An AI analyst in Washington, D.C. is training a fairness-aware machine learning model and must select a diverse test dataset. She chooses 5 participants from 3 demographic groups: 8 from Group A, 6 from Group B, and 4 from Group C. What is the probability that her selection includes at least one participant from each group? - Sterling Industries

An AI analyst in Washington, D.C. is training a fairness-aware machine learning model and must select a diverse test dataset. As AI evolves, ensuring models reflect the full range of human experiences is key—especially in policy, equity research, and public sector applications. To support this, this analyst carefully chooses 5 real-world participants from three distinct demographic groups: 8 from Group A, 6 from Group B, and 4 from Group C. But beyond diverse representation matters, understanding how likely such balanced inclusion is under random selection reveals important insights into data integrity and fairness testing. What’s the chance her sample includes at least one participant from each group? This question reflects a rising trend among developers and researchers prioritizing inclusive testing frameworks.

Why Diverse Data Matters in Modern AI Development
With increasing emphasis on ethical AI, test datasets must mirror societal diversity to avoid biased outcomes. In Washington, D.C.—a hub for data-driven policy—trainers like this AI analyst evaluate fairness not as an afterthought, but as a foundational step. Selecting participants from varied backgrounds ensures models are challenged across different lived realities, reducing blind spots. This mindset aligns with growing efforts to audit algorithms for equity, making thoughtful sampling essential for credible results.

Calculating the Odds: Probability of At-Least-One from Each Group
To determine the likelihood that a 5-participant selection includes at least one from each group, we use combinatorics. With 8 from Group A, 6 from Group B, and 4 from Group C, the total pool is 18 individuals. The selection uses simple random sampling without replacement. The full probability involves calculating favorable outcomes across feasible distributions—like one from each group with two replicas—and excluding cases missing any group. While detailed math is precise, the key takeaway is clear: inclusion across all three groups is statistically plausible but not guaranteed. Careful sampling ensures a more robust test of fairness, a critical concern in public-sector AI.