����ֱ��

Humans vs. Machines

Law professor explores whether consumers want people or algorithms to make their day-to-day decisions

By Claire Curry

Would you want a human or a computer to determine the interest rate on your bank loan? How about the chance to win a $20 gift card to your favorite coffee shop?

Algorithms—or calculations that can problem-solve and cast decisions—aren’t new. However, the rise of big data and increasing concerns about consumer privacy have some legal experts questioning what consumers actually prefer.

While most would assume that human decisions would override algorithms, a surprising majority of consumers don’t mind if machines determine the outcomes in situations that affect their welfare, according to co-authored by Michael Risch, JD, professor and vice dean of ����ֱ��’s Charles Widger School of Law.

In their study, Professor Risch and Derek Bambauer, JD, a law professor at the University of Arizona, surveyed 4,000 participants about whether they’d prefer a human or a machine to make the decisions in four scenarios: receiving a gift card, being held liable for a civil traffic offense, getting approved for a bank loan or being included in a clinical trial for medical treatment. Overall, 52 percent chose an algorithm while 48 percent chose a human—but there were conditions.

“We tested whether people would be more likely to choose a human as the stakes increased, and that was generally true,” Professor Risch says. “So, for example, people were more likely to choose a human for the medical trial or civil fine than they were for the coffee shop gift card, because the stakes were so much higher.”

However, if the algorithm could promise a better price, participants chose the machine. When it came to the bank loan, there was no real difference between participants’ likelihood to pick an algorithm or a human. “The study tells us something we didn’t know before,” Professor Risch says. “Sometimes people like algorithms and sometimes they don’t, and they do so in a rational way. If the algorithm is cheaper, faster, or more accurate, they tend to prefer it.”

According to Professor Risch, many of the existing regulations and policies that involve using algorithms in decision-making are not grounded in evidence or they’re based on outdated data, so this new research provides law and policymakers with new information to consider in regulatory choices and policy reform.