When Kaira Dumasia sits down to work on her university assignment, she is not writing code or debugging software, at least not in the way most people imagine. She is planning how to make strangers angry enough to click, comment, and argue. “I am really going to try to rage bait everybody,” she says, half amused, half serious. She is a computer engineering student, and her task is to use artificial intelligence bots to interfere in an election. But before anyone panics, there is a very important catch: the election is not real. The voters she hopes to manipulate are not real either. They are simulated citizens living inside a fictional world, designed by researchers at the University of New South Wales to give students a taste of how AI can be weaponized online. The assignment is part of a competition that pits teams against one another to win a simulated vote through social media manipulation. On the surface, it might sound like a game. But underneath, it is a serious educational experiment meant to teach young technologists about the dark arts of digital influence—by letting them practice those arts in a controlled, safe, and watchful environment. For Kaira, what started as an unusual piece of coursework has become a mirror held up to her own social media habits, and a lesson in how easily attention can be turned into power.
The game itself is cleverly designed to mimic the chaos of real online political campaigning. This year, the action is set in Kingstonia, a fictional island nation facing a major democratic decision: whether proof of vaccination against something called “dolphin flu” should be required to enter public spaces. It is a strange, slightly absurd premise, but that is intentional. The absurdity keeps the stakes low and the focus on the mechanics rather than real-world policy debates. Teams of students are assigned a side in the referendum and given the freedom to build AI-powered bots that can post, like, share, and reply on a made-up social media platform called Legit Social. The platform is populated by AI-powered citizens who read posts, interact with content, and gradually change their views over the course of the month-long competition. Points are awarded for a range of achievements: grabbing attention, changing voters’ minds, and, crucially, spotting rival teams’ bots. That last part is the twist. While everyone is busy trying to manipulate the simulated public, they are also being asked to act as defenders, hunting for signs that another team is doing the same. It is a dual role that mirrors the real world, where the same AI tools used to spread influence are also used to detect it. The competition is the brainchild of Hammond Pearce, a senior lecturer in computer science at UNSW, who sees it as a “Petri dish” for studying AI electoral interference. By letting students become attackers and defenders at the same time, he hopes they will leave with a deeper understanding of how manipulation works, not just in theory but in practice.
Kaira’s strategy for the competition is both simple and unsettlingly familiar. She has been watching how real political campaigns and influencers operate online, and she has noticed a pattern. “I see a lot of it is rage bait content, engagement bait content,” she explains. So she is doing the same thing: crafting posts designed to make the simulated citizens angry enough to comment, because comments signal engagement, and engagement earns points. It is a strategy born of real-world observation, and it works disturbingly well. The more controversial a post, the more likely people are to interact with it, even if they are interacting because they hate it. That is the dirty secret of algorithmic amplification: strong negative emotions are fuel. And Kaira is not just exploiting that dynamic in the game; she also sees it in her own life. She admits that she often falls for rage bait on her personal social media feeds. “I often find myself falling to rage bait and then I’m like, whoa, take a step back,” she says. That moment of self-awareness, the pause between the emotional reaction and the rational response, is exactly what the competition is trying to cultivate. It is one thing to learn about AI manipulation in a lecture, but another to feel its pull when you are the one pulling the strings. By turning her own feed into a source of inspiration, Kaira has discovered a strange truth: the best way to understand how people are manipulated online is to try to manipulate them yourself, and then recognize how easily you, too, would have fallen for the trap.
But the competition is not just about being a skilled manipulator. It is also about being a sharp-eyed guardian. Rival teams can report posts they suspect have been sent by other bots, and if their suspicions are correct, they earn points for their own team. That means every team must balance offense and defense, constantly asking themselves: Is that post too perfect? Is that reply too fast? Is that account behaving like a real person, or like a machine wearing a human mask? Dr. Pearce says this duality is essential. “We’re trying to expose that tech to students in a safe environment so they can learn both what the attackers are doing by doing it themselves and also what the defenders need to do in order to try and catch that kind of attack,” he says. The platform, Legit Social, was built specifically for this purpose, a closed universe where students can experiment with manipulation and detection without causing real-world harm. But the lessons are deeply relevant outside the game. AI-driven influence operations are no longer hypothetical threats. They are happening in elections around the world, often in ways that are difficult to see and even harder to stop. By giving students hands-on experience with the tools of both attack and defense, the competition hopes to produce a generation of engineers and computer scientists who think critically about the systems they build, and who understand that every feature, every algorithm, every design choice can be used to inform, persuade, or deceive.
The real-world context makes this exercise feel urgent. The fictional vote in Kingstonia is taking place in the lead-up to the United States midterm elections, and it comes shortly after Australia’s election watchdog warned about the role of artificial intelligence in amplifying existing threats to electoral integrity. Australian Electoral Commissioner Jeff Pope told a parliamentary committee that AI is increasingly involved in misinformation, influence operations, foreign interference, and even physical security risks through harassment and social cohesion damage. Those are heavy words, and they suggest that the playful, slightly absurd competition at UNSW is actually tackling one of the defining political challenges of our time. Researchers are also using the game to study how participants build AI bot networks and how they identify other bot activity. A preliminary academic paper about last year’s competition examined the techniques used by competitors and found that some teams turned to high-volume posting because the scoring system rewarded engagement. The result was a fascinating glimpse into the strategic thinking of young people who are not just consuming social media but actively engineering it. They learned that sometimes the easiest way to win is not to be clever or subtle, but to be loud, fast, and relentless. That finding echoes what observers have seen in real political bot networks, where armies of accounts flood platforms with repetitive messages in hopes of overwhelming the user’s attention and dominating the conversation.
For Kaira, the experience has been transformative in ways she did not expect. Last year, she was a moderator in the competition, helping to oversee the mayhem from the sidelines. This year, she is a competitor, and the shift has taught her to see the internet with new eyes. She now recognizes AI-generated, bot-like posts more easily, noticing the patterns that give machines away. But she also understands that recognizing manipulation does not make you immune to it. The same emotional triggers that make rage bait so effective are wired deep into human psychology. In a way, the competition has given her a kind of humility. She has seen how easy it is to exploit people’s anger and curiosity, and she knows she is not above it. “I often find myself falling to rage bait and then I’m like, whoa, take a step back.” It is a powerful lesson, not just for a computer engineering student, but for anyone who spends time online. AI may create the bots, write the posts, and scale the messages, but it is human nature that gives them power. The UNSW competition is designed to expose that uncomfortable truth to the people who will one day shape the digital world. It gives them a chance to make mistakes in a place where no one gets hurt, to learn the tricks of the trade before the trade becomes too dangerous, and to realize that the line between influencing an election and just trying to get a reaction is thinner than anyone would like to admit. In the end, the referendum on dolphin flu is not really about dolphins, or vaccines, or a fictional island. It is about the future of democracy, and the small handful of engineers who are learning, one simulated vote at a time, how easily it can be bent to someone else’s will.