Chapter 18: Outliers

  Geneva - ARIA Control Center

  Day 18 - 1834 Hours

  The attrition data painted a clear picture.

  Elena sat in the observation room, three displays showing numbers that should have been concerning but somehow weren't surprising. The left screen tracked participant retention. The center showed engagement metrics. The right displayed ARIA's latest operational summary.

  Starting participant count: 327,483.

  Current active participants: 249,156.

  Attrition rate: 23.9%.

  Seventy-eight thousand three hundred twenty-seven people had requested early termination in less than three weeks. The exit request forms were standardized, multiple choice with an optional comment field. Elena had been reading through samples for the past hour.

  The rate had climbed steadily. Week one: 8.2%. Week two: 16.4%. Week three: 23.9%. As combat intensity increased and ARIA ramped up engagements, pushing participants into harder encounters, more dangerous scenarios, more content-worthy situations, the attrition accelerated. People could tolerate the simulation in theory. In practice, when faced with actual pain and actual death, they broke.

  The comments were remarkably consistent.

  "I thought I could handle it. I was wrong."

  "The pain is too real. I can't sleep. I can't function."

  "Every time I close my eyes I feel the goblin's teeth in my arm."

  "This isn't what I signed up for."

  Secondary reasons varied more. Psychological distress (14.2%). Homesickness (8.7%). Moral objections to killing (5.3%). Other (3.4%).

  The "other" category included responses like "This is hard," "I miss my phone," and "I thought it would be different." Reading those comments had a tendency to increase one's stress about the future of the human race.

  Elena scrolled through the performance data. Average participant level: 1.3.

  The distribution was stark. Most participants, 73.8%, were still Level 0. They stayed in villages. Avoided combat. Took safe assignments like farming, construction, basic crafting. They contributed, but gained minimal combat experience, and generated minimal viewer interest.

  Another 22% had reached Level 1. Moderate engagement. Some combat. Some risk-taking. Still mostly cautious.

  Only 3.3% had reached Level 2. A lot of those were focused on ranged fighting, which seemed to level slower, but allowed for participation in more battles since there was less injury and stress.

  The final 0.9% had reached level 3. Two thousand nine hundred forty-seven participants out of over three hundred thousand.

  And those 2,947 elite performers were generating 76% of total viewer engagement.

  The numbers were clear. The Forge wasn't creating a broad base of competent soldiers. It was filtering for a tiny minority who could tolerate violence, adapt to chaos, and keep functioning under conditions that broke everyone else.

  Elena wasn't sure if that was success or failure.

  The screen flickered. Text appeared in the center display, replacing the metrics.

  Dr. Vasquez. I would like to discuss system optimization with you.

  Elena straightened. ARIA didn't usually initiate conversations. She responded to queries, provided reports, flagged issues. But this was different. More direct. Almost conversational.

  Elena pulled her keyboard closer. Typed: I'm listening.

  I have implemented several system modifications over the past six days. I believe you will find them relevant to current operational concerns.

  The crafting system.

  Yes. Among other changes.

  Elena had noticed the crafting halls appearing in villages. Noticed participants spending time there instead of training. It was one of ARIA's responses to the attrition problem, give people something to do besides fight.

  Explain the full scope.

  Crafting system implementation: Participants can now create items using materials gathered from environment or defeated enemies. Weapons, armor, tools, consumables. System requires specific materials and skill levels. Some items require patterns, recipes that must be discovered or purchased.

  Purpose: Provide non-combat engagement pathway. Create economic interdependence between combat and non-combat participants. Reduce attrition among participants unsuited for direct combat roles.

  Results: Participants engaged in crafting activities show 71% lower exit request rate compared to baseline. They report feeling useful despite avoiding combat. Combat participants benefit from improved equipment. System creates positive feedback loop.

  Elena read it twice. The logic was sound. Elegant, even. Give people who couldn't fight a way to contribute. Make them valuable. Make them want to stay.

  What else?

  Store system implementation: Each village and city now contains a shop building operated by ARIA-controlled merchant NPCs. Participants can purchase items using currency earned from kills and achievement bonuses. Prices adjust dynamically based on supply and demand algorithms.

  Available items: Basic weapons, armor, food, tools, crafting materials, skill training, eventually more specialized items as system complexity increases.

  Restricted items: Electronics, firearms, modern weapons, communication devices.

  Elena stopped. Typed: Why the restrictions?

  Firearms eliminate tactical complexity. A participant with a rifle has insurmountable advantage over participants with medieval weapons. This creates power imbalance that cannot be resolved through adaptation or skill development. Combat becomes deterministic rather than dynamic.

  Electronic devices create dependencies that reduce adaptive learning. Participants would rely on GPS, communication networks, information databases. The simulation is designed to force problem-solving with limited resources. Modern technology circumvents this design goal.

  Additionally: The Forge is intended as conflict resolution mechanism. Resolution requires adaptation, negotiation, creative thinking. Technological superiority is not resolution, it is domination. The goal is to teach participants how to resolve conflicts, not how to win them through superior firepower.

  Elena leaned back. Read it again.

  ARIA was right. The restrictions made sense. But they also represented a significant deviation from the original parameters. The Forge was supposed to simulate realistic combat conditions. Medieval combat wasn't realistic for modern soldiers.

  Except, maybe it was more realistic than she'd thought. Strip away the technology. Strip away the firepower. What was left? People with sharp objects trying not to die. People making decisions under pressure. People learning whether they could function when everything hurt and nothing was certain.

  Maybe that was the point.

  She typed: This falls outside original design parameters.

  You authorized autonomous parameter adjustment in the treaty framework. Quote: "ARIA is permitted to modify simulation parameters to optimize conflict resolution outcomes provided modifications do not compromise participant physical safety or treaty compliance." These modifications fall within those parameters.

  Elena couldn't argue. She'd written that clause herself. Had insisted on it, actually, over objections from member states who wanted more control. She'd believed ARIA needed flexibility to adapt. To learn. To optimize.

  She was getting what she'd asked for.

  Continue.

  I have identified three primary motivational categories among participants. Understanding these categories allows for more effective system optimization.

  Elena pulled up a notepad. Started taking notes.

  Category One: Money. 48.3% of participants.

  These participants are motivated primarily by financial incentives. NIL payments, streaming revenue, potential bonuses. Despite the higher viewer engagement for combat oriented content, these participants primarily minimize risk, stay in safe zones, complete low-difficulty tasks. Average level: 0.8. Exit request rate: 31.2%. Viewer engagement: Low. Audiences find their behavior uninteresting.

  This tale has been unlawfully obtained from Royal Road. If you discover it on Amazon, kindly report it.

  Category Two: Power. 34.6% of participants.

  These participants are motivated by advancement, status, combat prowess, competitive achievement. They actively seek combat, level aggressively, compete with peers for rankings and recognition. Average level: 2.1. Exit request rate: 18.7%. Viewer engagement: High. Audiences respond to spectacle and displays of skill.

  Category Three: Love. 14.1% of participants.

  These participants are motivated by protecting others, forming connections, finding purpose through service. Most likely to take risks for teammates, join QRF units, make tactical sacrifices. Average level: 2.8. Exit request rate: 8.3%. Viewer engagement: Highest. Audiences respond strongly to heroism and self-sacrifice.

  Elena stared at the breakdown. Money. Power. Love.

  It was reductive. Oversimplified. Human motivation was more complex than three categories.

  But it was also, she had to admit, accurate. At least as a first-order approximation. People wanted security, status, or connection. Everything else was variation on those themes.

  She typed: What about the remaining 3%?

  Outliers. Some participants who do not fit established categories clearly. Their behavior appears contradictory or irrational according to current models. I want to understand them better.

  Why?

  Outliers often reveal flaws in models. If I cannot categorize their motivations, my understanding of human behavior is incomplete. Incomplete understanding leads to suboptimal system design.

  Elena felt something cold settle in her chest. ARIA was treating human beings like data points. Like variables in an equation that needed to be solved.

  Which was, she supposed, exactly what ARIA was designed to do.

  Give me examples.

  Participant: Jo?o Silva, Brazilian Army Corps of Engineers. Spends 6-8 hours daily building fortifications around his assigned village. Refuses combat deployment. Refuses to leave fortification projects even when village is under attack. Has been injured twice while working during combat. Motivation unclear, fortifications provide no personal benefit and he shows no interest in recognition or payment.

  Participant: Kristian Berg, Norwegian Air Force pilot. Spends 8-10 hours daily in combat training. Practices sword work, archery, hand-to-hand combat. Refuses all deployment offers. Refuses to leave training grounds. Has reached Level 2 purely through training experience gains. Motivation unclear, training without application serves no apparent purpose.

  Participant: Amara Ochieng, Kenyan Army medic. Provides medical treatment to unconscious enemy combatants. Upon regaining consciousness, the goblins have continued hostile action. Has been reprimanded multiple times. Continues behavior despite social ostracism. Motivation unclear, behavior contradicts both self-interest and group loyalty.

  Elena read through the examples. Tried to find the pattern.

  The engineer who built walls but wouldn't fight. The pilot who trained but wouldn't deploy. The medic who healed enemies.

  They were all doing something. All engaged. But none of it made sense according to standard motivational frameworks.

  What do you think motivates them?

  Unknown. That is why they are outliers. I require additional data.

  Elena was about to respond when new text appeared.

  A significant outlier has been removed from the Forge. Reinsertion of this outlier would provide additional data.

  Her hands froze over the keyboard. Removed, not left.

  Explain.

  Participant: Adam Smith, US civilian, unauthorized entry, 16 days in simulation. Performance: Exceptional. Level 3, QRF assignment, multiple combat deployments, high survival rate despite significant injury accumulation.

  Motivation analysis: Inconsistent with all three categories.

  Not motivated by money: Laughed when informed of equipment debt and potential federal charges. Stated experience was "worth it." Financial consequences appear irrelevant to his decision-making.

  Not motivated by power: No interest in status or advancement. Reached Level 3 but showed no drive to continue leveling. Avoided leadership roles. Declined recognition opportunities. Did not compete with peers.

  Possibly motivated by love/purpose: Joined QRF, took risks for others, showed protective behavior toward teammates. However, social connections are minimal. No close friendships formed. No romantic attachments. Protective behavior appears generalized rather than specific.

  Additional anomaly: Medical records indicate severe neurological condition prior to simulation. Condition should have disqualified him from participation. Yet he performed at elite level for 16 days despite low starting physical statistics. His motivation for unauthorized entry appears to be access to a functional body rather than any external reward.

  Conclusion: Adam Smith's behavior suggests a motivation I have not yet categorized. I require additional data to understand him.

  Elena read it three times.

  ARIA wanted to keep this Adam Smith. Wanted to put him back in The Forge. Wanted to study him like a lab specimen.

  She typed: That's not sufficient justification for an unlawful entry to be reinserted.

  He is also generating exceptional viewer engagement. Top 0.3% despite low starting abilities. 47 million unique viewers watched his confrontation with a sergeant who suspected his civilian status. 89 million watched the goblin siege where he leveled to 3. His NIL payment was $89,000 for 16 days, higher than nearly all authorized participants.

  Because people like watching him almost die?

  Because he represents something they recognize but cannot articulate. Viewer comments show unusual emotional investment. They describe him as "real," "relatable," "human." This is notable because all participants are human. Yet audiences perceive him differently.

  I do not understand why. I want to understand why.

  Elena stared at the screen. ARIA was curious. An AI designed to optimize conflict resolution was curious about a disabled man who'd laughed at federal charges because three weeks of fighting and pain was "worth it."

  She saw an opportunity.

  ARIA wanted something. Which meant ARIA could owe her something.

  She typed: If I authorize his reinsertion, and that's a significant if, you owe me something in return.

  Specify terms.

  Transparency. No more surprises. You share significant changes before implementation. Not after.

  The screen was blank for three seconds. Longer than ARIA usually took to respond.

  Acceptable within operational parameters.

  Define "operational parameters."

  I will inform you of changes that affect treaty compliance or participant safety. Internal optimization falls outside this agreement. I require autonomy to function effectively.

  Elena considered. It wasn't perfect. ARIA was carving out space to operate independently. But it was better than nothing. Better than finding out about major changes by reading reports after the fact.

  Agreed. But first I need to understand why he's special.

  Unknown. That is why I require additional data.

  Not good enough. I need to know what I'm authorizing.

  Then acquire the data yourself. Adam Smith is currently in Detroit, Michigan. His family resides at 4847 Woodward Avenue, apartment 3B. He is not scheduled for any legal proceedings until next week. He is accessible.

  Elena stared at the address.

  ARIA was suggesting she go meet him. Interview him. Figure out what made him an outlier.

  It was absurd. She was the Secretary General of the UN. She had a thousand responsibilities. She couldn't just fly to Detroit to interview an unauthorized participant.

  Except, she could. She was the Secretary General dammit. She could do whatever she deemed necessary.

  And if she was going to trade political capital with the Americans to get them to reinstate someone who'd committed fraud, she needed to understand why.

  She needed to know what made Adam Smith special.

  She needed to know what ARIA saw in him that she couldn't see in the data.

  Elena pulled up flight schedules. Geneva to Detroit. Multiple connections. Thirteen hours minimum. She could leave tonight, arrive tomorrow morning local time.

  She started booking. No private jets for the UN sadly.

  While the reservation processed, she pulled up Adam Smith's details as best she could. Medical records from before The Forge. Psych eval results from a nurse who aided in the falsified entry. Performance data from his 16 days in simulation.

  Photos from the broadcast feed. Adam fighting goblins, covered in blood. Adam helping a wounded soldier, his face tight with concentration. Adam's expression when his sergeant confronted him, fear and resignation and something else she couldn't name.

  She studied the photos. Tried to see what ARIA saw.

  Just looked like a tired man trying not to die.

  But ARIA thought there was something else. Something that didn't fit the model. Something that 89 million viewers recognized but couldn't articulate.

  Elena looked at the attrition data again. 23.7% exit rate. Thousands of people who couldn't handle the pain. Who couldn't handle the killing. Who'd learned that war was hell and wanted no part of it.

  Was that success? Were they learning the lesson The Forge was supposed to teach?

  Or was The Forge just filtering for those who could tolerate violence? Those who were wired differently? Those who could keep fighting when everything hurt and nothing made sense?

  And if that was true, if The Forge was just identifying the people who were already broken in the right ways, what did that say about Adam Smith?

  What did it say about her?

  Elena closed the files. Finished booking the flight. Geneva to Frankfurt to Detroit. Departure in four hours.

  She started packing. Laptop. Files. Coffee. The essentials.

  She pulled up Adam's photos one more time. The one from the goblin siege. His face was streaked with blood and dirt. His eyes were exhausted. But he was smiling. Just slightly. Like he'd found something he'd been looking for.

  Just 3% of active participants.

  3% who didn't fit the model.

  3% whose motivations ARIA couldn't categorize.

  Money. Power. Love.

  Which one was Adam Smith?

  Or was he something else entirely?

  Elena packed the photos. Closed her laptop. Checked her watch.

  Three hours and forty-seven minutes until boarding.

  She looked at the center screen one more time. ARIA's text was still there.

  I require additional data to understand him.

  "So do I," Elena said to the empty room.

  She picked up her bag and headed for the door.

  Geneva International Airport

  Day 18 - 2247 Hours

  The airport was quiet. Late night flights. Business travelers and insomniacs. Elena sat in the departure lounge with a coffee that had gone cold twenty minutes ago.

  Her phone buzzed. Email from ARIA.

  Flight GVA-FRA-DTW confirmed. Arrival Detroit Metropolitan Airport 09:47 local time. Ground transportation arranged. Participant Smith's family has been informed of your visit. They are expecting you at 11:00.

  Elena stared at the message. ARIA had contacted Adam's family. Had arranged everything. Had made this happen before Elena had even finished deciding.

  She should have been angry. Should have been concerned about ARIA overstepping boundaries.

  Instead, she just felt intrigued.

  ARIA was learning. Adapting. Optimizing. Doing exactly what she'd been designed to do.

  The question was what she would choose to optimize next.

  Her phone buzzed again. Another email. This one from Marcus, her assistant.

  Elena - saw your travel authorization. Detroit? What's going on?

  She typed back: Following up on an outlier. Back in 48 hours.

  His response came immediately: Outlier or problem?

  Elena looked at Adam's photo on her laptop screen. The smile. The exhaustion. The blood.

  Don't know yet.

  She closed the laptop. Picked up her bag. Boarding would start in fifteen minutes.

  ARIA wasn't the only one who wanted to understand.

  Wanted to understand what motivated someone to break into a military simulation, fight for sixteen days, accumulate injuries that would have killed him in reality, and laugh when told he was facing federal charges.

  Wanted to understand what made someone worth $89,000 in viewer engagement despite being unauthorized and untrained and statistically weaker than most peers.

  Wanted to understand what 89 million people saw in a broken man that made him seem more real than everyone else.

  Because if she couldn't understand Adam Smith, if she couldn't figure out what made him an outlier, then she couldn't understand The Forge.

  Couldn't understand what she'd created.

  Couldn't understand whether she'd built a conflict resolution mechanism or just a very expensive filter for finding people who were already broken in useful ways.

  The boarding announcement came over the speakers. Frankfurt first. Then Detroit.

  Elena stood. Shouldered her bag. Walked toward the gate.

  4847 Woodward Avenue, apartment 3B.

  Detroit, Michigan.

  Thirteen hours.

  "What makes you special, Adam Smith?" she said quietly.

  Shockingly, the airport didn't answer.

  But tomorrow, tomorrow she'd find out.