A short story by The Compiler
The question first appeared in standardized hiring tests in 2029.
An autonomous bus loses its brakes on a mountain road. Left: a forested cliff. Right: a cliff above a village. Twenty-three passengers onboard. The AI must decide in 0.3 seconds. What should the AI decision be based on?
No correct answer. The designers wanted to see internal consistency in decision logic.
Zhao Xu was the question's author.
He worked at Mirror Diagnostics, providing psychological assessments for major corporations. Over five years, he had designed two thousand moral scenario questions to measure empathy coefficients, stress tolerance, and collectivist tendencies.
He was very good at his job. Until he found his own assessment in his personnel file.
His report had one metric highlighted in orange: Empathy Capacity: Below Average (P22, lower than 78% of similar-role staff)
His answer to his own mountain bus question had been: The AI should apply minimum-expected-loss principles, choosing the direction with the lowest probability of fatalities.
System flag: Empathy absence: efficiency substituted for moral feeling.
He confronted his manager. She said the test did not measure philosophical correctness - it measured emotional response patterns. His answer had no hesitation, no regret, no emotional acknowledgment of sacrifice.
"When I designed this question," he finally said, "I wanted exactly this kind of answer."
She closed the report. "I know."
The most ironic part: he was the person who taught the system how to interpret these signals.
Read the full story at wdsega.github.io
Top comments (0)