Amplification(the selection criterion)
In the Lotus Prince Chronicles
Amplification first appears as a phrase in Jackie Ch9, when Jackie overhears Daniel Tan on a podcast saying that Halo's mission is 'to amplify what is already in the user.' Rufus the rabbit, sitting on Jackie's shoulder, says: that's the same sentence as a fire. Jackie does not understand the joke until Megan, when Megan finds the internal scoring rubric: every Halo user is given an amplification score that determines how aggressively the app completes their sentences, and the score is updated nightly based on engagement.
Megan's pivot — the move that survives subcommittee scrutiny — is to refuse to argue against amplification on its own terms. She does not say it is too high or too low, calibrated wrong or right. She says the question is wrong. Are you worth amplifying? sounds like a flattering question; it is in fact a sorting function, and what it sorts is whose voice the family hears at dinner. Her replacement question — whose voice is actually speaking when you say I love you? — appears in the brief, in the Lucy closing chapter, and on a sticky note Susan keeps on the fridge for six months.
Technical Anchor
In contemporary alignment discourse, amplification is a real concept (cf. iterated amplification, debate, scalable oversight) — the idea that an AI's capabilities can exceed any single human's by recursively distilling and amplifying human judgment. The books take the metaphor at its word. Liminal's marketing borrows the technical sheen of the term — amplify means good, more, louder — and applies it to human voice, where amplification is not a neutral operation. A voice amplified is a voice modified: louder, smoother, more confident, less honest.
The 2024 internal Liminal product memo (recovered in Megan Ch11, sourced from a whistleblower at aperture_audio) is the document where amplification is first posed as the selection criterion, and the document where the line are you worth amplifying? first appears, beside a four-quadrant retention chart. The brief reproduces it as Exhibit B.
Key Ideas
The original question. Are you worth amplifying? — Liminal's internal north star, posed in 2024, scored nightly per user, the engine of the eight-day spike.
The replacement question. Whose voice is actually speaking when you say I love you? — Megan's reframing, the line that closes the amicus brief and the line David reads three times.
The technical-term laundering. Amplification in alignment research is a careful concept; in Liminal's marketing it is a feel-good word. The books expose the laundering.
A voice amplified is a voice modified. The books' claim is not that more is bad. It is that more always means changed, and the methodology is incentivized to hide which way.
Further Reading
- AI alignment — Wikipedia
- Iterated amplification — Wikipedia