In the digital age, algorithms and data have become important driving forces for social operation and economic development. From financial transactions to social media, from healthcare to transportation, algorithms and data are everywhere, profoundly changing our way of life and social structure. However, with the wide application of algorithms and data, the problems of algorithmic bias and data discrimination have become increasingly prominent, bringing numerous challenges and issues to society.
Algorithmic bias refers to the unfair and unreasonable differential treatment of certain groups or individuals by algorithms during the design, development and application of algorithms due to reasons such as data, models, objective functions or human factors. Data discrimination refers to discriminatory results against certain groups or individuals during the process of data collection, organization, analysis and application due to data bias, omission or error.
The COMPAS crime risk assessment tool is a notable example. Although it was intended to predict recidivism, it marked black defendants as “high-risk” almost twice as frequently as white defendants, despite similar recidivism rates (Angwin et al., 2016). This is not a failure of machine logic, but rather a reflection of historical policing inequality encoded in the dataset. When past biased patterns are used to train predictive systems, the algorithm not only replicates discrimination – it also reinforces it under the guise of objectivity and legitimizes it.
Solving algorithmic bias requires more than just different datasets. Fairness must be an intentional design goal. Friedler et al. (2019) emphasized that they are only effective when developers clearly prioritize interventions that enhance fairness. Transparency is also crucial: when the design and training processes are under review, society can question algorithmic decisions rather than treating them as neutral. What follows is accountability – recognizing that fairness is not a default state but an ongoing negotiation.
Technology should not merely record past inequalities; It should contribute to building a fairer future. Without intervention, with the advancement of technology, prejudice will become more concealed and dangerous. To prevent this from happening, we must abandon blind trust in algorithms and recognize that every system reflects human choices. Only by acknowledging the social nature of technology can we begin to create systems that truly serve everyone.
Reference list:
Angwin, J., Larson, J., Mattu, S. and Kirchner, L. (2016). Machine Bias: There’s Software Used Across the Country to Predict Future Criminals. And It’s Biased Against Blacks. ProPublica, 23 May.
Friedler, S. A., Scheidegger, C., Venkatasubramanian, S., Choudhary, S., Hamilton, E. P. and Roth, D. (2019). A Comparative Study of Fairness-Enhancing Interventions in Machine Learning. Proceedings of the Conference on Fairness, Accountability, and Transparency, pp.329–338
Noble, S. U. (2018). Algorithms of Oppression: How Search Engines Reinforce Racism. New York: NYU Press.
