Integrated Socio-technical Models for Conflict Resolution and Causal Reasoning

• Applicants: Professor Dr. Werner Damm; Professor Dr. Martin Fränzle; Professor Dr. Alexander Pretschner
• Subject Area: Theoretical Computer Science ; Security and Dependability, Operating-, Communication- and Distributed Systems
• Term: from 2020 to 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 433524788

Within the coordinated package proposal PAK 987/1 “Assuring Individual, Social, and Cultural Embeddedness of Autonomous Cyber-Physical Systems”, subproject 2 “Integrated Socio-technical Models for Conflict Resolution and Causal Reasoning” focuses on algorithmic means to detect, explain, resolve, and analyze the conflicts that arise when Human-Cyber-Physical Systems (HCPSes) interact with each other, with the physical world, and with humans in their vicinity. The project’s four work-packages (WPs) will develop a formal modelling framework expressive enough to cover the dynamics as well as legal and regulatory constraints and dynamically shifting personal preferences within HCPSes (within WP1), investigate algorithmic ways to find rational strategies avoiding conflict (WP2), develop means to rigorously explain and communicate local strategic decisions (WP3), and, if conflict could not be averted, means to analyze conflicts ex-post and reason over their causes (WP4). The modeling framework of WP1 will, beyond forming the formal basis for the algorithmic developments of the other work-packages, facilitate the communication and the model and method exchange among those work-packages and with the partner sub-projects (SPs) in the package proposal. From SP1 we will import models of human behavior and interaction, SP3 will provide elements of system designs, and SP4 will capture societal and legal requirements. WP2 will use game theory to describe the design space and model the interaction between agents and will develop an algorithmic game theory covering rational strategy design in heterogeneous multi-agent systems subject to partial observability and delayed feedback. The WP will thereby provide algorithmic means by which the agents can identify beneficial behavioral strategies permitting them to react to changes in their environment while respecting the pertinent legal and regulatory context and currently prevailing individual preferences. An extension of Justification Logic to heterogeneous multi-agent systems subject to regulatory constraints and dynamically changing individual preferences will be the scientific foundation for WP3. It will use it as a basis for the communication between agents and leverage its ability to provide provably correct online explanations of the reasons behind an agent’s choice, thereby helping foster behavioral alignment and trust across agents in general and the human-machine boundary in particular. Finally, WP4 will use the mathematical foundations of causality for ex-post reasoning over peculiar events. Operating ex-post and thereby being liberated from constraints applying to the design-time and run-time algorithms of WP2 and WP3, WP4 can detect and take account of unknown or unmodeled entities in- and outside the system boundaries, including knowledge not available at runtime, in order to reason over the actual causes of unexpected events, to attribute blame, and to guide improvements of the systems for the future.

• DFG Programme: Research Grants