Subjectivity and Suffering Boundaries of Data Science and Ethics - November 2025

Subjectivity and Suffering Boundaries of Data Science and Ethics - November 2025

The technical and mathematical aspects of data science and machine learning alone cannot adequately handle the "human suffering" and "ethical costs" of reality - this issue has become one of the most important issues in recent AI social implementation. No matter how accurate a prediction is, there is always a human being behind it who is being judged. When a false positive occurs, it is not just a "numerical error," but often a "life-destroying mistake.

This is an unavoidable problem when machine learning is used in critical social systems such as medicine, finance, crime prediction, education, labor, and government. In fraud detection and anomaly detection systems, both false positives and false negatives are inevitable, and both can lead to serious human rights violations and social distrust.

Moreover, it is not enough to simply improve performance. For example, if the threshold is adjusted to reduce false positives, then the original risk will be missed. Technically, this tradeoff is adjustable, but the question of "which damage is more unacceptable" requires a social and ethical decision. This is an area that cannot be solved by mathematics.

Furthermore, the problem of bias, in which the bias in the training data or the way it is pre-processed concentrates disadvantage on certain races, genders, age groups, etc., is also becoming apparent. An actual case in which a facial recognition AI misidentified a dark-skinned woman, leading to a false arrest, has been pointed out as not merely a technical error, but one that amplifies the imbalance in the social structure.

To address this reality, it is essential to incorporate ethical principles such as "fairness," "accountability," and "transparency" into algorithms. In addition, questions such as "Who will be harmed?" "Is there a remedy?" and "Can the decision-making process be held accountable?

Ultimately, data science is not about "accurately classifying" but an activity concerned with "how to protect human dignity and social trust. Acknowledging that suffering cannot be reduced to mathematical formulas, the sensitivity and responsibility to undertake the uncertainties and ethical burdens will be the qualities required of engineers in the future.