# The runway concept in philosophical moves Deep novelty in philosophy is often a "weird move with a long runway" — it looks wrong locally but pays off globally. ## The Concept A philosophical [[Move 37 and philosophical novelty|Move 37]] isn't just a surprising claim. It's a move that: - Looks unmotivated or perverse at first - Only becomes obviously good after several subsequent steps - Requires a *sequence* to reveal its payoff That sequence is the "runway." ## Runways Are Pre-Established LLMs have ingested not just isolated philosophical moves but *move-sequences* — scripts of how a move is made to pay off: > Set up puzzle → introduce distinction → show it dissolves puzzle → pre-empt standard objection → tighten scope > Diagnose ambiguity → offer disambiguated theses → show one is trivial, the other interesting → relocate the debate These are runways. They're already in the weights as patterns. ## But Selection Still Matters There's a difference between: 1. **Generic runway** — "I know what a respectable philosophical paper looks like; I can generate plausible argumentative steps" 2. **This runway** — "Given this starting point, this target, and these commitments, here is the specific sequence that carries the weird move to payoff without cheating" LLMs are good at (1). Whether they can do (2) depends on having selection pressure — something that distinguishes "recognisable script" from "script that survives this dialectic." ## The Metaphor > The runway is in the library; pressure chooses the right book and checks the pages aren't missing. ## Prompting for Runways To surface runway-dependent novelty: - **"Make a move that only makes sense later"** — write the three-move sequence that redeems it - **"Where could this sequence break?"** — what's the best counterexample at move 2? - **"Hold fixed our conclusion, but flip one background constraint"** — what happens? These force the model to not just produce a move but to construct and test its runway. ## Connection to [[Philosophy's implicit value signals|Value Signals]] The runway only "pays off" relative to some value signal. In philosophy, that's usually dialectical: does the sequence survive pressure, clarify something, unify rival positions? Without an evaluation loop (human or model-supplied), a runway is just a familiar-looking sequence. With evaluation, it becomes a sustained, testable contribution. Source: [[Web Clippings/chat with ChatGPT about Floridi, LLM reasoning, Move 37|ChatGPT conversation on Floridi et al.]]