OpenHCI
OpenHCI is the largest student-organized Human-Computer Interaction workshop in Taiwan. It brings together people from engineering and design backgrounds for a six-day intensive: a series of lectures on HCI topics, followed by cross-disciplinary team projects where participants go from ideation to working interactive prototypes. Our team participated in the 14th edition in 2024, where ChronoClones received the Best Technical Award.
Clonemator: The Starting Point
ChronoClones began with a paper. Clonemator presents a VR system that lets users create clones of their own avatar and configure them spatially and temporally to accomplish complex tasks. Clones can freeze at a body pose (static mode), mirror the user's movements in real time (synchronous mode), or replay a previously recorded sequence of actions (replayed mode). Combined with spatial operations like spawning, switching control, grouping, mirroring, and scaling, these temporal modes let users decompose multi-person problems into sequences of single-person demonstrations.
The key insight of Clonemator is treating the user's own body as a general-purpose automator. As the authors put it, "the user's avatar could be the generic automator waiting to be dispatched." It is essentially an open-ended sandbox tool: there are no predetermined goals, no win conditions, and no constraints on how or when clones are created. The user has full freedom to compose whatever spatiotemporal arrangements they can imagine.
This openness is exactly what we wanted to constrain.
ChronoClones: Turning a Tool into a Game
Our design question was straightforward: what happens when you take Clonemator's expressive clone system and add rules?
In ChronoClones, the player faces multi-person puzzles alone. The core mechanic is a time loop: each round, the player's actions are recorded. When the round ends, a clone is spawned that replays those exact actions in the next round. Round after round, the player accumulates clones, effectively building an entire team out of nothing but their own past decisions. If a puzzle requires four people to press four buttons simultaneously, the player must solve it across four rounds, each time coordinating with the ghosts of their previous selves.
The design decisions that separate ChronoClones from Clonemator are all subtractive. We removed free-form clone spawning: clones only appear at the start of each new loop. We removed the ability to choose interaction modes: all clones are in replayed mode, always. We removed real-time control switching: you cannot jump into a clone's body mid-round. Every degree of freedom we took away was a deliberate constraint that turned an open sandbox into a goal-directed game.
When we were building this, we internally described what we were doing as "gamification through constraints", making an open system game-like not by adding points or badges, but by restricting what the user can do. It was an intuitive design move at the time. We only looked into the formal literature afterward.
It turns out the idea is well-established in game design, though not under that exact name. Deterding et al.'s widely cited gamification taxonomy (2011) identifies game design patterns and mechanics (time constraints, limited resources, turns) as a core layer of what makes something feel like a game. More directly relevant is the concept of structured sandbox design: the best sandbox games succeed not by removing rules, but by carefully choosing which rules create the most expressive space. The design question is never "how do we remove constraints?" but rather "which constraints generate the most interesting problems?"
This framing maps precisely onto what we did with ChronoClones. Clonemator's three temporal modes, four spawning methods, and open-ended spatial manipulation give users enormous freedom, but also enormous ambiguity. By collapsing that design space down to a single temporal mode (replay), a single spawning trigger (loop boundary), and a fixed goal (solve the puzzle), we created a system where every player decision is legible, consequential, and composable with their past decisions. The constraints didn't limit creativity; they gave it structure.
Beyond the Game: Multi-Agent Coordination Data
After settling on the game design, we started thinking about what players are actually doing when they play ChronoClones.
Each round, the player is essentially programming one agent in a multi-agent system. They observe the current state (including the behavior of all existing clones), form a plan for how their new agent should coordinate with the others, and execute that plan through embodied action. Over multiple rounds, they iteratively construct a complete multi-agent strategy, one agent at a time.
This is remarkably close to what swarm robotics researchers need (this connection was suggested to us by Prof. Lung-Pan Cheng, who is also the advisor behind the Clonemator paper): diverse, human-generated coordination strategies for groups of agents operating under constraints. Clonemator's own discussion section mentions multi-robot control as a future direction, the idea that clone configurations could map onto physical robot swarms. ChronoClones, with its structured constraints, might actually be better suited for this purpose than the original open-ended system. The time-loop structure forces players to produce strategies that are temporally decomposed, fully reproducible, and naturally indexed by round number. Every successful playthrough is, in effect, a complete multi-agent coordination plan, generated by a human, ready to be analyzed.
We didn't build this into a formal pipeline, and it remained a discussion-stage idea. But the realization that a game designed for fun could double as a behavioral data collection platform is what led to the next part of the story.
Speculative Design
During OpenHCI, one of the lectures introduced us to speculative design, a practice rooted in the work of Anthony Dunne and Fiona Raby, where designers create artifacts not to solve problems but to ask questions. Speculative design doesn't predict the future; it constructs plausible alternative scenarios to provoke reflection on the present. The question it asks is always some version of: "if this trajectory continues, what world do we end up in, and do we want to live there?"
That framing stuck with me. We had just been discussing how ChronoClones could serve as a data collection platform. So I pushed the idea to its extreme and wrote a short piece of speculative fiction. (This wasn't discussed much during the workshop, it's something I had fun thinking about on my own after the event, so don't take it too seriously, LOL.)
The day they came for me, I was debugging a physics engine for a mobile game about stacking pancakes. A black sedan had been parked outside my apartment since morning. By noon, two men in suits were sitting in my living room, asking me to come with them.
They brought me to a government briefing room. They played the audio first. Signal-9963, converted from raw electromagnetic data into something human ears could process. It wasn't language. It wasn't rhythm. It was a low, continuous drone, just barely within the range of hearing, like a hum trapped inside the walls of a room you can't leave. The briefing lasted three hours. I don't remember most of what was said. I remember the sound.
Something was moving toward Earth. They had enough observational data to model its physical capabilities, its movement patterns, its apparent methods of attack. Enough to simulate. But knowing what something can do is not the same as knowing how to stop it. Every simulation the military ran ended the same way. It was faster, more numerous, and more coordinated than anything we could field. No strategist, no war-gaming AI, no think tank had produced a viable survival plan. The question wasn't how to organize a defense. The question was whether a defense was even possible.
That's why they wanted a game.
Someone on the defense council had read a paper I'd co-authored on multi-agent coordination in virtual environments. They didn't want me to fight this thing. They wanted me to build a VR game. One level. One scenario. The full model of what happens when it arrives. And then they wanted to give it to everyone.
The reasoning was desperate, which made it honest. Eight billion people on this planet, and some of them think in ways no institution can replicate. What if one of them could see something nobody else had seen? What if one person, given the right tools, could find the crack in the wall?
It took four months to build. ChronoClones. You play alone, but through a time-loop mechanism, each round gets recorded as a clone. A ghost that replays your exact actions in the next round. Round after round, you accumulate clones, building an entire team from nothing but your own decisions. If you clear the level, you've single-handedly devised a complete strategy for surviving what's coming. A proof of concept that humans can live through this.
The game opens with silence. Then, faintly, the drone fades in. The same fourteen-second audio clip, looped and layered, growing louder as the threat approaches. Every player who has ever put on the headset knows that sound.
When the game was ready, the governments went public with everything at once. The threat, the timeline, the audio, the game. Fourteen seconds of that sound was enough. You didn't need to understand astrophysics to know something was wrong. You just had to listen. People started calling it the Hum. And alongside the fear, they were given something to do about it. Play this, and you might be the one who figures out how we survive.
I have never experienced anything like the months that followed. The entire world was playing. Teenagers in Manila. Retirees in Oslo. Soldiers in Fort Bragg. Forums, livestreams, universities offering course credit for gameplay analysis. People quit their jobs to play full-time. Nobody questioned whether it was worth it. If nobody clears the level, we have no plan. If even one person clears it, we have a chance.
I worked twenty-hour days refining the simulation, tuning parameters, making sure the physics matched the best available intelligence. Every variable I adjusted, every collision model I calibrated, felt like it mattered in the most literal sense possible. The survival of the species ran through my code.
Nobody cleared the level.
Months passed. The best players in the world threw themselves at it, round after round, hundreds of clones deep. They found strategies that lasted longer, formations that held for a few extra minutes, resource distributions that delayed the inevitable. But nobody found the answer.
That was the thing that kept the world playing. The absence of a solution didn't mean one didn't exist. It just meant nobody had found it yet. Every failed attempt was another dead end eliminated, another step closer to the one path that might work. The desperation never faded. It couldn't. The Hum was still coming.
Then the first delay was announced.
New analysis suggested a later arrival window. Six more months, maybe eight. The world exhaled. Not relief exactly, but something close to it. More time to try. More time to find the answer.
Six months later, another announcement. Communication had been established, or something resembling it. Negotiations were underway. Each update bought another few months, another year. The public received each delay with gratitude. More time meant more chances that someone, somewhere, would finally crack it.
And because nobody had cleared the level, the fear never went away. The problem was still unsolved. The game was still the most important thing any person could be doing with their time.
The negotiations became the backdrop of daily life. News anchors reported on them the way they used to report on stock markets. "Talks resumed today in what officials describe as a cautiously optimistic development." Or, weeks later: "Negotiation channels have gone silent. Officials urge continued vigilance." When the talks were going badly, engagement spiked. When they were going well, people relaxed just enough to sleep. The rhythm of global anxiety was set to a dial that someone, somewhere, was turning.
This is when they started expanding the game.
The original ask made sense. If we can't beat them, maybe we can survive them. New levels began to appear. Evacuation coordination. Underground shelter logistics. Medical supply distribution during surface abandonment. Resource rationing across sealed habitats. If the main level was about defeating the Hum, these were about staying alive in a world where we couldn't. I built them gladly. It felt like a natural evolution. If we can't find the sword, at least build the shield.
But the levels kept coming. And slowly, so slowly I didn't notice for a long time, they began to drift.
One was essentially a municipal logistics problem. How to move supplies across a disrupted transportation network. Another was industrial workflow optimization. The Hum was still there in the background audio, the lore still referenced the threat, but the underlying problems had nothing to do with an alien species. I asked about it. I was told the scenarios were abstractions. That the mathematical structure of supply chain coordination translates directly to post-contact survival logistics. General-purpose readiness. It sounded reasonable. I let it go.
Six months later, I asked again about a scenario that was functionally a consumer behavior prediction model. I was told it was for modeling civilian response patterns during an evacuation. That sounded less reasonable. I let it go anyway.
You let things go when the alternative is too expensive. When your entire identity is built around the belief that what you're doing matters, you develop an incredible tolerance for things that don't quite add up. Every doubt has a half-life, and mine kept decaying before they could accumulate into anything critical.
Until the logo.
It was a Thursday. I was reviewing a design specification for Scenario 34, a new level involving coordinated resource distribution in a simulated urban grid. Page eight, bottom right corner, partially cropped by the margin: a logo. Sleek, sans-serif, unmistakable. One of the largest commercial AI companies on the planet. Not a government agency. Not a defense contractor. A company that sells predictive analytics to retail chains and advertising networks.
Their logo had no business being on that document.
My cursor stopped moving. I sat there for what felt like a very long time, listening to the hum of my workstation fan. For a moment I couldn't tell if the sound was the fan or the memory of the signal or something else entirely.
That night, I started looking. I had access to systems most people didn't. Not because I had security clearance, but because I'd built the infrastructure. I knew where the data pipelines were. I knew how the telemetry was routed. I just had never looked at where it went.
Every action every player had ever taken in ChronoClones was being ingested by a central processing node and distributed outward. Dozens of endpoints. I recognized some: military contractors, defense research labs. But there were others I had never seen. Tech conglomerates. Logistics firms. AI training facilities. Private research institutes with names that returned nothing on public searches.
My game, the thing I had poured my life into, the thing millions of people played because they believed it would save humanity, had become the largest behavioral data harvesting operation in history. And the people feeding it didn't know. They thought they were saving the world.
I leaned back and stared at the ceiling and tried to feel something specific. What I felt was not anger. Anger would have been easier. What I felt was the quiet, total collapse of a load-bearing wall inside me. The wall that separated "what I do" from "what is being done with what I do." They had been the same thing all along. I just hadn't looked.
The worst part is that I understand the logic. I can see the argument they would make if confronted. The data from ChronoClones has accelerated progress in fields we didn't know were stagnant. AI systems trained on player coordination patterns have improved emergency response times in twelve countries. Traffic fatalities have dropped. Supply chain efficiency has increased measurably. The utilitarian math probably works out. Probably.
But nobody asked.
Every one of those players consented to one thing. Contributing their strategic thinking to the survival of the human species. Not to ad targeting. Not to military logistics for conflicts that have nothing to do with the Hum. Not to training proprietary AI systems that generate revenue for companies they've never heard of. They signed up for one sacred purpose, and that purpose was quietly hollowed out and replaced with something else while the shell was kept intact.
And the negotiations. The ones that have been "ongoing" for years now. The ones that conveniently stall whenever engagement metrics dip and resume whenever there's risk of unrest. I don't think they're real. I don't think they've ever been real. I think they figured out a long time ago that the Hum was never hostile, or that it was never coming for us at all. And instead of telling the world, they kept the fear alive, because the machine it powers is too valuable to turn off.
I can't prove all of this. I can prove enough.
It's 2:47 AM. The admin interface is open in front of me. There's a counter in the top right corner of the screen: 4,237,891 players currently online. Four million people, right now, at this moment, inside the thing I built, believing that every move they make matters in the way they were told it matters.
My hand is on the mouse. The cursor is hovering over the button that sends out yet another update. I've been sitting like this for eleven minutes.
I built this. The time-loop mechanic, the clone synchronization, the physics engine that makes every round feel consequential. That's mine. The best work I've ever done. And now it's something else. Or maybe it was always going to become something else. Maybe every system built for a sacred purpose eventually gets optimized for a profitable one. Maybe the question was never whether they'd hollow it out, but when.
The office is silent. Not completely. There's the fan. The faint electrical buzz of the monitors. And underneath it, so faint I can't tell if it's real or if I'm imagining it, a low, steady hum.
The number on the screen ticks up. 4,237,903.
Twelve more people just logged in to save the world.
A Note on HCI
One thing I've taken away from the OpenHCI experience is just how expansive the HCI space is. A single project can start with a VR interaction technique from an academic paper, pass through game design theory, touch on swarm robotics, and end up as a piece of speculative fiction about surveillance and consent, and none of those transitions feel forced (at least to my amateur mind). HCI has room for all of it. That breadth is something I find genuinely exciting. It's always both fun and important to keep "human" and "society" in mind when developing technology, and getting familiar with HCI is a worthwhile experience whether you're a researcher in the field or not.
I'll be serving as one of the general coordinators for OpenHCI 2026. If you're a student in Taiwan interested in the intersection of technology and design, come join us!