When We Question the Measure of Life Where Objective Functions Quietly Draw the Line December 2025

When We Question the Measure of Life Where Objective Functions Quietly Draw the Line December 2025
When we entrust AI to make decisions, we often offer a single yardstick, the objective function. However, the direction of that yardstick does not always correspond to fairness or justice. For example, the objective function used in the medical field to maximize the number of years of life saved appears at first glance to be rational and suitable for optimal allocation of resources. But inside this formula, social conditions such as age, income, residential status, race, and medical history quietly creep in and have the power to put certain groups at a disadvantage.
A study of the U.S. health insurance algorithm, using health care spending as a proxy for health demand, revealed a structural bias that prevents black patients from receiving the full range of health care services they originally needed. The amount of spending is a result of access to health care, not health status itself. Nevertheless, the objective function interpreted low spending as low demand and built inequity into the system.
The objective function takes on more social and ethical significance than a mathematical formula. The choice of what to maximize and what to ignore is inextricably linked to the decision of who to prioritize and who to sacrifice. Equity studies have also shown that small changes in the objective function can significantly alter output disparities, and discrimination often arises at the objective-setting stage.
For this reason, the EU AI Act and the OECD AI Principles emphasize the importance of verifying not only the transparency of algorithms but also the validity of the objectives themselves. Since the very act of setting objectives determines the direction of future values, it is essential to question them from an ethical perspective before entrusting AI to make decisions.