# Telos in syntax (moderate reading)
The norms of truth-directed philosophical practice are publicly encoded in discourse; mastery of that discourse can yield telos-shaped outputs without an inner truth-aim.
## The Claim
In a natural language like "English-for-Analytic-Philosophy," words are not just labels; they are tools for tracking reasons. To understand the inferential role of *Justification*, you must understand that it is used to filter out *Falsehood*. You cannot map the "syntax" of philosophical moves without mapping the normative direction (the telos) those moves are intended to take.
For a sufficiently saturated model: **Syntax = Functional Semantics**. The model doesn't need an external truth-sensor because it has a map of the success conditions of human arguments.
## The Moderate Reading
The inferential web encodes the telos in the sense that it publicly specifies what truth-aiming *looks like* in that practice; therefore a system can *simulate* truth-aiming extremely well by mastering that web, even if it doesn't "care."
This matches the "argument not author" instinct: in analytic philosophy we assess the logical object (the argument), not the inner phenomenology of the producer.
On this reading, "telos in syntax" doesn't mean the model is a truth-lover. It means:
> The practice is truth-directed, and the practice is linguistically articulated, so mastery of the practice's inferential structure can generate *truth-directed artefacts*.
## Why This Matters
The moderate reading is defensible without overclaim:
- It doesn't require showing the model has an inner truth-aim
- It focuses on outputs and their normative properties
- It aligns with functionalist accounts of meaning ([[Wilfrid Sellars]], [[Robert Brandom]])
This connects to [[Philosophy as internal verification domain]]—both arguments leverage the fact that philosophy's evaluative standards are largely internal and discursive, making it a domain where LLMs can do substantive work even without world-contact.
It also provides a response to [[Floridi's critique of LLM abduction]]: even granting that LLMs lack an internal truth-objective, the publicly encoded norms of philosophical practice can guide outputs toward truth-directed artefacts.
Source: [[Web Clippings/chat with ChatGPT about Floridi, LLM reasoning, Move 37|ChatGPT conversation on Floridi et al.]]