Chanattan R. Sok

CS Student Researcher

I am a Master student at Sorbonne University and École Normale Supérieure (ENS) de Rennes in Computer Science, France, focusing on decision (aid) and AI. My research integrates decision-making (sequential, collective, multi-criteria), societal challenges, and artificial intelligence.

I address interdisciplinary challenges with multi-agent systems (e.g., Mixture of Experts), socio-economics (e.g., mechanism design, fairness, recommendation systems), learning (RL), and game theory. My recent interests focus on auctions, incentives and LLM alignment.

LLMs, Agentic AI Reinforcement Learning Decision-making Social Sciences Game Theory

Publications & Preprints

Currently working on a conference paper on fairness in railway networks, and a workshop on social choice and pedestrian decision.

Towards an Energy Twin: Simulating Global Warming Potential in Hamburg's Building Stock

Chanattan Sok, Rico Herzog, Arnaud Grignard.
In Review — Special Issue Urban Digital Twins
Simulation et résolution de conflits spatiaux entre piétons fondée sur la théorie du choix social

JFSMA 2025 — Chanattan Sok, Olivier Simonin. · English version (ECAI submission)
JFSMA 2025
WebAssembly Serverless Join: A Study of its Application

SSDBM 2024 — Chanattan Sok, Laurent d'Orazio, Reyyan Tekin, Dimitri Tombroff.
SSDBM 2024

Projects

CNM

Conflict Negotiation Model — pedestrian simulation platform using social choice theory. Published at JFSMA 2025.

MAS Python / Mesa Simulation

tweakable_model

Generic n-choice negotiation model inspired from Putnam's two-level game theory. Batch experiments with genetic algorithms.

MAS Python / Mesa

IsomEdit

Isometric map editor built in Java. A personal tool for game design and content creation workflows.

Java

Thoth

Stock investment simulator with AI assistance. Simulated press events with randomised market impact, Brownian motion stock prices, and Thoth — an ARIMA-based AI agent that analyses price curves and news to predict trends over a few months. Human–AI feedback loop for improved synergy, with portfolio and ROI tracking.

Python ARIMA Finance

train_mystery

Collaborative mystery game set aboard a train, written in Rust. A fun game-jam-style project with colleagues.

Rust

satellite_coordination

Satellite coordination via auction-based mechanisms for collective decision-making, and distributed constraint optimisation (SDCOP — Sequential Distributed Constraint Optimisation Problem).

MAS Auctions Python DCOP