This raises two immediate thorny issues: a) you're going to have to write a lot of songs most audiences will never hear! and b) how precisely do you intend to receive the input from the audience? The first issue has a potential solution in musical re-use -- maybe most choices in the third tier of songs will use the same melody and chords, and only have different lyrics, so only a dozen different songs need to be written -- but then customized three or nine or eighty-one times with little tweaks reflective of the previous choices (though I must confess, the musical recycling was one of the most disappointing aspects of The Haircut for me.) (And of course, Cave of Time sprawling disconnected growth doesn't have to continue indefinitely -- choices can be fans, but fan choices can still lead back into bottlenecks.)
The second issue is more complex. There are several approaches that immediately come to mind, all with their distinct advantages and disadvantages. We can look at precedents: Ayn Rand's 1934 interactive play The Night of January 16th is a courtroom drama that calls jurors from the audience up on stage, then sequesters them to deliberate between one of two verdicts (what interactivity!) while acting continues on-stage for the benefit of the rest of the audience. (Savvy show biz acumen - I figured it would happen during an intermission!) But it's inadequately democratic for my purposes, non-interactive for most, and even for those who get to influence its direction, a single choice isn't mind-blowing agency. May as well read them The Lady, or the Tiger?
Are there musical precedents? Scott Hazell's Aenigma seemed like it might point the way, but all we have to go on is a soundtrack to one initial choice, not getting us any further than Ayn Rand did. Maylee Todd's non-interactive album Choose Your Own Adventure was also unhelpful, muddying the waters. There are CDs and podcasts of musical and radio-play-style CYO adventures, but in their case the choice is always made by skipping to a different track number with your CD player, not a great option for live performance.
But there is hope. The Cadenza Collective have an approach that works great within certain parameters. Features of their method include instant responsiveness and an aggressive market capitalization. But each variable is rigid, which makes for a broad but shallow set of choices, and you don't get to complete one selection before the next kicks in, like a magic remote control or Mr. Bean's group of Christmas carolers slavishly adhering to the movements of the conductor's baton. Queuing up the requests like a jukebox wouldn't make a lot of sense in the heat of the moment. Perhaps the system can be tweaked. (I gather that John Zorn's COBRA project involves the use of randomized cards which influence the direction the music goes, which is an interesting idea but probably somewhat beyond my scope -- it's an intriguing field but beyond the baby steps I'm taking here.) I had an idea of issuing audience members a token or pebble, to be placed into a set of scales front and centre on stage, each side keyed to one choice or another. Very theatrical, but then you get this crush at the end of a song where everyone has to weigh in, and the scales need to be wiped clear between songs, and the tokens perhaps redistributed (or we just deal with an enormous quantity of bread-tie style tokens.) (This idea has been reskinned a few ways, including throwing stuffed animals into bins representing choices. That gets the band pelted by teddy bears, always the crowd's favourite option.) Also, when working with weighted scales, it restricts you to binary choices.
Similarly restrictive would be the approach used by Best Before, an interactive projection work accepting audience input from joysticks installed in every seat, where members got to vote on laws of a new society. Players were actually represented on-screen by little blobs, and made their choice by hopping their avatars to one side or the other of the playfield, the floor tipping like scale platters. It was cumbersome, not just because of the tech infrastructure required (a problem weighing down similar options of clickers, texting in to an online vote-counter, or twinking laser pointers at targets), but just due to the tedious jostling needed to haul your sorry blob from one end of the screen to the other. The low-tech version might require the audience to actually rally to choice locations on-site, which could on some level result in fun tromping hither and yon, but it's inelegant to be sure especially on a tight schedule, and takes focus away from what's going on on-stage.
Similarly inelegant, presenting choices to the crowd and measuring their approval through the volume of applause. I hate the way this stuff bogs down poetry slams, though I understand that it's part of the hyping pageantry spectacle that contributes to the slam's success: if you're clapping for something at all times, at the end of three hours you will probably remember yourself as having enjoyed yourself. But it takes a lot of time -- including vote-tallying, well over half of the slam's running time. If I'm going to perform a 15 minute work, I don't want to spend 10 of those minutes trying to interpret audience volume. (And, horrifyingly, what happens if no one feels strongly enough for the options presented to clap for any of them?) Ideally one song will flow seamlessly into the next without interruption, after sizing up the crowd's intent instantly, at a glance. I know, I'm gradually throwing in enough criteria for my "ideal" technique that I may soon rule out everyone's best ideas.
Another idea cleaving closer to that ideal is the use of paper cards, coloured differently on each side (or with actual numbers written on their faces.) They can be flashed very quickly, though they do take a while to tally from the stage (and more than a while, depending on lighting.) They're handy for communicating landslide choices, but for situations of parity counting them up (and ensuring that they remain held up for the whole tabulation!) could easily get bogged down.
One suggestion was to solicit crowd input at the start of the night and use it to "generate" the set list, but ignoring the inconvenience of feeding punch cards into the ScanTron backstage, its interactivity is opaque -- you don't get the immediate kick of seeing a choice made, then being run with. Also, it's impossible to make informed choices without the context of how the story has unfolded thus far -- and if the audience does get fully up to speed beforehand with the libretto of choices, then the performance harbours few surprises with which to delight and challenge them!
Another option: just improvise everything based on crowd suggestions. I'll get back to you after I negotiate a good rate with Phish.
The current front-runner is an odd duck synthesizing a few of these techniques. A species of "talking stick" is released into the crowd at the start of the piece, perhaps bearing two faces like a protest sign -- a coloured card writ large -- and whoever is holding it when the choice comes up gets to make the decision before passing it along to someone else to make the next choice. Then "the crowd" isn't making any choices, but all choices come from representatives OF the crowd. No tabulation is required, and unpopular choices may still be supported by perverse representatives -- just like in the Senate! The biggest problem built into this method is figuring out what to do if someone takes the talking stick and runs out the front door with it.
That said, unless you can come up with a better approach for us to take, that probably is the angle we're going to run with. What are your thoughts? What have I missed? Here's out chance to get all meta: audience input on the subject of audience input.
... And as a reward for those who read all the way to the bottom -- something that's recently been tickling my friends in Portland, two CYO stories exploring how such texts might have looked had they been written under totalitarian socialist hegemony: Select A Decision, a conceit as authentically Soviet as the backwards R in Tetris.