ministry of Science. He acts as the coordinator of the EURO Working Group on MCDA and president-elect of the INFORMS MCDM Section. His primary research interests include preference learning, robustness analysis, and strengthening the interfaces between MCDA and other sub-disciplines of algorithmic decision theory, such as interactive and evolutionary multiple objective optimization, data envelopment analysis, machine learning, and group decision. He has published more than 110 papers in premier international journals, including European Journal of Operational Research, Omega, Decision Support Systems, INFORMS Journal on Computing, Group Decision and Negotiation, Information Sciences, and Information Fusion.

2026 INFORMS GDN Section Award Winner

Institute of Computing Science,
Poznań University of Technology
& Systems Research Institute of the Polish Academy of Sciences

Keynote address:
Constructive Preference Learning

Decision problems are inherently multidimensional. This multidimensionality has three main sources: (i) multiple decision makers, or voters; (ii) multiple evaluation criteria, or objectives; and (iii) multiple uncertain states of the world, leading to risky outcomes, such as gains or losses. In the mid-twentieth century, these three sources gave rise to three major subdisciplines of decision theory: (i) social choice and group decision making, (ii) multiple criteria decision analysis and multiobjective optimization, and (iii) decision making under risk and uncertainty.

In the multidimensional decision problems studied within these subdisciplines, the only objective information stemming from the problem statement is a dominance relation in the set of possible alternatives, such as actions, solutions, candidates, acts, or investments. This dominance relation is a partial order: it is transitive but incomplete. As a result, it does not generally allow one to choose the best alternative, assign alternatives to predefined ordered classes, or rank them from best to worst.

To make alternatives more comparable, one needs to aggregate the multiple dimensions into a preference model richer than the dominance relation. Since these dimensions are usually conflicting, such aggregation should involve the decision makers’ preferences, expressed through subjective preference information. This talk offers a short personal synthesis of constructive preference learning from subjective preference information [1].
We focus on indirect holistic preference information expressed in terms of decision examples provided by decision makers, rather than on the direct elicitation of preference model parameters. The rest of the presentation is concerned mainly with Multiple Criteria Decision Aiding [2].

Decision examples processed through ordinal regression usually do not lead to a unique preference model. For this reason, robust ordinal regression was proposed to account for all instances of an employed preference model that are compatible with the decision examples. Since robust ordinal regression is already fairly well known, we focus instead on a new methodology of preference modeling based on an extended preference structure defined by a new seven-valued logic [3].

Within this framework, we propose procedures for aggregating multiple criteria while accounting for a plurality of perspectives and for indeterminacy in the parameters of the employed preference models. This leads to a mixture of true, false, and unknown preference information. This information is then used to construct a seven-valued preference relation and to formulate robust recommendations based on multiple value functions, outranking models, or if–then decision rules with a specially adapted syntax. These models represent the preferences of a single decision maker or multiple decision makers more faithfully than traditional approaches. Since the construction may be sensitive to small changes in model parameters, we also propose a fuzzy extension of the seven-valued preference system, enabling smooth transitions between preference values and avoiding edge effects. The presentation is illustrated with instructive examples.

References

  • [1]. Hüllermeier, E., Słowiński, R.: Preference learning and multiple criteria decision aiding: differences, commonalities, and synergies. 4OR – A Quarterly Journal of Operations Research, Part I: 22 (2024) 179–209, https://doi.org/10.1007/s10288-023-00560-6, Part II: 22 (2024) 313-349, https://doi.org/10.1007/s10288-023-00561-5.
  • [2]. Greco, S., Słowiński, R., Wallenius, J.: Fifty years of multiple criteria decision analysis: From classical methods to robust ordinal regression. European Journal of Operational Research, 323 (2025) 351-377, https://doi.org/10.1016/j.ejor.2024.07.038.
  • [3]. Greco, S., Słowiński, R.: Preference representation in multiple criteria decision aiding with seven-valued logic. 2025, submitted.

Roman Słowiński is a Professor and Founding Chair of the Laboratory of Intelligent Decision Support Systems at Poznań University of Technology. As a full member of the Polish Academy of Sciences, he was its Vice-President in 2019–2022. In his research, he combines Operational Research and Artificial Intelligence for Decision Aiding. He is a recipient of the EURO Gold Medal and an Officer of the Academic Palms of France. He has been awarded the title of Doctor Honoris Causa by six universities worldwide, including Université Paris Dauphine. He is a laureate of the 2005 Prize of the Foundation for Polish Science, the 2020 Scientific Award of the Prime Minister of Poland, the 2023 Humboldt Research Award (Germany), and the 2026 Jiangsu Friendship Award for foreign experts (China). He is also a member of Academia Europaea and the Accademia delle Scienze in Bologna (Italy), and a Fellow of IEEE, INFORMS, IFIP, IFORS, AAIA, IRSS, IAITQM, CORE, and AIIA. Since 1999, he has been a coordinating Editor-in-Chief of the European Journal of Operational Research (Elsevier). His Google Scholar h-index is over 100.

Google Scholar: https://scholar.google.com/citations?hl=en&user=yCX-JrQAAAAJ

Personal www site: https://fcds.cs.put.poznan.pl/IDSS/rslowinski/cv_en.htm