The Algorithm

Chapter 1: The Clattering of Keyboards

The clattering of keyboards filled the sterile, white-walled room. Takeshi Jaka Mulyana adjusted his glasses and stared at the screen before him. Rows of patient IDs and diagnostic results glowed on his laptop. The hospital had entrusted him, a postdoc specializing in data science, with validating their ambitious new project: an AI algorithm designed to detect early-stage pancreatic cancer. It was hailed as a revolutionary step in diagnostics, the kind of work that could save thousands of lives—if it worked.

Takeshi's supervisor, Dr. Saito, had emphasized the importance of his task that morning. "You're our final check," she said, handing him a flash drive with the training data. Her tone was encouraging but firm. "No shortcuts. The hospital board is counting on us to make sure this algorithm is ready before next month's conference."

He nodded, eager to prove himself. For Takeshi, this was more than just another project. It was a chance to make a mark in the field, a step closer to securing a future beyond the endless cycle of short-term research positions. The thought of achieving something meaningful buoyed him, but as the hours ticked by, a nagging sense of unease began to creep in. What if something went wrong? What if his analysis fell short of expectations?

By the second week, the algorithm's performance was beginning to look almost too good to be true. Takeshi's tests showed it identifying pancreatic cancer cases with an accuracy that seemed borderline miraculous. It flagged early-stage cancers that even seasoned doctors might miss. His colleagues in the adjacent office had been buzzing with excitement, speculating about the recognition their team might receive if the algorithm performed as advertised. And yet, something felt off.

Takeshi's work required detail attention and the more he scrutinized the data, the more strange pattern he began to notice. The algorithm excelled at identifying younger patients and those from urban areas, but when it came to patients over 70 or those from rural regions, its accuracy seemed to drop significantly. He ran multiple tests, comparing results across demographic subgroups, and the trend became increasingly clear: the algorithm wasn't performing equally for all patients.

One evening, as the office emptied and the soft hum of the central AC became his only companion, Takeshi decided to dig deeper. He pulled up the demographic breakdown of the training data and compared it to the test results. His brow furrowed as a pattern began to emerge. The training data, it seemed, was heavily skewed toward certain demographics. Younger patients from urban areas were overrepresented, while older and rural patients appeared as a statistical afterthought.

Takeshi leaned back in his chair, letting out a deep sigh. A quiet voice in the back of his mind whispered what he had been trying to ignore all day. Something wasn't right.

Could the imbalance in the training data explain the discrepancies? If the algorithm had learned to detect patterns primarily from urban patients, it might struggle with others. But how could such a fundamental flaw have gone unnoticed until now? The hospital's reputation was on the line, not to mention the lives of patients who might depend on the accuracy of this tool.

He took a sip from his now-cold coffee and rubbed his temples. Reporting his concerns would mean slowing down the project, something his team's leadership would undoubtedly frown upon. But if he ignored the issue and moved forward, the consequences could be catastrophic. Patients who trusted the algorithm could be misdiagnosed, their lives potentially at risk.

Takeshi glanced at the clock. It was nearing midnight. The office's fluorescent lights flickered slightly, casting faint shadows on the walls. He returned to his laptop and began drafting a detailed report, highlighting the trends he'd discovered.

The sound of footsteps startled him. He turned to see Dr. Saito standing in the doorway, her expression unreadable.

"You're still here," she said, stepping inside. "That's dedication."

Takeshi hesitated, then nodded. "I... I found something," he admitted. "Something that doesn't add up."

Dr. Saito crossed her arms, leaning against the doorframe. "What is it?"

He gestured to the screen. "The algorithm's performance isn't consistent across all demographics. Older and rural patients are being misclassified far more often. It's the training data—it's skewed."

Her expression shifted, a hint of concern breaking through her usual calm demeanor. "How confident are you?"

"Confident enough to know we need to take a closer look," Takeshi said, his voice firm. "If we move forward without addressing this, it could lead to serious consequences."

Dr. Saito sighed, pinching the bridge of her nose. "This isn't the news I wanted to hear, but it's better to catch it now than later. Draft your findings and share them with me first thing tomorrow. I'll figure out how to approach the team."

Takeshi felt a wave of relief. "Thank you, Dr. Saito."

She gave him a small nod before heading toward the door. "Go home soon, Takeshi. Don't overwork yourself."

As she disappeared down the hall, Takeshi stared at the screen, the glow of his findings reflecting in his glasses. He had a long night ahead, but for the first time that day, he felt a renewed sense of purpose. Whatever the cost, he knew he had to do the right thing.