Logical Institutional and Narrative Dynamics
Aus International Center for Computational Logic
Logical Institutional and Narrative Dynamics
Vortrag von Fedor Grigoryevich Pogulskiy
- Veranstaltungsort: APB-2026
- Beginn: 13. August 2026 um 11:00
- Ende: 13. August 2026 um 12:00
- Forschungsgruppe: Automatentheorie
- Event series: Research Seminar Logic and AI
- iCal
Human societies are full of heterogenous rule-based structures (norms and institutions) which coexist with each other temporally, but also evolve and transform over time under interaction, documented through primary sources. This presents a modelling challenge, since some generative social models (e.g. game-theoretic ones) compress textual data into simplified, temporally local theoretical assumptions while others use statistical learning over longer times but struggle to generate off-distribution outputs. Meanwhile decision-makers are increasingly interested in scenario models for fields such as privacy and systemic risks to democratic institutions. This student talk will present a proposal for a Master's thesis project, a generative agent-based model that is both data-driven and interactive in an interpretable way. The plan is to extract institutional transition dynamics from prosopographies - temporal KGs of named individuals. Additionally, agents would be equipped with personal narratives of rules and facts. Following the symbolic interactionism approach in sociology, these two elements will act as the rules and constraints under which the social KG evolves. Because all of these components can be expressed in one unified logical representation, the model can be tentatively used both to analyze documented structures (e.g. identifying communities via constraints) and generate counterfactuals to test hypotheses, without any encoding/decoding or collapse of concepts into scalars. Currently, similar work has been done to analyze privacy policies with ASP, but another consideration is Standpoint Logic.
