Selected Publications.

• Mikhail Soutchanski   Planning as Reasoning in the Real World. Workshop on Highlights of Reasoning about Actions, Planning and Reactive Synthesis (ActSynt) at the 27TH European Conference On AI (ECAI), 19-24 October 2024, Santiago de Compostela.

• Mikhail Soutchanski and Ryan Young   Planning as Theorem Proving with Heuristics. This is a revised and improved version of our Arxiv paper. We fixed a minor bug in hash function implementation in October 2023, and this resulted in the significant improvement of the performance of our TPLH planner. We use the usual IPC score (based on plan length) and a similar metric IPC score but based on the number of situations/states visited as metrics for comparison. The experimental results show that our TPLH planner is comparable to the state-of-the-art in planning: FastDownward (FD) and Best-First Width Search (BFWS) planners. These results show that deductive planning is a productive research direction. Our revised paper is also available in the workshop proceedings published by https://ceur-ws.org/Vol-3585/ (CEUR, volume 3585).
  This paper was blindly peer reviewed by the two independent program committees of the 11th IPS and the 30th RCRA workshops that are part of the the 22nd International Conference of the Italian Association for Artificial Intelligence (AI*IA 2023), November 6-9, 2023. It was accepted for publications in the joint proceedings of these workshops to appear as a CEUR online volume.
• Shaun Mathew and Mikhail Soutchanski   Heuristic Planning for Hybrid Dynamical Systems with Constraint Logic Programming. We discussed a Bouncing Balls Example in the paper. The Prolog source code of this example works with ECLiPSe Constraint Logic Programming System.
  This paper was blindly peer reviewed by the 3 independently program committees and accepted for presentation at the AIRO-2023 workshop, the 30th RCRA Workshop (Experimental Evaluation of Algorithms for Solving Problems with Combinatorial Explosion), and the 11th Italian Workshop on Planning and Scheduling (IPS-2023). These workshops are held in conjunction with the 22nd International Conference of the Italian Association for Artificial Intelligence (AI*IA 2023), November 6-9, 2023. Our paper is also available in the workshop proceedings published by https://ceur-ws.org/Vol-3585/ (CEUR, volume 3585).
• Afifa Tariq, Richard Valenzano, Mikhail Soutchanski   Action Schema Networks for Numeric Planning. Published: 15 Jun 2023, Last Modified: 29 Jun 2023. Heuristics and Search for Domain-independent Planning (HSDIP workshop at ICAPS 2023). How to cite: bibtex.txt
  Abstract: Planning is the fundamental ability of an intelligent agent to reason about what decisions it should make in a given environment to achieve a particular set of goals. Generalized planning is the task of finding a generalized policy that applies to a set of planning instances that share a standard model. Action Schema Networks (ASNets) is an approach to find generalized policies for classical planning problems. In this paper, we extend ASNet to work with numeric planning problems. We use a technique to propositionalize numeric variables, which converts them from infinite ranges to a finite domain, and update the training procedure to use exploration in order to increase the diversity of states encountered. We also use a non-generalized numeric planner, Expressive Numeric Heuristic Search Planner (ENHSP), to teach ASNet to solve numeric planning problems by learning to mimic the actions chosen by ENHSP for problem instances. ASNet finds a generalized policy and weights after training, allowing it to share these to solve unseen problem instances of the same domain. We analyze our approach through an extensive experimental study aimed at evaluating the performance of ASNet on several numeric planning domains. The results show that our numeric ASNet can effectively solve problems in many numeric planning domains.
• Mikhail Soutchanski and Ryan Young   Planning as Theorem Proving with Heuristics. Submitted to arXiv on March 23, 2023.
  Abstract: Planning as theorem proving in situation calculus was abandoned 50 years ago as an impossible project. But we have developed a Theorem Proving Lifted Heuristic (TPLH) planner that searches for a plan in a tree of situations using the A* search algorithm. It is controlled by a delete relaxation-based domain independent heuristic. We compare TPLH with Fast Downward (FD) and Best First Width Search (BFWS) planners over several standard benchmarks. Since our implementation of the heuristic function is not optimized, TPLH is slower than FD and BFWS. But it computes shorter plans, and it explores fewer states. We discuss previous research on planning within KR and R and identify related directions. Thus, we show that deductive lifted heuristic planning in situation calculus is actually doable.
• KR-2022 Tutorial: Hybrid Temporal Situation Calculus for Planning with Continuous Processes @ the Federated Logic Conference (FLoC-2022).
• Bita Banihashemi, Shakil M. Khan, Mikhail Soutchanski From Actions to Programs as Abstract Actual Causes. Proc. of the 36th AAAI Conference on Artificial Intelligence (AAAI-2022), pages 5470-5478. (Acceptance rate: 15%, after double blind peer-review process.) https://doi.org/10.1609/aaai.v36i5.20485 (DOI).
• Shakil M. Khan, Mikhail Soutchanski Necessary and Sufficient Conditions for Actual Root Causes. Proc. of the 24th European Conference on Artificial Intelligence (ECAI), Santiago de Compostela, Spain, 10-12 June 2020 (about 20% acceptance rate).
  ActualCauses-Blocks_2022.pl This is a PROLOG program mentioned in our paper. This program demonstrates how reasoning about actual causes can be implemented using the well-known Blocks World domain. But any other domain can be similarly encoded and explored for the task of computing actual causes. Available under the MIT License for research evaluation purposes. As of Summer 2022, there is no well established set of benchmarks for computing actual causes correctly and efficiently except of a few small examples popular in the literature about actual cause. Moreover, there are no generally accepted and well established criteria for comparison of different definitions of actual cause. This PROLOG program is provided to stimulate development of rigorous criteria, implementations and benchmarks that can be used by the research community to evaluate the existing and new definitions of what is actual cause.
• Vitaliy Batusov and Mikhail Soutchanski   A Logical Semantics for PDDL+, International Conference on Automated Planning and Scheduling (ICAPS2019), July 11-15, 2019 in Berkeley (CA), USA. This paper builds on "Hybrid Temporal Situation Calculus" linked below. (Acceptance rate: ~25%, double blind peer-review process).
• Shakil M. Khan, Mikhail Soutchanski Towards A Logical Account of Epistemic Causality. Our paper was peer-reviewed and accepted for presentation at the 4th Workshop on Formal Reasoning about Causation, Responsibility, and Explanations in Science and Technology (CREST 2019), part of ETAPS 2019 series, Sunday, 7 April, Charles University, Prague, Czech Republic. Available also from Electronic Proceedings in Theoretical Computer Science vol 308, pp. 1–16. ArXived at: http://dx.doi.org/10.4204/EPTCS.308.1
• Vitaliy Batusov, Giuseppe De Giacomo and Mikhail Soutchanski Hybrid Temporal Situation Calculus, arXiv:1807.04861, Submitted on 12 Jul 2018. Comments are welcome!   A shorter version (peer reviewed) of this paper has been accepted to the 32nd Canadian Conference on Artificial Intelligence, Kingston, Ontario, Queen’s University, May 28 to May 31, 2019. The proceedings are published by Springer, Lecture Notes in Artificial Intelligence, vol. 11489, Pages 173-185. It can be downloaded from Springer Web page, ISBN 978-3-030-18304-2. Even shorter peer-reviewed version has been published in the proceedings of the 34th ACM/SIGAPP Symposium On Applied Computing (SAC2019), KRR - Knowledge Representation and Reasoning Track, (24% acceptance rate), pages 1162-1164, Limassol, Cyprus, April 8-12, 2019.
  Abstract:
  The ability to model continuous change in Reiter's temporal situation calculus action theories has attracted a lot of interest. In this paper, we propose a new development of his approach, which is directly inspired by hybrid systems in control theory. Specifically, while keeping the foundations of Reiter's axiomatization, we propose an elegant extension of his approach by adding a time argument to all fluents that represent continuous change. Thereby, we insure that change can happen not only because of actions, but also due to the passage of time. We present a systematic methodology to derive, from simple premises, a new group of axioms which specify how continuous fluents change over time within a situation. We study regression for our new temporal basic action theories and demonstrate what reasoning problems can be solved. Finally, we formally show that our temporal basic action theories indeed capture hybrid automata.
• Shakil Khan and Mikhail Soutchanski Diagnosis as Computing Causal Chains from Event Traces, the AAAI 2018 Fall Symposium on Integrating Planning, Diagnosis, and Causal Reasoning, Oct 19-21, 2018, West Arlington Gateway, Arlington, VA, USA.
• Vitaliy Batusov and Mikhail Soutchanski ``Situation Calculus Semantics for Actual Causality" ( AAAI2018 ). Proceedings of the 32nd AAAI Conference on Artificial Intelligence, February 2–7, 2018, New Orleans, Lousiana, USA. This version includes proof sketches relating our defitinion of achievability actual cause with the modified Halpern-Pearl definition based on structural causal models. (Acceptance rate: 25%, double blind peer-review process).
• Vitaliy Batusov and Mikhail Soutchanski: A different version of our paper with a broader definition of actual causality has been presented at the 13th International Symposium on Commonsense Reasoning (Commonsense-2017), held at the University College London, UK, November 6-8, 2017. This version includes maintenance causes in addition to achievability causes. Our paper (peer-reviewed by a program committee) has been scheduled first to open this symposium, apparently as the best paper of this symposium. Our paper is included into the Proceedings of the 13th International Symposium on Commonsense Reasoning, CEUR, ISSN 1613-0073, vol 2052.
• Denis Ponomaryov and Mikhail Soutchanski   Progression of Decomposed Local-Effect Action Theories, a long (38 pages) version of our paper with all proofs included. Abstract. Published in the ACM Transactions on Computational Logic (TOCL), volume 18, Issue 2, June 2017, Article #16. DOI in ACM digital library.
• A collection (a ZIP file, 29MB) of exam problems on   organic chemistry synthesis (updated July 20, 2017, initially posted on September 21st, 2016).   They are formulated in the basic STRIPS fragment of PDDL with types that is extended with negation of equality. The organic synthesis problem requires finding a sequence of reactions that can produce the target molecule from a given set of initial molecules. Most of these problems are taken from the past undergraduate organic chemistry exams offered at the MIT (Massachusetts Institute of Technology). This set of exam problems is distributed as a benchmark to encourage development of new efficient Artificial Intelligence planners that can compete with humans solving the organic synthesis problem. The Planning Domain Definition Language (PDDL) is a computer language developed for international planning competitions. This benchmark and its provenance are described in the SPARK-2016 and GCAI-2015 papers linked below. Please cite these papers if you publish your research related to this benchmark. The GCAI-2015 paper mentions a different formulation with the PDDL derived predicates (axioms), but the PDDL domain in the linked ZIP file does not use them, for simplicity. (Warning: you may need to introduce the :negated-preconditions requirement to the :requirements list to make sure your planner can properly process this domain file.)

  This benchmark was part of the domain submission that received the Outstanding Domain Submission Award (scroll to the bottom of the Web page) awarded by the Organizers of the 2018 International Planning Competition to Hadi Qovaizi, Arman Masoumi, Anne Johnson, Russell Viirre, Andrew McWilliams, and Mikhail Soutchanski (Faculty of Science, Ryerson University). The results were presented at the 28th International Conference on Automated Planning and Scheduling on Friday, June 29 in Delft, Netherlands.
  The 2018 International Planning Competition included a set of planning problems related to alkene reactions. This set of problems was designed by Dr. Russell Viirre (Dept of Chemistry, Ryerson University, Toronto). It includes exercises related to Markovnikov's rule that can be challenging for undergraduate Chemistry students. Hadi Qovaizi converted this set of exercises into PDDL as part of his Master's thesis research. Hadi was the only CS person who contributed to this Alkene project. If you use this PDDL code for your research please acknowledge both Dr. Russell Viirre and Hadi Qovaizi. This work is licensed under a Creative Commons Attribution, NonCommercial, ShareAlike 3.0 Unported License.
  The original set of several problems from the MIT organic chemistry exams was encoded by Abraham Heifets using a specialized cheminformatics language. Our papers describe the revisions that we had to do before we could produce a PDDL encopding of the exam problems.

• Rami Matloob and Mikhail Soutchanski
  Exploring Organic Synthesis With State-of-the-Art Planners. Proceedings of Scheduling and Planning Applications woRKshop (SPARK), pages 52-61, held on June 13, 2016 in conjunction with the 26th International Conference on Automated Planning and Scheduling, London, UK.
• Arman Masoumi, Megan Antoniazzi and Mikhail Soutchanski
  Modeling Organic Chemistry and Planning Organic Synthesis.
  Proceedings of the 1st Global Conference on Artificial Intelligence (GCAI 2015): Tbilisi, Georgia, October 16-19, 2015. This paper was published in the EasyChair Proceedings in Computing (EPiC Series in Computing): Volume 36, pages 176-195, 2015 (issn 2040-557X).
• Diego Calvanese, Giuseppe De Giacomo, Mikhail Soutchanski On the Undecidability of Basic Action Theories Extended with Description Logic Ontologies, the 24th International Joint Conference on Artificial Intelligence (IJCAI), 2015, Knowledge Representation and Reasoning track, Buenos Aires, Argentina, July 25th -- July 31st, 2015.
• Arman Masoumi, Mikhail Soutchanski and Andrea Marrella Organic Chemistry Synthesis as Artificial Intelligence Planning (15 pages, single column, LNCS style). Published in the proceedings of the International Semantic Web Applications and Tools for Life Sciences Workshop ( SWAT4LS-2013), Edinburgh, Dec. 10th, 2013.
• Sotirios Liaskos, Shakil Khan, Mikhail Soutchanski, and John Mylopoulos Modeling and Reasoning with Decision-Theoretic Goals (14 pages). Runner-up for the best paper award. The proceedings of the 32nd International Conference on Conceptual Modeling ( ER 2013), Hong Kong, November 11-13, 2013. Lecture Notes in Computer Science, volume 8217, pages 19-32.
• Denis Ponomaryov and Mikhail Soutchanski Progression of Decomposed Situation Calculus Theories (7 pages, not including references). This is a revised version of the paper published in the proceedings of the Twenty-Seventh Conference on Artificial Intelligence (AAAI-2013), held in Bellevue, Washington, USA, from July 14-18, 2013.
• Arman Masoumi and Mikhail Soutchanski. Reasoning about Chemical Reactions using the Situation Calculus. Published in the proceedings of the Discovery Informatics Symposium: The Role of AI Research in Innovating Scientific Processes. AAAI Technical Report FS-12-03, pages 35-44. This symposium was part of the AAAI Fall Symposium Series at Arlington, VA, USA, November 2-4, 2012. Original 9 pages version of this paper with a better formatting.
  A preliminary version of this paper was presented to the 2nd Deep Knowledge Representation Challenge (DKRC2012) Workshop, co-located with the International Workshop on Qualitative Reasoning (QR-2012), on July 17, 2012 in Playa Vista, California, USA.
• Wael Yehia and Mikhail Soutchanski Towards an Expressive Practical Logical Action Theory (11 pages, single column version). Proceedings of the 25th International Workshop on Description Logics (DL2012), Rome, Italy, June 7-10, 2012. Published online as CEUR Workshop Proceedings, vol 846 (ISSN 1613-0073). Another (8 pages, double column) version has been also accepted to the 8th International Cognitive Robotics Workshop, July 22-23, Toronto (Canada), held as a part of the AAAI-12 Workshop Program.
  A longer version of this paper (19 pages, single column, PDF file) with technical background and proofs has been published in the proceedings of The Alan Turing Centenary Conference (Turing-100), Manchester (UK), June 22-25, 2012, editied by Andrei Voronkov, EPiC Series, volume 10, pages 307-325, published by EasyChair, issn 2040-557X.

• Wael Yehia, Hongkai Liu, Marcel Lippmann, Franz Baader, Mikhail Soutchanski Experimental Results on Solving the Projection Problem in Action Formalisms Based on Description Logics. Proceedings of the 25th International Workshop on Description Logics (DL2012), Rome, Italy, June 7-10, 2012. Published online as CEUR Workshop Proceedings, vol 846 (ISSN 1613-0073).

• Wael Yehia, Mikhail Soutchanski Evaluation of OWL Reasoners on Solving the Projection Problem: Experimental Study. Proceedings of the 1st International Workshop on OWL Reasoner Evaluation (ORE-2012). In conjunction with the 6th International Joint Conference on Automated Reasoning (IJCAR), Manchester, UK, July 1st, 2012. Published online as CEUR Workshop Proceedings, vol 858 (ISSN 1613-0073).

• Fangzhen Lin and Mikhail Soutchanski. Causal Theories of Actions Revisited. Published in the proceedings of the Tenth International Symposium on Logical Formalization on Commonsense Reasoning, held as part of the AAAI Spring Symposium Series at Stanford University, March 21-23, 2011.
• Yilan Gu and Mikhail Soutchanski. Order-Sorted Reasoning in the Situation Calculus, published in the Proceedings of Commonsense-2009, the Ninth International Symposium on Logical Formalization on Commonsense Reasoning, that was held on June 1-3, 2009, at the Fields Institute, Toronto, Canada. This paper was also published by "Digital Commons @ Ryerson" on May 4, 2009, as A Logic of Actions Revisited, Technical Reports, Paper 1. A revised and improved version (20 pages) is published in the book ``Knowing, Reasoning, and Acting: Essays in Honor of Hector J. Levesque", edited by Gerhard Lakemeyer and Sheila A. McIlraith, College Publications, vol 16, 2011, ISBN-13: 978-1-848900448.
• Yilan Gu and Mikhail Soutchanski. "A Logic For Decidable Reasoning About Actions", Technical Reports, Paper 2: this version had been completed on October 6, 2008. Published by "Digital Commons @ Ryerson" on May 4, 2009. An extended and revised version is published by Annals of Mathematics and Artificial Intelligence journal, Volume 58, February 2010, Numbers 1-2, pages 3-83, with the title "A description logic based situation calculus", DOI 10.1007/s10472-010-9176-z
• Mikhail Soutchanski and Paulo Santos. Reasoning about Dynamic Depth Profiles (draft). Published in the Proceedings of the 18th biennial European Conference on Artificial Intelligence, pages 30-34. University of Patras, Patras, Greece, July 21-25, 2008.

• Yilan Gu and Mikhail Soutchanski. Reasoning about Large Taxonomies of Actions. In the proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence, pages 931-937. Chicago, Illinois, USA, July 13-17, 2008.

• Yilan Gu and Mikhail Soutchanski. Modular Basic Action Theories. In the proceedings of the Seventh IJCAI International Workshop on "Nonmonotonic Reasoning, Action and Change", Sunday 7 January 2007, Hyderabad, India.

• Yilan Gu and Mikhail Soutchanski. Decidable Reasoning in a Modified Situation Calculus. In the proceedings of the Twentieth International Joint Conference on Artificial Intelligence (IJCAI-07), Hyderabad, India, January 2007. Slides of the presentation.

• Yilan Gu and Mikhail Soutchanski. A Logic For Decidable Reasoning About Services. In the proceedings of the 4th International Workshop on AI for Service Composition (page 1-8), held in conjunction with the 17th European Conference on AI (ECAI 2006). Also in the proceedings of the workshop AI-Driven Technologies for Services-Oriented Computing held in conjunction with the Twenty First National Conference on Artificial Intelligence (AAAI-06), Boston, Massachusetts, USA, July 16 2006. Another version of this paper (with the title "The Two-Variable Situation Calculus") has been accepted to the 3rd European Starting AI Researcher Symposium ECAI-06 in Riva del Garda, Italy, on August 28-29, 2006.

• Mikhail Soutchanski, Wee Pham and John Mylopoulos. Decision Making in Uncertain Real-World Domains Using DT-Golog, (6 pages version) published in the proceedings of the ECAI'06 W/Sh on Planning, Learning and Monitoring with Uncertainty and Dynamic Worlds (pages 36-41).
  Mikhail Soutchanski, Wee Pham and John Mylopoulos. Decision making in large-scale domains: a case study, a long (14 pages) version of the paper that has been accepted (with this title) as a poster to the 17th European Conference on AI, Riva del Garda, Italy (Aug 28th - Sept 1st, 2006). You can download Golog programs, a logical model and simulator mentioned in this paper.

• High-Level Robot Programming in Dynamic and Incompletely Known Environments (Ph.D. thesis). PostScript files: letter and A4 formats. PDF files: letter and A4 formats. Code from Appendices (includes off-line DTGolog interpreter and on-line DT-Golog interpreter)

• High-level Robot Programming and Program Execution: letter and A4 formats. Published in the proceedings of the Workshop on Plan Execution, 10th June 2003, held in conjunction with International Conference on Automated Planning and Scheduling (ICAPS'03), Trento, Italy.
• An On-line Decision-Theoretic Golog Interpreter, IJCAI-2001, Seattle, Washington, August 4-10, 2001: Postscript and PDF files.
  Slides of my talk.
• Technical report: On-line Decision-Theoretic Golog Interpreter (the version of August 15, 2000). The 2nd International Cognitive Robotics Workshop (held in conjunction with ECAI-2000), Berlin, Germany, August 21-22, 2000.
• Decision-Theoretic, High-Level Agent Programming in the Situation Calculus. C. Boutilier, R. Reiter, M. Soutchanski and S. Thrun. AAAI-2000, Seventeenth National Conference on Artificial Intelligence, Austin, Texas, July 30-Aug 2, 2000.


• C. Boutilier, R. Reiter, M. Soutchanski and S. Thrun. Decision-Theoretic, High-Level Agent Programming in the Situation Calculus. Proceedings of the Workshop on Decision-Theoretic Planning (ed. by Sven Koenig), Seventh International Conference on Principles of Knowledge Representation and Reasoning (KR2000), Breckenridge, Colorado, Apr 12-15, 2000.
• Execution Monitoring of High-Level Temporal Programs. Accepted by the program committees of the IJCAI-99 Workshop "Robot Action Planning" (31 July 1999, Stockholm) and of the IJCAI-99 Workshop "Scheduling and Planning meet Real-time Monitoring in a Dynamic and Uncertain World" (August 2, 1999, Stockholm).
• Execution Monitoring of High-Level Robot Programs (by Giuseppe De Giacomo, Ray Reiter and Mikhail Soutchanski), the revised version. Proceedings of the 6th International Conference on Principles of Knowledge Representation and Reasoning (KR'98), Trento, Italy, June 2-5, 1998. PostScript file, PDF file.
• Execution Monitoring of High-Level Robot Programs (same authors), a first version of that paper. This version includes Prolog code of the implementation (typos are corrected thanks to Marcus Bjareland). Abstract. Published in working notes of the 4th Symposium on Logical Formalizations of Commonsense Reasoning (London, UK, Jan. 6 - Jan9, 1998).
  PostScript file, PDF file. Slides of my talk.
• Spatial Reasoning and Robotics: the document written for my depth oral exam, slides and bibliography archive (a PDF file ).
• Revised and improved version of Transition System Semantics for Deterministic Situation Calculus (PDF file) and abstract of this paper (here is the link to compressed PostScript file). Original version has been published in the proceedings of the 3rd Symposium on Formalization of the Commonsense Reasoning, Stanford (CA, USA), January 6-8, 1996. Compressed PostScript file with slides of my talk is also available.
• Vladimir Akimov and Mikhail Soutchanski. Automata Simulation of N--Person Social Dilemma Games (April 1993): abstract (text file), PostScript file (210K), and PDF version. Published in: Journal of Conflict Resolution, 1994, v. 38, N 1 (March).
• Logical Theory of Concurrent Actions: English version (a PDF file, version of August 1993) and a published Russian version (March 1993). Russian version was published in "Izvestiya of Russian Academy of Sciences, seria Tekhnicheskaia Kibernetika", 1993, N5. The official translation of this version was published in: Journal of Computer and System Sciences International, 1994, vol.32, N6, pages 72-86.
• Qualitative Reasoning about Physical Systems, review (in Russian): draft in PDF format This review was written in April 1992. Here is a draft in PDF format version that was published in the journal "Izvestiya of Academy of Sciences, seria Tekhnicheskaya Kibernetika", 1992, N5, p.82-96 (in Russian); translated in: Journal of Computer and Systems Sciences International, 1994, vol. 32, N 1, pages 82-96.
• Mikhail Soutchanski An adaptive algorithm to find a weakly efficient action while interacting with a stationary random environment (in Russian). Published in the journal "Izvestiya of Academy of Sciences, seria Tekhnicheskaya Kibernetika", 1987, N2, p.164-174. English translation in: Journal of Computer and Systems Sciences International, 1987, vol. 25, N 3, pages 148-157. This paper proposes a finite automaton that can determine which action out of finitely many alternatives can be considered an "optimal" choice, where optimality is understood as a balance between multiple criteria that might be competing with each other. For each selected action and for all criteria, the feedback evaluating the action according to one of the criteria is provided randomly with a fixed probability. These probabilities can be different for different criteria. The probabilites are not known in advance. However, it is known that the probabilities do not change over time. This paper provides only theoretical results about the proposed finite automaton. The phrase "weakly efficient action" in the title of the paper refers to an alternative that is not stictly dominated by other alternatives, as defined in this Wikipedia article. The proposed automaton is motivated by the task of constructing a semi-decentralized mechanism for finding a Pareto optimal solution in N-person generalization of Prisoner's Dilemma, also known as a Social Dilemma, when all players interact with unknown stationary random environment. This setting is challenging because Nash equilibria in Social Dilemma games are not desirable, and the players have to balance exploration of the stochastic environment with exploitation of individually optimal actions and with fiding an action that benefits the society of players as the whole, but not just an individual player.