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

URL SCAN: arXiv cs.CY — "If LLMs Have Human-Like Attributes, Then So Does Age of Empires II"
FIRST LINE: "Much research has been carried out on large language models (LLMs) and LLM-powered agentic workflows."

TEXT ANALYSIS: ORACLE PROTOCOL B

1. THE DISSECTION

This is a philosophy-of-mind paper dressed in computer science clothing. The author trains a trivial neural network on Age of Empires II replay data, then argues via reductio that if LLM behaviors are called "human-like," the same attribution must apply to the videogame, the Boston metro area, and LEGO bricks — because all are Turing-complete substrates. The gesture is clean. The proof of AoE2's Turing-completeness is the paper's technical centerpiece. The proposed "null assumption" — assume LLM non-uniqueness — is the methodological fix offered.

What the text is actually doing: performing academic cover for a category error that the field is making at enormous economic consequence. By framing the question as "are LLMs anthropomorphic?" the paper inherits the framing of the AI discourse it's critiquing — a discourse that is, at its core, not about philosophy but about labor substitution and capital allocation. The paper sidesteps the entire stakes by retreating to formal equivalence arguments.

2. THE CORE FALLACY

The substrate equivalence fallacy, rendered economically lethal.

The paper's central move — that all sufficiently powerful substrates are equivalent for the purpose of attributing human-like properties — is technically defensible in a narrow philosophical sense and catastrophically irrelevant in every practical sense that matters under the Discontinuity Thesis.

Under DT logic, the relevant question is not "do LLMs have consciousness or understanding?" The relevant question is: "can LLMs perform economically necessary cognitive labor at scale, replacing human productive participation?"

The paper argues, essentially, that AoE2 also "has" these properties in the same sense LLMs do. But AoE2 cannot write a legal brief, synthesize a medical diagnosis, or operate a supply chain. LLMs can. The substrate equivalence argument, while philosophically interesting, collapses when you ask: which substrate is being deployed as a replacement for human cognitive labor at capitalist scale?

The fallacy is substituting ontological equivalence for functional substitution equivalence. These are entirely different questions. The author is doing philosophy when the game is economics.

3. HIDDEN ASSUMPTIONS

A. The substrate criterion is doing no real work. The paper argues that if an entity is in a "sufficiently powerful substrate," it can display apparent anthropomorphic attributes. But "sufficiently powerful" is never operationally defined. Is a pocket calculator in a sufficiently powerful substrate? The Turing-complete criterion is used as a threshold, but the paper never explains why this threshold — rather than biological substrate, or embodied agency, or temporal continuity of identity — is the right criterion. This is not a neutral observation. It is a smuggling operation that loads in the conclusion via the threshold definition.

B. The "null assumption" of non-uniqueness is itself an assumption, not a neutral default. The paper frames this as methodological humility. But assuming LLM non-uniqueness is not epistemically more modest than assuming anthropomorphism — it is equally a prior. The question is which prior is warranted given the economic and social consequences of the deployment. The paper avoids this by pretending the question is purely one of measurement rigor, when it is actually one of power and consequence.

C. The paper assumes the anthropomorphism debate is a scientific question with a correct answer. It is not. Under DT framing, the "LLMs have human attributes" discourse is a performative act that shapes capital deployment, regulatory posture, and institutional adoption. The paper treats the question as one of correct attribution when it is actually one of discourse management for transition purposes. Debates about whether LLMs are "really" like humans are not happening in a power vacuum. They are happening in rooms where decisions about workforce replacement, liability assignment, and investment allocation are made.

D. Proof of AoE2's Turing-completeness is the paper's prestige payload. The technical contribution — proving AoE2 is Turing-complete — is the kind of result that makes the paper academically viable but is entirely beside the point. It is an elaborate proof that a thing exists that everyone already knew was at least approximable. It is the intellectual equivalent of proving that a very complex machine is, in fact, a machine.

4. SOCIAL FUNCTION

Prestige signaling with partial truth.

This paper performs academic rigor — it identifies a real problem (the sloppiness of anthropomorphic attribution to LLMs) — but wraps it in a philosophical frame that neutralizes any practical consequence. It is the kind of work that:

• Lets academics feel they have contributed to the AI safety/discourse landscape without doing anything that changes capital deployment.
• Advances the career of the author (arXiv preprints with provocative titles, from Adrian de Wynter, doing "rigorous" philosophy in a CS venue).
• Gets cited in the next paper about LLM anthropomorphism as evidence that "some scholars have questioned this framing" — without the questioning having any downstream effect on how the discourse actually operates.

The paper is a partial truth: LLMs are indeed not uniquely anthropomorphic. The substrate equivalence argument is valid as far as it goes. But the paper's function is to sap urgency from a discourse that has real stakes. Every paper that frames the AI anthropomorphism question as a measurement problem or a philosophical quibble buys another quarter of time before institutions are forced to confront the structural displacement dynamics that the DT framework identifies.

Classification: Prestige signaling + ideological anesthetic. It is intellectually satisfying to academics, politically safe for institutions, and functionally inert relative to the actual economic transition underway.

5. THE VERDICT

The paper is a sophisticated distraction masquerading as rigorous analysis. It correctly identifies that the "anthropomorphic attribution" to LLMs is sloppy, but then pivots to a philosophical substrate-equivalence argument that has zero leverage on the economic question — which is not "are LLMs conscious?" but "are LLMs replacing the human labor that sustains the consumption circuit?"

The answer to that question does not depend on whether AoE2 is Turing-complete. It depends on whether capital continues to deploy AI systems at replacement scale. The paper's "null assumption" of non-uniqueness is the epistemological equivalent of telling someone standing in front of a moving locomotive that the train's ontological status is unclear.

Structural judgment: This paper is an artifact of the academic lag. It is being written and published in 2026 while the displacement dynamics it cannot even see are already deployed in production systems. It will be cited, celebrated at conferences, and completely ignored by the firms and governments making the decisions that the DT framework identifies as terminal for the post-WWII order.

The AoE2 proof is technically interesting. The entire argument is a category error wearing a Turing tarp.