The Messy Locker Room: Mia’s Impossible Search Problem

The fluorescent lights of Westbrook High’s cafeteria hummed so loud Mia could almost feel the buzz in her molars as she stared at the tech fair registration form taped to the pizza counter. The annual competition was open to every student, but the prize for “Most Impactful Student Project” was the one she’d had her eye on for months: a $500 grant to build a tool that actually helped kids at school, not just another app for tracking homework or ranking lunch menu popularity. Three weeks earlier, she’d spent 45 minutes wandering the empty hallways after school, backpack heavy with textbooks, trying to find a quiet spot to finish her 10-page AP Bio research paper before it was due at midnight. The library was packed with a sophomore study group blaring K-pop through headphones, the art studio was locked for after-school pottery club, and the only empty classroom she found had a leftover birthday party mess of confetti and half-eaten cupcakes on the desks. She’d finally curled up on the floor of the old supply closet behind the gym, the smell of disinfectant and old basketballs in her nose, and promised herself she’d never let another kid have to do that.

Her CS teacher, Mr. Henderson, had spent a unit that semester teaching basic information retrieval, and Mia had aced the quiz on keyword matching. It made sense, right? If you want to find something, you just search for the words that describe it. She spent the next 10 days pulling every public campus resource she could find: the full lunch menu, library and study room hours, club recruitment posts, lost and found listings, nurse’s office availability, event calendars, even the list of vending machine locations and what snacks they stocked. She organized it all into a 1,200-entry knowledge base, each entry tagged with a handful of keywords she thought kids would use to search for it. She even named her bot Westbrook Helper, scribbling the name in bubble letters on the top of her coding notebook.

Her first test worked perfectly. She typed “what’s for lunch today?” into the simple chatbot interface she’d built, and it pulled up the exact menu: chicken tikka masala, vegan chili, and those weird but delicious matcha cookies everyone fought over. She whooped so loud the kid at the next table dropped his fork. This was it. She was going to win that grant, and by the end of the semester, every kid at Westbrook would be asking her bot where to find everything they needed. She’d even added a fun little Easter egg: if you typed “who’s the best CS teacher?” it would reply “Mr. Henderson, duh” with a little emoji of a computer.

Two days later, her best friend Liam, a senior stressing about AP Calculus homework due the next day, leaned over her shoulder in the computer lab. “Hey, can I test your bot?” he asked. Mia nodded, proud as a peacock. Liam typed: *Where can I get a quiet place to finish my math homework?* The bot blinked for a second, then spat back: *No results found for “quiet place to finish my math homework.”* Mia blinked. “That’s weird,” she said, typing the query herself. Same result. She frowned and opened her knowledge base, scrolling to the “study spaces” section. There it was, plain as day: *Library Study Rooms: Quiet individual and group study spaces available until 10PM on weekdays, 6PM on weekends. Sign up at the library front desk.* “Wait,” Liam said, squinting at the screen. “That’s exactly what I asked for. Why didn’t it find that?” Mia’s cheeks burned. “Uh, maybe it’s a glitch. Let me try another query.” She typed: *quiet place to do homework*. This time, the bot pulled up the study room listing immediately, along with a note from the librarian posted last month saying the second-floor study nook is reserved for quiet individual work after 4PM. “See?” Mia said, forcing a smile. “It works. You just have to type the exact right words.” Liam raised an eyebrow. “Who’s going to type ‘quiet place to do homework’? I typed exactly what I was thinking. If I’m stressed about math and need a spot to work, I’m not gonna waste time guessing the magic keywords your bot wants. This is dumber than the chatbot we built in middle school tech class.”

Mia spent the next three days testing the bot with every question she could think of, and the failures piled up faster than the confetti in that empty classroom. When she typed *I lost my blue basketball with a signature on it in the gym*, the bot returned nothing—until she changed the query to *blue basketball signature lost gym*, at which point it pulled up a lost and found entry from three weeks prior: *Blue Spalding basketball with “MVP” written in silver marker, found in gymnasium court 2, claim at front office.* The kid who’d lost it had typed *basketball*, not “basketball with a signature,” and the entry had never mentioned the word “blue.” When she typed *What club is good for people who like drawing anime?* the bot came up empty, because the anime club’s recruitment post only used the term “ACG fandom” and “2D art,” no mention of “anime” or “drawing.” When she typed *Is there a fun event happening next week to relax?* the bot didn’t pull up the upcoming Westbrook Student Council Lawn Music Festival, because the event listing only said “Free live performances, food trucks, and yard games for all students, next Friday 4-8PM on the main field,” no mention of “relax” or “fun.”

By the end of the week, Mia’s notes app was full of rants, and her test accuracy rate was a miserable 32%. She’d tried adding every possible synonym she could think of to each entry: she added “homework,” “study,” “test prep,” and “finish assignments” to the study room listing; she added “anime,” “manga,” “drawing,” and “cartoon” to the ACG club post; she added “fun,” “relax,” “stress relief,” and “hang out” to the lawn music festival entry. But every time she added a new set of keywords, a new batch of test queries would fail. There were infinite ways to ask the same question, and she couldn’t possibly tag every entry with every possible word a kid might use. She even tried adding common misspellings, like “omework” for “homework” and “libary” for “library,” but the list of edge cases just kept growing. Liam had been right: no one was going to waste time guessing the magic keywords her bot wanted. If she couldn’t fix this, her tech fair project was going to be a joke.

That night, she sat at her kitchen table staring at her laptop, a half-eaten bowl of cereal going soggy next to it. She’d always thought of computers as these super smart, all-knowing machines, but this bot was dumber than a goldfish. It couldn’t even tell that “quiet place to finish math homework” and “library study room” were related, even though any human at Westbrook would make that connection in half a second. She pulled up her CS class notes from the unit on information retrieval, scrolling past the quiz she’d aced to the lecture slides Mr. Henderson had used. There it was, a term she’d skimmed

The fluorescent lights of Westbrook High’s cafeteria hummed so loud Mia could almost feel the buzz in her molars as she stared at the tech fair registration form taped to the pizza counter. The annual competition was open to every student, but the prize for “Most Impactful Student Project” was the one she’d had her eye on for months: a $500 grant to build a tool that actually helped kids at school, not just another app for tracking homework or ranking lunch menu popularity. Three weeks earlier, she’d spent 45 minutes wandering the empty hallways after school, backpack heavy with textbooks, trying to find a quiet spot to finish her 10-page AP Bio research paper before it was due at midnight. The library was packed with a sophomore study group blaring K-pop through headphones, the art studio was locked for after-school pottery club, and the only empty classroom she found had a leftover birthday