Responsible AI cannot be retrofitted after deployment. That is the central argument of a special issue of the INFORMS Journal on Computing, published September 3, 2026, which presents 13 studies on building ethical considerations into AI systems from the earliest stages of development. The research examines how data science can create meaningful societal impact across judicial systems, education, communication, health care, bias and fairness, and interpretability.
"Responsible AI cannot be achieved by addressing ethical consequences after the systems are built," said Ram Ramesh, Ph.D., area editor of the journal and professor of management science and systems at the University at Buffalo School of Management, who co-edited the issue. "Social goals must be incorporated throughout the AI life cycle, from deciding what data to collect and how models learn to determining what they optimize, how they interact with people and institutions, and how they are governed."
Fairness as a design objective
The researchers argue that AI systems do far more than generate predictions. They allocate resources, shape opportunities, influence institutional processes and ultimately affect long-term societal outcomes. One study in the health care field found a machine-learning appointment scheduling system had Black patients waiting approximately 30% longer than non-Black patients. Researchers eliminated that disparity by incorporating race into the system's optimization objective while maintaining scheduling efficiency.
Bias can also develop through less obvious channels. Historical inequities embedded in data, incomplete feedback and interactions among different institutions can all produce disparities. Because algorithms can influence the data generated by future decisions, those disparities can compound over time.
"Fairness should not be viewed as an external constraint imposed on otherwise optimal systems but as a core design objective," Ramesh said.
Building social goals into the AI life cycle
The special issue, titled "Responsible AI and Data Science for Social Good," covers six domains where AI systems intersect with public institutions and individual well-being. The studies span judicial systems, education, communication, health care, bias and fairness, and interpretability. Each domain presents distinct challenges for researchers who must evaluate who benefits, who may be harmed and how systems can be optimized for social good.
For professionals working in AI for Science & Research, the findings reinforce a methodological shift. Rather than treating ethical review as a final checkpoint, the research demonstrates that fairness metrics, data collection protocols and model optimization targets must be defined concurrently with technical specifications. The full study was published by INFORMS under DOI: 10.1287/ijoc.2026.ed.v38.n4.
Why this matters for science and research professionals
Researchers designing AI systems for institutional use need to document how fairness constraints were incorporated during model development, not after. The scheduling study shows that bias mitigation can succeed without sacrificing performance when it is treated as a first-order optimization goal. For teams building models that allocate resources or influence institutional decisions, the takeaway is practical: define who might be harmed during the problem formulation stage, and bake those constraints into the objective function before training begins. Professionals can deepen their approach to these methods through AI Research Courses that cover responsible design frameworks.
Your membership also unlocks:
