Audience intent as an input

Problem

In a club, the dancefloor talks to the DJ all the time, without saying a word. People lean in, drift to the bar, put their hands up, or go very still in a way that means “don't you dare change this”. A good DJ reads it and adjusts. On a stream, that channel is mostly gone. There's a chat, which is a different conversation, usually about something else.

At the same time, the live system already accepts intent as an input. It takes direction from me in words, from a controller, from what I play. The question is whether listeners who aren't in the room can be given a small, honest share of that.

Hypothesis

If listeners can vote on a few simple directions (more energy or less, more tension or release), and the aggregated vote enters the system as one input among the others, the set will respond to the audience in a way people can feel, without being steered by it.

Two things have to be true for this to work. The vote has to be aggregated and smoothed, so that it describes the mood of the room over the last minute or two rather than whoever clicked last. And it has to stay one input among several, weighted against my intent and the current musical state, with my decision on top. A set run by majority vote would end up at the most typical version of itself, which is the pull I spend the rest of the time working against.

Setup

Planned, not built.

• On the SYSTEM ONLINE view, during a live session, listeners get a small set of controls: energy up or down, tension up or release.
• The server aggregates the votes over a rolling window and turns them into a single intent value per dimension, with a limit on how fast it can change.
• That value goes to the live system the same way the rest of the state does, through the Max for Live device, and the agents treat it as a suggestion with a fixed, modest weight.
• I see the aggregated intent next to my own. I can follow it, ignore it, or switch it off for a stretch of the set.
• Visitors see that their vote was counted, and they see the aggregated direction. They don't see a promise that it will happen.

Result

Nothing yet, because nothing has run. Questions to answer on the first sessions:

• Do listeners use the controls at all once the novelty wears off? TBD.
• Does the aggregated vote carry a real signal, or does it hover around the middle? TBD.
• How often does it agree with where I was going anyway? TBD, and interesting either way.
• Does it change how people listen, even when the set doesn't follow it? TBD.

What changed

So far only the framing. My first instinct was to call it “the audience controls the set”, which sounds good on a poster and isn't true. What's being tested is smaller: the audience as one more voice in a system that already has several, with a human who answers for the result. If that turns out to be boring, that's a result too.