MENU
A cloud computing company uses machine learning software to screen the resumes of job seekers. The company's aim is to reduce potential human prejudice in the hiring process. To set up the software and train the machine learning model, the company provides information from its hiring decisions over the past five years. After several months of using the software, the company runs an audit and finds that the software screens out minority job seekers at a much higher rate.
Which behavior is the trained machine learning model displaying?
The machine learning model unintentionally discriminates against minority job seekers because it was trained on historical hiring data that contained biases.
Why Exposing Bias in Data?
Machine learning systems reflect the biases present in their training data.
If past hiring decisions favored certain groups, the AI will replicate and reinforce those biases.
The audit revealed systematic discrimination, indicating bias in the data, not the AI itself.
Why Not the Other Options?
B . Acting fairly toward candidates -- The AI is not fair if it disproportionately excludes minority job seekers.
C . Exhibiting intent to harm applicants -- The AI does not have intent; bias is an unintended consequence of flawed data.
D . Scoring low humility measurement -- 'Humility' is not a standard AI metric; bias exposure is the actual issue.
Thus, the correct answer is A. Exposing bias that exists in the data, as the AI amplifies historical discrimination in hiring.
Reference in Ethics in Technology:
O'Neil, C. (2016). Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy.