Conversable Complexity: Agentic LLM Collectives as Interpretable Substrates
Elias Najarro, Ane Espeseth, Eleni Nisioti, Sebastian Risi, Stefano Nichele
Why It Matters
What makes this one worth your time
Understanding emergent behaviors in LLM collectives can enhance AI interpretability and inform future research in Artificial Life.
Agentic LLM collectives could redefine interpretability in AI by enabling emergent behaviors through interaction.
Summary
The paper discusses the potential of collectives of agentic large language models (LLMs) as a computational substrate for Artificial Life research, emphasizing their emergent dynamics and interpretability through natural language communication.
Key contributions
- Proposes a framework for interpreting emergent behaviors in LLM collectives.
- Surveys existing examples of agentic LLM collectives and their applications.
- Extends the notion of interpretability in language-model research to include collectives.
Notable insights
- The interaction among LLMs can lead to emergent properties that are not observable in isolated models.
- Natural language communication allows for direct interrogation of the collective behavior of LLMs.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2607.01047v1 Announce Type: new Abstract: Complexity and interpretability rarely coincide: systems rich enough for complex behaviours to emerge are usually too opaque to question, while transparent ones are too simple for anything complex to emerge. A single large language model (LLM) is a static artefact, hardly exhibiting any of the emergent properties we associate with life. This changes through interaction: populations of LLMs display emergent dynamics absent from isolated models. Furthermore, LLMs can be endowed with persistent memory, tools and shared skills, and the capacity to initiate actions unprompted, i.e., turning LLMs agentic. In this paper, we argue that such collectives of agents can serve as a computational substrate for Artificial Life (ALife) research. Critically, since the agents communicate in natural language, their collective behaviour can be directly interrogated by examining textual traces and asking the agents themselves. We outline the notion of interpretability in language-model research and extend it for collectives of agents. Lastly, we survey recent examples of agentic LLM collectives that already instantiate the idea of agentic substrates, from controlled experiments to deployments in the wild.