|
Published: 11th July 2023 DOI: 10.4204/EPTCS.379 ISSN: 2075-2180 |
The TARK conference (Theoretical Aspects of Rationality and Knowledge) is a conference that aims to bring together researchers from a wide variety of fields, including computer science, artificial intelligence, game theory, decision theory, philosophy, logic, linguistics, and cognitive science. Its goal is to further our understanding of interdisciplinary issues involving reasoning about rationality and knowledge.
Previous conferences have been held biennially around the world since 1986, on the initiative of Joe Halpern (Cornell University). Topics of interest include, but are not limited to, semantic models for knowledge, belief, awareness and uncertainty, bounded rationality and resource-bounded reasoning, commonsense epistemic reasoning, epistemic logic, epistemic game theory, knowledge and action, applications of reasoning about knowledge and other mental states, belief revision, computational social choice, algorithmic game theory, and foundations of multi-agent systems. Information about TARK, including conference proceedings, is available at the website http://www.tark.org/
These proceedings contain the papers that have been accepted for presentation at the Nineteenth Conference on Theoretical Aspects of Rationality and Knowledge (TARK 2023), held between June 28 and June 30, 2023, at the University of Oxford, United Kingdom. The conference website can be found at https://sites.google.com/view/tark-2023
The conference is enlivened by four invited talks, given by:
The Programme Committee received 82 regular paper submissions. Of these, 40 were selected for this volume in a reviewing process during which every paper received three independent expert reviews. Decisions were often difficult and were based on lively discussions between PC members. Of the 40 accepted papers, 21 will be presented as an oral lecture and 19 as a poster presentation accompanied by a flash talk. This volume evidences the interdisciplinary nature of research on theoretical aspects of rationality and knowledge: Several papers are contributions to more than one of the fields listed above, and all of them have been written to be understood by readers across discipline boundaries.
The members of the Programme Committee for the conference were:
Many other people assisted with the reviewing process, including: Leyla Ade, Edoardo Baccini, Philippe Balbiani, Jacques Bara, Fausto Barbero, Gaia Belardinelli, Francesco Berto, Patrick Blackburn, Giacomo Bonanno, Richard Booth, Martin Caminada, Sourav Chakraborty, Michele Crescenzi, Ramit Das, Adam Dominiak, Soma Dutta, Peter van Emde Boas, Jie Fan, Peter Fritz, Asta Halkjær From, Satoshi Fukuda, Paolo Galeazzi, Rustam Galimullin, Avijeet Ghosh, Patrick Girard, Olga Gorelkina, Davide Grossi, Pierfrancesco Guarino, Shreyas Gupta, Jens Ulrik Hansen, Adrian Haret, Aviad Heifetz, Wesley Holliday, Prosenjit Howlader, Neil Hwang, Stephan Jagau, Dominik Klein, Barteld Kooi, Jan Lastovicka, Dazhu Li, Grzegorz Lisowski, Shuige Liu, Emiliano Lorini, Maaike Los, Munyque Mittelmann, Niels Mourmans, Eric Pacuit, Anantha Padmanabha, Timothy Parker, Mina Young Pedersen, Rafael Peñaloza, Elise Perrotin, Charlie Pilgrim, Robert Routledge, Ocan Sankur, Katsuhiko Sano, François Schwarzentruber, Ted Shear, Chenwei Shi, Sonja Smets, Tomasz Steifer, Katrine B. P. Thoft, Paolo Viappiani, Yanjing Wang, Yì Nicholas Wáng, Nic Wilson, Fabio Massimo Zennaro, Stanislav Zhydkov, Gabriel Ziegler, Aybüke Özgün.
I would like to thank the members of the Programme Committee and all other reviewers for the time, professional effort and the expertise that they invested in ensuring the high scientific standards of the conference and its proceedings and for providing a lot of useful suggestions for further improvements to the authors. It was an honor and pleasure for me to read your thoughtful reviews and share in the discussions about the papers. I also thank the authors for their excellent contributions. Moreover, I thank Rob Glabbeek of EPTCS for bringing this volume to publication and for his kind support to the authors and to me as editor.
I want to express my thanks to the organizers of the conference, Mike Wooldridge and Jenny Dollard, for their dedication in bringing TARK 2023 to life in the beautiful grounds of Worcester College at the University of Oxford. Special thanks go to the TARK General Chair Joe Halpern who started the conference series in 1986 and who supported us with his advice in all phases of the conference preparations.
Rineke Verbrugge
Programme chair, TARK 2023
Groningen, The Netherlands
June 20th, 2023
Bayesian equilibrium
constitutes the prevailing solution concept for games with incomplete
information. It is known that from an ex-ante perspective Harsanyi's seminal
notion is related both to Nash equilibrium as well as to canonical correlated
equilibrium. We provide an epistemic characterization of Bayesian equilibrium
from an interim perspective by means of common belief in rationality and a
common prior. Since these epistemic conditions also characterize correlated
equilibrium in the special case of complete information, our result
substantiates that Bayesian equilibrium forms the incomplete information
analogue to correlated equilibrium - and not to Nash equilibrium - in
terms of reasoning.
The goal of the paper is to examine distributed knowledge in groups with differently introspective agents. Three categories of agents are considered: non-introspective, positively introspective, and fully introspective. When a non-introspective agent knows something, she may fail to know that she knows it. On the contrary, when a fully introspective agent knows something, she always knows that she knows it. A fully introspective agent is positively introspective and, when she does not know something, she also knows that she does not know it. We give two equivalent characterizations of distributed knowledge: one in terms of knowledge operators and the other in terms of possibility relations, i.e., binary relations. We show that two different cases emerge. In the first, distributed knowledge is fully determined by the group member who is sophisticated enough to replicate all the inferences that anyone else in the group can make. In the second case, no member is sophisticated enough to replicate what anyone else in the group can infer. As a result, distributed knowledge is determined by a two-person subgroup who can jointly replicate what others infer. The latter case depicts a wisdom-of-the-crowd effect, in which the group knows more than what any of its members could possibly know by having access to all the information available within the group. Finally, we show that distributed knowledge is not always represented by the intersection of the group members' possibility relations. Depending on how introspective agents are, distributed knowledge may be represented by strict refinements of the aforementioned intersection. A full version of the paper is available at https://arxiv.org/abs/2305.08729.
This paper studies extensions of incomplete preferences over basic alternatives to preferences over lotteries on those alternatives. I begin by introducing the normatively and descriptively plausible axiom of parity, an analog of typical principles of dominance: for an outcome $o$ and a lottery $L$, if, for every outcome $o'$ in the support of $L$, both $o \not \succeq o'$ and $o' \not \succeq o$, then both $L \not \succ o$ and $o \not \succ L$. The main result of the paper shows that, in a natural setting, where incomplete preferences are sensitive to an underlying space of totally ordered 'dimensions', the axiom of parity is incompatible with all natural ways of extending preferences from basic outcomes to lotteries over them. In particular, the axiom of parity is inconsistent with the natural principle that, if the outcomes in the support of a lottery only vary along a single dimension, the decision-maker should be indifferent between the lottery and its expected value.