Emergent Language as an Approach to Conscious AI
This work offers a new methodology for studying consciousness in AI by ensuring causal attributability to task demands, but it is currently a proof of concept with limited empirical results.
The paper proposes using emergent language in multi-agent reinforcement learning as a generative methodology to study consciousness-relevant structures, avoiding human language priors. As a proof of concept, agents developed self-referential communication, including an echo-mismatch detection circuit, in a minimal environment.
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.