The Algorithm

Chapter 7: Buried Warnings

Takeshi leaned back in his chair, staring at the old email thread on his screen. Someone had flagged concerns about the algorithm months ago, and those concerns had been dismissed. His mind raced with questions. Who was this person? Why were they ignored? And where were they now?

He scrolled down, checking the sender's name: Dr. Kenji Murakami. The name was unfamiliar. Takeshi quickly searched the hospital's internal directory—nothing. Then, he expanded his search to external databases. After a few minutes, he found something: Murakami had left the hospital six months ago.

Takeshi tapped his fingers on the desk. That's around the time the algorithm moved into its final validation phase. Was it a coincidence, or had Murakami left because of the algorithm's issues?

There was only one way to find out.

Murakami's contact information wasn't listed, but a quick search revealed that he was now working at a private AI research firm in Tokyo. Takeshi hesitated before drafting an email.

Subject: Inquiry Regarding Previous Research at KyoTech Hospital
Dear Dr. Murakami,
My name is Takeshi Jaka Mulyana. I'm a postdoctoral researcher working on the hospital's AI diagnostics project. While reviewing past documentation, I came across your analysis of the algorithm. I'd like to ask about your findings and why they were dismissed. Any insights you can share would be greatly appreciated.
Best,
Takeshi

He hit send, hoping for a response.

The reply came quicker than expected. Less than an hour later, an email from Murakami appeared in Takeshi's inbox.

Subject: Re: Inquiry Regarding Previous Research at KyoTech Hospital
Takeshi,
I was wondering when someone would ask about that. It's complicated, but I can tell you this—what you're seeing now isn't new. Some of us tried to bring up the issue before, but we were told to "trust the model" and not interfere. I had my doubts, but ultimately, the hospital's leadership decided the risks were acceptable.
If you're looking into this, be careful. Pushing too hard won't get you anywhere.
Kenji

Takeshi read the email twice. Murakami's words confirmed his suspicions—the bias in the algorithm wasn't overlooked, it was accepted.

Later that afternoon, Takeshi sat in Dr. Saito's office, recounting what he had learned. She listened in silence, her expression unreadable.

"So they knew," she said finally, tapping her fingers on her desk. "They knew and chose to move forward anyway."

Takeshi nodded. "Murakami warned them, and they ignored it."

Dr. Saito sighed. "I wish I could say I was surprised."

Takeshi frowned. "What do we do?"

She exhaled. "That depends. The hospital sees this conference as a major milestone. If we push too hard now, we risk sidelining ourselves. If we let them present it as is, we may get an opportunity to fix it afterward."

"But what if we don't?" Takeshi asked. "What if this gets buried again?"

Dr. Saito didn't answer right away. Instead, she leaned forward. "Takeshi, think carefully about what you're willing to risk. You've done incredible work, but you're still a postdoc. If you become the face of resistance, it won't go well for you."

Takeshi swallowed. She wasn't wrong. He had no real power in the hospital—no influence. If he pushed too hard, he'd be the one to suffer for it.

For the next few hours, Takeshi wrestled with his thoughts. He had two choices:

Step back and let the algorithm be presented at the conference, hoping they could fix the issues later.

Continue investigating, knowing that if he pushed too hard, he might be cut off entirely.

He was leaning toward the first option—until he found a document he wasn't meant to see.

While searching through old files, he stumbled upon a draft of a presentation prepared for an internal meeting before the project went public. His eyes scanned the text quickly, but one section stood out:

"Known Limitations & Justifications"

Demographic bias acknowledged; model adjustments deemed unnecessary for initial deployment.

Early trials suggest strong predictive accuracy for target groups; priority given to securing research funding for next phase.

Takeshi's breath caught. This wasn't negligence. The hospital had knowingly ignored the flaws to get the algorithm past early testing phases.

He sat back, pulse racing.

This wasn't just a technical problem. It was a decision.

And now, he had to decide what to do about it.