When Accuracy Was Not Enough by Bishop

When Accuracy Was Not Enough by Bishop - July 2026 Scholarship Essay

At first, I thought a good machine learning model was simply one that just got a high score. Then DermaBridge came in and taught me that a number can look impressive and still hide who the model is failing.

Indeed, The academic topic that challenged me most was fairness in artificial intelligence, especially in healthcare. Before studying it deeply, I understood accuracy, training data, and model performance in a basic way. If a model reached a high percentage, I assumed it was doing well. But while working on DermaBridge, my skin-cancer screening research project, I began to see that the real question was not only, “Is the model correct?” It was also, “Who is it correct for?”

That question completely changed the way I think.

In our project, we studied how artificial intelligence could support earlier skin-cancer screening and clearer risk guidance. Since skin conditions can appear differently across different skin tones, I had to think beyond general performance. A model could perform well overall but still struggle with underrepresented groups. A dataset could look large but still be unbalanced. A result could look strong on paper but still be unsafe if it missed the people who needed it most.

This challenged my critical thinking so much because it forced me to slow down and question my assumptions. I had to look at data distribution, class imbalance, missing information, validation accuracy, precision, recall, and whether the model’s results were actually meaningful. I learned that a high score does not automatically equal fairness. I also learned that technology should not be trusted just because it is complex. It has to be tested, questioned, and explained.

Honestly, this topic also connected to my life outside the classroom. After losing my grandpa, I became more sensitive to how delayed care, unclear information, and unequal access can affect families. That pain helped me understand why fairness in healthcare technology matters. It is not just an academic issue. It can decide whether someone receives attention early or is overlooked.

My small African community initiative, The B Foundation, also shaped how I saw this topic. Through providing school supplies, food, mentorship, and encouragement to children and struggling families, I have learned that access is never just a word. It can be the difference between hope and discouragement. Studying AI fairness helped me connect my technical education to that same belief: systems should be built with people in mind, especially those who are usually left out.

As The respected scientist Albert Einstein once said, “The important thing is not to stop questioning.” That is exactly what this topic taught me. It made me question numbers, assumptions, datasets, and even my own definition of success.

Overall, Fairness in artificial intelligence was difficult at first because it demanded more than just technical understanding. It required patience, humility, and responsibility. But it strengthened my critical thinking because it taught me to look beneath the surface. Now, when I see a result, I do not only ask whether it works. I ask who it works for, who it misses, and how it can be made better.