If LLMs Have Human-Like Attributes, Then So Does Age of Empires II

A peer-reviewed paper challenges the assumption that large language models possess uniquely human-like attributes by demonstrating that simpler systems—including the video game Age of Empires II—can exhibit similarly complex behaviors when given sufficient computational substrate. The research argues that attributing anthropomorphic qualities to LLMs requires explicit measurement criteria rather than subjective interpretation, and proposes a methodology that assumes non-uniqueness to avoid circular reasoning.

This arXiv paper addresses a fundamental methodological problem in AI research: the tendency to ascribe human-like properties to language models without rigorous empirical grounding. By training a neural network on Age of Empires II and demonstrating that it exhibits behaviors superficially similar to those attributed to LLMs, the authors expose how substrate—the underlying system executing the computation—fundamentally shapes interpretation rather than reflecting inherent properties.

The research builds on decades of philosophy of mind and cognitive science debates about what constitutes genuine understanding, intentionality, or morality. However, it applies these critiques specifically to contemporary AI hype, where researchers and commentators frequently claim LLMs demonstrate reasoning, understanding, or even consciousness. The paper's core insight is that any sufficiently complex computational system can generate outputs that appear meaningful to observers, but this appearance depends on the observer's interpretive framework rather than the system's intrinsic nature.

For the AI industry and research community, this work has significant implications. It suggests that many published claims about LLM capabilities rest on unstable empirical foundations. Venture capital investments in AI agents, regulatory frameworks premised on AI risk models, and safety research based on assumed LLM properties all depend on these anthropomorphic attributions being meaningful and measurable. The paper's null-hypothesis approach—assuming non-uniqueness until proven otherwise—would require substantially more rigorous experimental design in AI research.

Looking forward, expect increased scrutiny of published claims about LLM cognition and emergence. The research suggests funding agencies and peer reviewers should demand explicit measurement criteria and substrate-independent validation before accepting claims about novel AI properties.

• →Simple systems like Age of Empires II can exhibit behaviors superficially identical to those attributed to LLMs, undermining claims of unique human-like properties
• →Perceived anthropomorphic attributes in AI depend on substrate and observer interpretation rather than reflecting inherent system properties
• →Current AI research lacks explicit, measurable criteria for distinguishing genuine capabilities from patterns that appear meaningful to humans
• →The paper proposes assuming AI non-uniqueness as a null hypothesis until rigorous evidence demonstrates otherwise
• →Substrate-independent validation is necessary to make empirically grounded claims about any computational system's cognitive-like properties