Researchers at the College of Information & Computer Sciences have designed a new tool to create algorithms that prevent undesirable behaviors in machine learning, giving users more control in specifying safer and less biased results in applications.