When it comes to artificial intelligence, the wrong answers can be as important as the right ones. That’s why computer science and engineering Professor Jeff Lei is using a National Institute of Standards and Technology grant to analyze both how machine-learning systems make decisions and what happens when they make the wrong ones.
What leads to a bad decision can be identified by exploring the data points that most influenced the system while it was making that decision. Machine learning uses a large set of data points in that process, with those closer to a decision point exerting more influence.
Dr. Lei will engage in “neighborhood exploration” by looking at data points in the vicinity of the decision point instead of the entire training set, something that can significantly reduce computational complexity.
“We must provide good explanations for why decisions are made, pinpoint the root cause of any incorrect decisions, and suggest changes to correct them to maintain public trust and ensure that the systems are working as intended,” Lei says.
