Automated Semantic Rules Detection (ASRD) for Emergent Communication Interpretation

TL;DR

This work tackles the interpretability challenge of emergent communication in multi-agent systems by introducing Automated Semantic Rules Detection (ASRD), an algorithm that extracts semantic rules from agent messages and ties them to input data attributes and hyperattributes in the Lewis Game. ASRD operates through four steps—grouping messages by attribute values, identifying constant token positions, analyzing co-varying positions, and removing globally invariant tokens—to produce patterns that relate message structure to data properties. Experiments on two image-centric datasets, MOPRD and VGHAC, reveal that ASRD can uncover consistent, interpretable patterns in modular, compositional settings (notably with MOPRD and certain image-feature transforms), while more holistic language in VGHAC yields fewer extractable rules and lower TopSim. The findings demonstrate ASRD’s potential to reduce manual interpretation effort in emergent languages and guide future refinements for enhanced interpretability across diverse tasks and datasets.

Abstract

The field of emergent communication within multi-agent systems examines how autonomous agents can independently develop communication strategies, without explicit programming, and adapt them to varied environments. However, few studies have focused on the interpretability of emergent languages. The research exposed in this paper proposes an Automated Semantic Rules Detection (ASRD) algorithm, which extracts relevant patterns in messages exchanged by agents trained with two different datasets on the Lewis Game, which is often studied in the context of emergent communication. ASRD helps at the interpretation of the emergent communication by relating the extracted patterns to specific attributes of the input data, thereby considerably simplifying subsequent analysis.

Figures (4)

• Figure 1: Speaker's architecture chaabouni_emergent_2022vanderplaetse_influence_2024.
• Figure 2: Listener's architecture chaabouni_emergent_2022vanderplaetse_influence_2024.
• Figure 3: Image generated for a combination ($\bullet$, $\times$, $\nwarrow$) from MOPRD.
• Figure 4: Sample images from VGHAC.