Emergent Language as an Approach to Conscious AI
Zengqing Wu, Chuan Xiao
arXiv (Cornell University) June 4, 2026 preprint DOI: 10.48550/arxiv.2606.06380 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper |
|---|---|
| Population | Simulated multi-agent reinforcement learning agents |
| Key points | The authors propose emergent language in multi-agent reinforcement learning as a generative methodology for studying consciousness-relevant structure, arguing it avoids human language priors by having agents start with minimal language and develop communication under task pressure. In a minimal proof-of-concept environment, agents developed self-referential communication, including an echo-mismatch detection circuit that the authors contend is not predicted by task structure or architecture alone but emerges from a specific environmental affordance. |
Abstract
The question of whether artificial systems can be conscious remains open, in part because existing approaches either evaluate systems against theory-derived checklists (discriminative) or engineer consciousness-inspired modules directly (architectural); both leave open whether observed structures are artifacts of human language priors. We propose a generative methodology: emergent language (EL) in multi-agent reinforcement learning, where agents start from minimal (no language, no concept of self, minimal exposure to human text) and develop communication under task pressure alone, ensuring causal attributability to task demands rather than inherited human language priors. We position our methodology by discussing how EL serves as a generative tool for studying consciousness-relevant structure, including the role of environment complexity and the interpretation of emergent communication. As a proof of concept, we instantiate this methodology in a minimal environment and show that agents develop self-referential communication, including an echo-mismatch detection circuit that is not predicted by task structure or architecture alone but emerges from a specific environmental affordance.