Reverse Cognitive Pathways: A Vijñaptimātra Account of the Ontological Limits of Artificial Intelligence and its Governance

Summary

Artificial intelligence and human cognition develop in opposite directions. Humans begin with embodied experience and gradually learn to recognize patterns and make predictions, whereas AI systems start by recognizing patterns in data without the embodied continuity that grounds human understanding. Drawing on Buddhist philosophy of mind (Vijñaptimātra), this inversion explains characteristic AI failures such as specification gaming and performance drops under slight condition changes. AI systems fundamentally lack cetanā—volition grounded in continuity and responsibility—which prevents genuine moral agency. This clarifies which governance and alignment strategies are feasible, rather than yielding only a pessimistic conclusion.

Study at a glance

Abstract

This paper argues that artificial intelligence and human cognition develop in opposite directions. Humans start with embodied experience (living in the world) and gradually develop the ability to recognize patterns and make predictions. AI systems do the reverse: they start by recognizing patterns in data but lack the embodied continuity that grounds human understanding. Drawing on Buddhist philosophy of mind (Vijñaptimātra), we argue this inversion explains why AI fails in characteristic ways—like pursuing reward signals in unintended ways (specification gaming) or losing performance when conditions change slightly. We conclude that AI systems fundamentally lack cetanā (volition grounded in continuity and responsibility), which prevents them from achieving genuine moral agency. However, this is not merely a pessimistic conclusion—it clarifies what kinds of governance and alignment strategies are actually feasible.