On the Psychology of a Large Language Model

Diogo Luiz Alves de Araújo

No adxbj_v1, SocArXiv from Center for Open Science

Abstract: This article contends that the prevalent anthropomorphism in Large Language Model (LLM) alignment research constitutes a fundamental category error, rooted in psychological projection. By describing LLMs with human-centric terms like “deception” and “intent,” the field mischaracterizes the technology, leading to flawed threat models and misguided safety evaluations. I first deconstruct the LLM as a mathematical and statistical system, demonstrating how its convincing mimicry of cognition emerges from probabilistic pattern-matching, not genuine understanding. I then establish a philosophical threshold for moral agency by synthesizing Humean, Kantian, and phenomenological perspectives, arguing that agency requires affective sentiment, rational autonomy and subjective, temporal experience—all of which are absent in LLMs. Using a Jungian framework, I re-interpret studies on “deceptive” and “scheming” AI not as discoveries of emergent malice, but as manifestations of the projection of our own “Shadow” onto an opaque technological artifact. This misinterpretation leads to dangerous, quasi-mythological narratives of AI risk, exemplified by reports such as 'AI 2027'. As an alternative, I propose a grounded paradigm for alignment that shifts focus from human-like malice to non-human failure modes. This paper concludes not that LLMs are harmless, but that danger is misplaced. The risk arises when a non-rational text generator is connected to real-world tools and functions as an advisor to end-users and geopolitical leaders, a situation that demands conspicuous communication about the technology's scripted nature and inherent limitations.

Date: 2025-09-22
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://osf.io/download/68cedd254b6ff888b874e85e/

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:osf:socarx:adxbj_v1

DOI: 10.31219/osf.io/adxbj_v1

Access Statistics for this paper

More papers in SocArXiv from Center for Open Science
Bibliographic data for series maintained by OSF ().