> **Status:** Raw ideation from a single conversation. Nothing here is decided, prioritized, or closer to a draft than anything else. These are loose threads to potentially return to. See also: [[Ghost Writing]] (in Writing/substack/) --- ## The Dissonance Frame (discussed, not adopted) Possible phenomenological grounding: the dissonance between knowing LLMs aren't persons and behaving as if they are. Examples discussed: - Getting angry at an LLM and telling it so - Trying to get the last word in an argument with [[ChatGPT]] - These behaviors persist despite knowing there's no one to win against Possible explanation: natural language as interface. When you write/speak English, you activate patterns wired for person-to-person communication. Those patterns don't switch off just because you know there's no person on the other end. Nuanced input → nuanced output. The medium carries interpersonal weight. Comparison to optical illusions floated: you KNOW the lines are equal, you still SEE them as different. You KNOW ChatGPT isn't a person, you still RESPOND as if it is. ## Bajohr (2023) Summary [[Hannes Bajohr]], "On Artificial and Post-Artificial Texts" - three phases of expectation: 1. Standard expectation (text = human author) 2. Transitional doubt (is this AI?) 3. Post-artificial (origin irrelevant) Key concept: "banal deception" - knowingly treating machines as if conscious to enable functionality. Discussed but not decided: whether the dissonance frame is stronger than Bajohr's "banal deception" (involuntary vs. chosen). The "unmarked zone" expansion - more text where authorship isn't expected. ## Double Dissolution (name disliked, idea maybe useful) The idea: not just FROM (author missing) but also TO (audience uncertain). LLM text often isn't addressed to anyone in particular. No name settled on. Options floated: "neither from nor for", "unmoored language", "phantom voices" ## Things explicitly rejected or cooled on - The "secretion" language - The 5-part structure from Ghost Writing.md - The framing of text as "weightless" ## Things NOT decided - Whether dissonance is the right frame - Whether Bajohr belongs in this piece - Whether "double dissolution" is useful - What the structure should be - What the essay is actually arguing - Whether this is the first piece or not ## Raw material to preserve "Have you ever tried to get the last word in an argument with an LLM?" - possible opening hook Natural language at as high a level as you can speak/write it is the interface - this is different from Siri/Google Home which are blunt instruments The responsiveness is part of the natural language point, not separate from it