Thought Leaders in Artificial Intelligence: Josh Sullivan, SVP and Angela Zutavern, VP of Data Sciences at Booz Allen Hamilton (Part 2)

Sramana Mitra: Can you parse that out? What are the tasks that machines do better and what are the tasks that humans do better in that context?

Josh Sullivan: They were having people look over their existing loyalty club members and trying to figure out how to upsell or get them to use their points that they have accumulated in different ways. They were missing something even bigger that said, “How do I go and entice people who are not already loyalty club members? Build me a predictive model of what it would take in order to covert those people.”

They had an entire set of machine learning path that would call through all of their stays. They said, “Let us go and figure out how to mine that and build a model.” It was incredibly successful. Then they started a series of experiments where a marketing team had 30 or 40 different incentives they wanted to try.

Instead of trying them one at a time and measuring how effective they were, they essentially stimulated different attributes of the profiles that they had built for the people who’ve stayed in their properties before but weren’t loyal. They did a probability match and let the computers decide which of the people would get which offers. It’s a very intensive way of thinking differently about how to segment your client base and how to tailor offers to them.

Sramana Mitra: Given large amounts of data to work with, machines will always make better decisions. Can you give me an example where humans make better decisions?

Josh Sullivan: We can talk about a different example. There’s a health startup called JOOL Health. They let people talk about wellness from an app. Some of the earlier recommendations of the algorithms in terms of how to improve wellness were very crude. They were very simplistic. The algorithms had a very hard time matching the complexity of human behavior and human thought. It had lots and lots of data.

Once they started interviewing some of their clients and asking different questions, that’s what changed JOOL’s path. JOOL is becoming even more successful now. Although the original generation of algorithms had a lot of statistical data, it wasn’t making good health and wellness recommendations in terms of human behavior.

The algorithms had a mathematical context of this person. They didn’t really have the nuance of human behavior. It took a whole different set of data points and thinking in terms of machine learning algorithms before it started being able to make better recommendations.

Angela Zutavern: The machines are good at solving problems. There are cases where machines can do things that may cross an ethical line. One of the examples we have in the book is about the US Census Bureau. They’re moving to a completely data science-driven, innovative model for conducting the 2020 census.

The idea that they have will save their enumerators. These are the people who travel around and go door-to-door collecting data. It’s going to save them thousands of hours of time and travel. However, there maybe data sources where they don’t want the government to collect that data and cross into an area of privacy. We need the people and the leaders to apply judgement about how far we should take machine intelligence, and what are the privacy and ethical issues we should address.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Josh Sullivan, SVP and Angela Zutavern, VP of Data Sciences at Booz Allen Hamilton
1 2 3 4