Folks, welcome. It's a pleasure to have you with us. This is our program on the effects of artificial intelligence on the management of people and the related issues there. My name is Peter Cappelli. I've been a professor here at the Wharton School for 35 years. I run our Center for Human Resources, among other things. I've been studying the workplace for a very long time, and most recently I've been looking at changes in how people are paid, and changes in how performance is managed, and more generally about the move we've been seeing over the last generation toward more open labor markets and movements across the employees. My colleagues and I on this program have been looking at the role that artificial intelligence is playing in managing people. Matthew Bidwell, you'll hear from shortly, a colleague here who's created the people analytics courses here at the Wharton School and colleague Sonny Tambe, who is a reformed computer scientist, I guess you could say, who studies workplace issues and particularly how various electronic markets affect issues like hiring. We're going to spend the next four hours or so with you, depending when you watch this, talking about those issues. Let me begin with the general point about managing people. There are some people probably watching this who have not had a lot of experience managing people before, particularly if you're coming at this from the data science or artificial intelligence side. Managing people is a huge task. If we look at the costs involved in a typical organization, for example, two thirds of all the costs come from employees or labor of various kinds. If you look just in the United States, every year we hire 66 million people, and that's out of a workforce of 160 million, so you get a sense that some of those people are going out one door and into another, it's another issue. Along the way we have to talk about how to manage people's performance, because the big question is how do we get people to do what we want them to do and what we need them to do? We have to think about how we pay them, the careers through which they move and how we help them move, all those issues. To get a sense of how big this is, if you look just at the industry that has been formed by people pushing work out, outsourcing, managing people issues, payroll and performance issues, but also staffing and hiring issues, if you total up the size of that industry, it's half a trillion dollars just in the United States. That's almost the size now of the entire US construction industry. This is a big issue and one with real money consequences for organizations and for societies. Particularly, for businesses, one of the things they care about is how the company's brand is affected by people management issues. We're seeing a lot of concern now about sexual harassment, diversity and inclusion issues is how they affect the company's financial performance, frankly. Lots of companies now are asked to demonstrate their compliance with various diversity inclusion policies and anti-harassment policies, etc. When we think about how companies compete, the key issue in business strategy, at least it seems to me, has been this question of competencies, and that is how does a business get good at something in a way that allows them to compete and succeed against competitors around them? Almost all that conversation now comes down to questions of people in the way we manage them, because that's about all it differs, frankly, across companies. Of course, ultimately, a reason for concern about all these topics is that managing people affects their lives. Managing people badly, causes a range of health problems, stress related problems in particular, and frankly just makes people miserable. If we can manage people well, we can make their lives a whole lot better. That just by itself is something worth taking very seriously. Let's talk just a little about what the issues are we're going to talk about here over the course of the next few hours with you. We begin with this issue of what is artificial intelligence? Let's see if we could talk about that for just a minute. It's a moving target, artificial intelligence, and the idea behind it is those decisions, that intelligence, which can only be made by people, by humans, and the problem is that it is a moving target because we get better with computer science and decision making, etc, we discover that we can do more and more decision making with those. What we have at the moment, though, when people are talking about artificial intelligence, is mainly data science. What data science is about is an engineering approach to data. This is the equivalent, you might say, of as engineering is to science, data science is to statistics. Statistics is the underlying discipline on which data science is based. Data science is particularly applying statistics and related data analysis tools to questions of optimization. What do we mean by that? The examples would be thinking about supply chains, thinking about how we make sure we get just the right amount of stuff to our stores at just the right amount of time. You can think about something like optimal scheduling with truck fleets, for example, to make sure that we know we have just the right number of truck fleets at just the right time. The problem when we think about people though, is that unlike maybe trucks, people get sick, people also get angry and they act out when they're angry, they quit. If they're not engaged in their work, they don't care about it. They can slow things down. They can actually sabotage. It was an old technique in the union world called work to rule, and that was if you were really annoyed with the management, what you do is just take the work rules right out of the manual and follow them word for word and you could basically just bring the organization to a stop. There are a lot of consequences about how you manage people and a lot of implications which are quite different than the optimization norms of data science, which came out of things like studying machines and studying equipment. The fundamental debate in management over the last century or so or since the beginning of management, something we'll talk about a little later in the program, is about the divide between management views, leadership views that think about people as rational actors and think about managing people along those equivalent lines. Versus, a movement that began 1930s but didn't hit its stride till the 1950s informed more by psychology, thinking about how people actually respond. Sometimes people call this theory X. Douglas McGregor's famous distinction here, theory X was people are rational. That means they don't really want to work unless you pay them and if you don't check, they won't work hard, etc. Theory Y, which says if you manage them correctly, they might actually do what you want them to do. The way to test where you think you are on this distinction is to think is it better to tell your employees what to do or ask them what to do? Kind of two extreme views right?. The former is a theory X view. The latter is a theory Y view. The theory X view aligns itself with engineering principles, as we'll see a little later and data science and the management of data that goes with it is based on techniques like machine learning. You apply those to equipment, for example. What are they trying to do? We're trying to figure out when the ball bearings might fail on this very expensive piece of equipment. Now, how do we want to do that? Well, we might say we don't particularly care. As long as you can predict it, we'll just listen, maybe put microphones on there or look for filings in the oil or whatever it might be, as long as it predicts when the bearings are going to fail, we don't care how it's done, that's all we care about. That's good enough. Once you move to employees and if you asked a question like, well, when are these employees going to quit? You did things like listening to their telephone calls, for example, or go to their social media and check in, I think the difference there is that we care how you get that answer, the answer is not simply a good prediction on when people are going to quit. People get very irritated, if you have a good explanation, we can predict with 100 percent who's going to quit if it turns out we did that based on listening to your phone records. One of the things, of course, it's most different about managing people is that these issues, like fairness, matter a lot. Fairness questions get embodied in the law, for example, in government regulations of all aspects of employment. When we start making employment issues and decisions, we start bumping up against these fairness questions and in practice, that means we're bumping up against the law a lot. You don't see that in other aspects of data science. One of the big issues related to managing people when we confront data science with that is the problem of change management. It's a general capability that is really important in management. Change management basically means we have organizations that have been doing things one way we want them to change. Before data science came along, we had a very intricate and you might say very messy system for managing employees. If you don't think so, go to your human resource department and ask to see the documents and manuals that are there that describe how employees are supposed to be managed. You'll see maybe it's all electronic now, but you'll see lots and lots of files and hundreds of pages explaining how we should manage people. We've invested a lot of time in training people as to how to manage employees. If with data science, we come into it with a very different approach as to how to make decisions on something like hiring, we confront this problem of change. We got this complicated system in which people are deeply invested already and we're asking them to do it differently now and so that is going to be a big question. We also have to think about a little bit, even when we have data science tools, which are really good and they do what we want them to do, how do we get people to adopt them? That's partly a story about how do we get them to give up some of the decision-making that they currently are exercising to turn it over to these algorithms. A simple way to think about this is if you're applying data science to something like allocating the hotel rooms, let's think about optimal pricing to make sure that we keep our rooms 100 percent occupancy or as close as we can get. The hotel room is not going to sue you if you make a mistake with that. But if you are allocating employees across work tasks and jobs, for example, and it turns out that those applications have some implications for fairness, it's quite possible the employees might end up suing you. The same algorithms in different context lead to really, really different outcomes. One of the things we have to think about is the resistance from not just the human resource community, but line managers who are currently doing things in a quite different way. There are good reasons for thinking about trying to do a better job carefully with managing this process of change. Let's think just a little bit about the opportunity here and I think in my lifetime, looking back at the things that I have seen coming along that changed business and changed the managing of people. This could be one of the very biggest ones, because its potential is the knock on effects that we change the way we make decisions by imposing data science tools on it, that changes the way we collect data, that changes the allocation of decision rights across people, that impact some of these fairness questions, all that kind of stuff. The good news is that there is a big opportunity here in introducing data science and the reason is because a lot of these decisions we're talking about, we just don't do very well right now. As I'm sure you know, on questions like hiring, there's a lot of concern that not only are we not doing a great job getting the right people into the right jobs and finding good employees, but that there's a lot of bias involved and a lot of discrimination in these outcomes. There's a lot of room to get better, it's not like we're tweaking a model which is running really, really well. If we get better, this makes people's lives better as well. That's where we are with this, it is a real opportunity to get better. It is a real challenge to think about how to apply these tools, which were developed for a very different context in things like manufacturing and applying them to people. As I said earlier, I'm going to be here with you throughout this program at different points. My colleague Matthew Bidwell will be here as well for a while, and our colleague Sonny Tambe will be as well. You'll see us in and out over the course of this program. We look forward to seeing you all the way through to the end.
