From Teleology to Adequacy=A Spinozistic Layer for World-Model AI

Contemporary AI systems increasingly offer natural-language explanations of their behavior. These explanations are often teleological: they describe what the system “wants”, “tries”, or “is meant to do”. Teleological language can be useful as a form of narrative compression for humans, but it also invites anthropomorphism and miscalibrated trust. At the same time, recent work on world models and Joint Embedding Predictive Architectures (JEPA) emphasizes that autonomous agents need compact causal representations of the world, not just next-token prediction.