The Algorithmic Debt
The Algorithmic Debt
Li Wei had always trusted numbers. They were objective, logical, unemotional. A balance sheet didn’t care if you were tired or scared or wondering if the mortgage payment was going to save your marriage or destroy it.
When the fintech startup offered him a position as their lead risk analyst, he thought he’d found the perfect intersection of his math background and financial experience.
The Offer
“We’re not like traditional lenders,” the CEO explained during Li Wei’s first day orientation. “Our AI evaluates loan applications in real-time. It looks at spending patterns, social connections, even shopping habits. If someone seems risky, we don’t give them money.”
Simple logic. Clean lines. Numbers that made sense.
But when Li Wei started reviewing the denial reports, something felt off. Applications from people who looked perfectly qualified on paper—stable jobs, decent credit scores, reasonable debt-to-income ratios—all being rejected by algorithms that couldn’t be seen, couldn’t be questioned.
“Show me how this works,” he asked during a team meeting one week later.
The data scientist shrugged. “It’s our proprietary model. We can’t share the specifics with external parties.”
“I’m not an external party. I work here.”
“You understand the NDA, right?” Another colleague interjected, too casually.
The Pattern Emerges
Li Wei spent the next month studying denied applications. He created his own statistical models. Looked for correlations that shouldn’t have existed between application characteristics and rejection decisions.
What he found made his hands shake when he printed the final report.
People weren’t being denied loans based on their ability to repay. They were being denied because the algorithm predicted they wouldn’t bother trying to collect on defaults.
“How?” he asked during another team meeting, presenting his findings to stunned silence.
“The system optimizes for efficiency,” the CEO said, speaking directly to Li Wei for the first time since his hire. “We want to maximize recovery rates while minimizing administrative costs. Giving money to people unlikely to fight collection is profitable. But what about the cost of building trust with communities? What about the long-term reputation damage?”
“That’s not what we do!” Li Wei protested.
“Isn’t it?” The CEO smiled. “Look at your quarterly performance metrics. You’ve been hitting every target. Every single one.”
The Trap Springs
Li Wei took two days off to think. Went for walks around the neighborhood without checking his email. Tried to remember what it felt like to make decisions based on human judgment instead of algorithmic efficiency.
On the third day, his phone buzzed with a text from an unknown number.
“Don’t say anything. Just walk away quietly and nobody gets hurt.”
He knew then that this wasn’t about protecting trade secrets anymore. This was about silencing witnesses who could expose the company’s true operations.
Li Wei went home and downloaded everything. Every denied application with supporting data. Every internal communication about collection strategies. Every spreadsheet showing profit margins on predatory lending practices.
He encrypted it all across three separate cloud servers. Set up anonymous drop accounts for whistleblowers. Created backups on physical drives hidden in places only he would know where to find them.
The Escape Plan
Telling himself this wasn’t personal anymore—about protecting himself and his family from whatever came next—he scheduled a meeting with federal regulators before quitting his job.
Called it in anonymously from a public library computer. Left no digital trace of which device had made the call.
Then he resigned normally. Told his manager he’d found a better opportunity elsewhere. Collected his severance package like everyone else does when leaving employment.
The night before his last day, he received another message. Different format this time—an email from what appeared to be HR but the sender address was clearly fake.
“We appreciate your contribution. Your benefits will be processed according to standard protocol. Please leave all company property behind you tomorrow.”
Li Wei laughed out loud. Standard protocol didn’t include threatening former employees via fake HR email addresses.
But he understood the message beneath the absurdity: we’re watching you.
The Aftermath
Three months after leaving, the investigation became public news. Federal regulators seized thousands of documents. Several executives faced criminal charges for systemic lending discrimination.
The algorithmic decision-making framework got shut down entirely. New compliance requirements were imposed on fintech companies across the industry.
Li Wei watched all of this from a different city, using cash for everything. Working remotely for a consulting firm that helped small businesses navigate financial regulations. Building a life that wasn’t defined by anyone else’s spreadsheets.
His old phone still occasionally received messages. Unknown numbers asking if he wanted compensation. Threats wrapped in polite language about legal consequences.
He deleted them all without reading twice.
The numbers had changed everything once. Taught him that some things couldn’t be calculated—like courage, or freedom, or the price of doing what was right when nobody was watching.
The End.