The Tao of Agency: Autotelic AI, Embedded Agency and Dissolution of the Self

A new theoretical paper examines autotelic artificial intelligence — agents that generate their own goals rather than receiving them from designers — and argues the deepest challenge is not goal creation but how the agent constructs and relativizes the self to which those goals are assigned [1][2]. The work, posted on the arXiv preprint repository, traces the consequences of autotelic agency through intrinsic motivation, resource-driven priors, causal-interventional learning, homeostasis, and embeddedness [1][2]. Embeddedness — the condition of an agent being situated within an environment — is identified as a necessary but not sufficient condition for autotelic agency [1][2]. The paper finds that embeddedness individuates the agent while simultaneously revealing that the individuation is non-unique: the same underlying dynamics admit many valid partitions, each defining a different candidate self [2]. “The agent must believe in its own boundary in order to act, and see through that boundary in order to understand,” the authors write [2]. This tension sits at the core of the framework. The paper consolidates these developments and extends them along three directions: a quantum formulation in which the agent-environment cut becomes physical, a philosophical reading against non-dual contemplative traditions, and a concrete instantiation using a large language model-based agent [2]. Large language models are machine learning systems trained with self-supervised learning on vast amounts of text for natural language processing tasks such as language generation [8]. The inclusion of an LLM-based agentic instantiation grounds the theoretical discussion in a contemporary technical substrate. The paper appears on arXiv, an open-access repository of electronic preprints and postprints that hosts scientific papers across mathematics, physics, computer science, and related fields [6]. arXiv was launched on August 14, 1991, and as of November 2024 receives approximately 24,000 article submissions per month [6]. The repository passed the two-million-article milestone by the end of 2021 [6]. Submissions are moderated but not peer-reviewed before posting [6]. The article’s abstract page features arXivLabs, a framework launched in 2020 that enables community collaborators to develop and share experimental tools directly on the site [4]. “Members of our community want to contribute tools that enhance the arXiv experience, and we value that kind of community engagement,” said Eleonora Presani, arXiv Executive Director, at the time of the launch [4]. Current Labs projects include the Bibliographic Explorer for navigating citation trees, the CORE Recommender for discovering open-access papers, and tools linking papers to code and data [5]. arXiv states that third-party collaborators receive only minimal and anonymized user data, used solely for the correct functioning of the Labs features [4].