But for all its promise, the algorithm lived on a tightrope of paradox. It could only be as good as the data fed into it, and the data, in turn, came from a world steeped in inequality. Janet had spent countless nights wrestling with the model’s “fairness” constraints, adjusting loss functions, and adding layers of privacy preservation. The deeper she dug, the more she realized that “pure” might be an unattainable ideal.

Janet leaned forward. “What do you want me to do, Score X?”

Janet took a breath. “Option C,” she said, “but we must flag the result as provisional and provide a transparent explanation to the user.”

Months later, in a modest community center, a young woman named Maya walked in, clutching a printed copy of her Score X report. She sat across from Janet, who smiled warmly.