iv · visszhangtárs — echo companion

ismétlődéses · imitation — a partner made of sound

Two keys. Play something and stop; something answers. It is imitating you, but not exactly, and you may imitate it back. There is no score and no end — you leave when you leave. Every so often it will change one thing on purpose and then listen to what you do about it, which is the only thing on this page that is really being measured: when your partner hands back a slightly altered version of what you played, do you correct it, or do you adopt it?

The partner is not clever and is not pretending to be. It has four numbers — eagerness, fidelity, initiative, patience — drawn once from this session’s seed and then held fixed. It has no model of you, no memory beyond the last few phrases, and nothing to say about how you are doing. Everything that happens between you is made by the two of you out of timing, and that is the claim: a partner does not need to be intelligent to be a partner, it needs to be consistent, and to be listening.

what this strand isolates

Social stance. This is the strand that is least about a faculty and most about a person, and the source says so plainly: what is examined in the end is the player’s relationship to the play partner.

The design hinge is the phrase two-way imitation (kétirányú utánzás). Simon — Ralph Baer and Howard Morrison’s four buttons and four tones, 1978 — is one-way: the machine calls, you repeat, forever. Making it two-way changes the genre entirely. If the partner can also imitate you, then it can imitate you imperfectly, and the imperfection becomes a probe. That single binary — adopted or corrected — logged over a few hundred exchanges, is a musical conformity assay, and it needs no questionnaire and no language to run. It is the strand’s signature number and it is on the panel above as adoption.

The source’s other axis — aimless play against the accumulation of successes — is a stance, not a skill. Some players will treat the partner as a mirror, some as an opponent, some as company. All three are correct, and nothing here distinguishes between them by rewarding one.

Where this needs care. A synthetic play partner deployed to people who are lonely — in care settings, in dementia wards, with isolated children — is ethically loaded, and the loading does not disappear because the partner is made of sound rather than a face. The position this project takes is that these pieces are instruments for people to play together through, and are not offered as substitutes for human company. Where a solo version is used in a care context that framing has to be explicit in the deployment, not implied in a document.

the rest of the strand

1. I2 Hűtlen visszhang · The Unfaithful Echo
   Ten minutes, once. The partner begins as a perfect echo; over ten minutes its fidelity decays from 100% to roughly 40%, and what replaces the imitation is material of its own — a partner becoming a worse mirror, or a better improviser, depending entirely on what the player wanted from it. Nothing is explained; the player’s response is the whole content. Nothing about the piece is unfair, and the partner never punishes, yet in testing people reliably describe it in emotional terms. What a player says the machine was doing to them is more informative than any of the logged numbers. measures — attitude classification from behaviour alone: follow / correct / compete / abandon · time to abandonment · post-session anthropomorphism, from a one-question interview
2. I3 Bújócska · Hide and Seek
   Binaural, or a ring of six loudspeakers. The partner hides somewhere in the sound field and will only imitate you from where it is standing; to find it you must make sounds and listen to where they come back from. Then the roles swap, and you must produce sounds that are hard to place. Players discover on their own that a low sine is nearly impossible to localise while a broadband click gives you away instantly — the duplex theory of sound localisation, learned by hiding badly. It carries a real Hungarian children’s-game name and earns it, and it is the piece in which a blind player will routinely and visibly beat a sighted one. measures — absolute localisation error by bearing and elevation · front–back confusion rate · whether hiders learn to prefer low sinusoids, and in how many rounds

I2 is a ten-minute one-way trip that would be spoiled by a reload button, and I3 needs measured head-related transfer functions and a short individualisation routine, because in that game localisation accuracy is the measurement and a generic HRTF would be measuring the wrong head.