The concept sounds simple enough: fans vote to choose which characters join AIdeal. In practice, building a voting system that's fair, engaging, manipulation-resistant, and capable of handling global participation at scale took us about six months of design and iteration before we were confident enough to open the first election.
This post is an honest account of what that design process looked like — including the problems we didn't anticipate and the rules we had to change after learning from the first round.
Phase One: Nominations and Community Development
AIdeal's formation process doesn't begin with a vote. It begins with a nomination phase, which we run for approximately four weeks before a formal election opens. During nominations, fans propose candidate concepts: a character name, a visual aesthetic direction, a vocal archetype, a rough personality sketch. These nominations are submitted through our community platform and are visible to all registered fans.
What happens during the nomination phase is more interesting than just collecting ideas. Community threads form around promising nominations. Fans refine concepts collaboratively — adding detail to visual direction, debating whether a proposed character concept overlaps too much with an existing confirmed member, arguing about whether a particular aesthetic niche is underserved in the unit. By the time the formal nomination window closes and we move toward a candidate shortlist, the most viable candidates have been substantially developed by the community, not just the person who originally submitted them.
This community co-development phase is one of the things we're most proud of in the process. It means fans arrive at the formal voting period already invested in the candidates, already knowledgeable about the reasoning behind each concept. The vote becomes a conclusion to a conversation, not a cold poll.
How We Handle Vote Eligibility
Open voting without identity verification is a solved problem in the sense that everyone knows how it goes: bots, duplicate accounts, and coordinated manipulation campaigns make the results meaningless. We decided early that we needed verified accounts to cast votes, and that each verified account could cast a limited number of votes per election.
Our current eligibility model requires email verification and account age — accounts created in the final 72 hours before an election opens cannot participate in that election. This is a blunt instrument, but it creates a meaningful barrier against the most common manipulation tactic, which is bulk account creation immediately before a vote.
We also weight votes by participation history. An account that has been active in the community for several months, has participated in discussions, and has voted in previous elections casts a slightly higher-weighted vote than a new account casting its first ballot. The weight differential is not large — we don't want a small number of highly active long-term fans to dominate elections — but it does mean that systematic attempts to flood the vote with new accounts face diminishing returns.
We're not claiming this system is undefeatable. Someone sufficiently motivated and resourced could probably still manipulate an election. What we're saying is that the cost of doing so is high enough relative to the benefit that it hasn't been worth attempting, and that we've built enough transparency into the results publication that the community would likely notice if something had gone wrong.
The Three-Week Election Window
The formal voting window for member elections runs three weeks. This length was chosen after thinking carefully about several competing constraints.
Too short, and participation skews toward fans in time zones and schedules that happen to align with the open window. A 72-hour flash vote rewards fans who are online during those 72 hours and shuts out everyone else. Too long, and the momentum and drama of the vote dissipates — there's no urgency to participate, and the community conversation becomes repetitive rather than escalating.
Three weeks gives enough time for word to spread, for fans who discover the election late to still participate meaningfully, and for community advocacy to build without the whole thing becoming a grinding endurance test. We publish intermediate vote tallies twice weekly during active elections, which creates natural moments of drama without giving away too much information too early.
The Problem We Didn't Anticipate: Regional Coordination
In our first election, we encountered something we hadn't specifically designed for: highly organized regional fan communities who coordinated votes with an efficiency that made their preferred candidates' performance in the tallies quite different from what community discussion threads would have suggested.
This wasn't cheating. Coordinated fan voting is a legitimate and time-honored practice in K-pop fandom — fan communities running organized voting drives, setting timers to remind members to vote simultaneously, streaming music to boost chart positions. These are cultural practices we had intellectually anticipated but hadn't fully priced into our model of how the election would run.
The result was that candidates with deep resonance in one regional community had a structural advantage over candidates with broader but shallower international interest. Whether that's a problem depends on your philosophy. We decided it was something worth addressing — not by suppressing regional coordination, but by adjusting how we present intermediate results to make the geographic distribution of support more visible. Fans can see where a candidate's support is concentrated, which adds useful information to the community debate without changing the fundamental math.
After the Vote: Transparency in Results
When an election closes, we publish a post-election report within 48 hours. It includes total vote counts by candidate, geographic distribution of participation, turnout compared to previous elections, and a brief summary of any anomalies we investigated (account flags, irregular patterns) and how they were handled.
This level of results transparency is unusual in entertainment and, frankly, in most voting contexts. It makes us accountable in ways that are occasionally uncomfortable. But it's the price of running a participatory system that fans trust. The community needs to be able to see that the results are genuine, that the process was followed, and that any edge cases were handled consistently with the published rules.
After three completed elections and three confirmed AIdeal members — Airi Kazoku, NEON, and Hoshi — the process has stabilized into something we're genuinely confident in. The fourth member slot is still open, and the next election will use an updated set of rules reflecting everything we've learned. The specifics of those updates will be published to the community before the nomination phase opens.
Building a fan-participatory system well is harder than it looks from the outside. But we haven't found a reason to believe the alternative — going back to agency-controlled selection — would produce a community with anywhere near this level of genuine investment in AIdeal's formation.
