Algorithmic Management: Navigating the Gig Economy’s New Frontier | Lindsey Cameron

How Algorithmic Management Is Changing Gig Work

Dan Loney: Well, as more and more people add gig work to their professional mix, we also see an increase in algorithmic management. This means that workers are following the instruction of so-called managers, who are the process of algorithms. Recent research examines what that relationship is like and how workers feel about being managed by tech, and not by directly another human being. Pleasure to be joined by Lindsey Cameron, assistant professor of management here at the Wharton School, who’s led this research. Lindsey, great to talk to you. It’s been a while. Always great to touch base.

Lindsey Cameron: Yeah, definitely. Feel the same over here.

Loney: I know how much you have focused on a lot of these elements in your research. But I guess when you look at that relationship, it becomes even more interesting with how gig work factors in.

Cameron: Exactly. One thing I love to say about the gig economy is that it punches above its weight class. It’s only about 1% of the U.S. workforce, but 97% of people have heard of Uber and Lyft. You know, about 60 to 70% of Americans have gotten into one of those cars. And you think about all the sweeping legal regulations that we’ve talked about before, Dan. In California, Massachusetts, Minnesota. All about redefining gig work, which has all these ripple effects for other types of independent contractors. So the gig economy is really just a microcosm that helps us understand broader changes in our economy.

Loney: Right. I guess where this research is concerned, it’s unique to look at how these workers are being managed, and what that might mean for the concept of management moving forward.

Cameron: Exactly. We always think of management as having a boss. And like, “How do I avoid the bad boss and get the good boss?” But we have these longer and larger conversations about AI and control. Really, what is AI doing? It’s trying to algorithmically manage people. Outsource part of the work, the managerial function, to the algorithms.

And Uber is a really great example, from end-to-end. I worked part-time as a driver for three years. Never spoke to a single employee of Uber. Hiring, firing, evaluating, disciplining, all done by an algorithmic management system. And it really gives us a precursor or a view into the future.

Loney: What did that experience really mean to you? I mean, it’s unique. As you said, we’re so used to coming into the office and meeting with a boss, or talking with other people. And it’s a totally different experience, isn’t it?

Cameron: It is. You do have a lot of schedule flexibility. You can work around your schedule. But when things don’t go right, then you don’t know what to do. There was this whole period — I drove in in the DC, Virginia area. And there were three hours I was sitting near Dupont Circle. It was raining. I was seeing all these other cars go by, and people getting in the back seat. I’m like, “That’s an Uber. I think it’s an Uber. But why am I not getting any rides? Why am I here in the dark, not earning income?” And I was texting, and they were like, “No, you’re logged on. You’re eligible to give rides.” I had no real way to have voice and resolve an issue. And the next day, they apologized. They said something was down with the system. I was logged online but not getting rides.

There’s this joy in schedule flexibility, and then there’s this issue of when things don’t go right —whether it’s pay, something with a customer, or not getting rides. Then you’re talking to a robot and it’s hard to get resolution.

Loney: What does it mean, then, for the types of tasks that may be brought that person’s way in terms of dealing with them and handling them, depending on who or what is managing the process?

Cameron:  The types of tasks that you’ll see algorithms step in [to manage] are going to be sliced down to the smallest possible unit. We call that deskilling. But it’s in a much deeper way than, like, factory work, where you’re at an assembly line. That is a bit of deskilling. But this is like, can a task be completed in two seconds? You know, micro tasks.

When you think about, say, ride hailing, for example, it’s lots of little micro tasks. Will I accept this ride or not? Am I going to follow the GPS, Uber’s or Waze’s, to go where I’m going? Am I going to talk to the customer? Which way will I drive? And then do I rate them or not? They’re all these very, very small components. But because they’re so small, they can be algorithmically managed. And at the same time, the workers feel like they have choice. Because there are all these little individual elements. I have a very small but very real amount of choice. And I think that’s one of the reasons why people like this work so much, is this feeling of choice. Small but real.

The Good-Bad Job: Navigating the Paradox of Gig Work

Loney: But then you also have, and you use the term in your research, the “good-bad job.” How does that come into play?

Cameron: You know, I trust what my workers tell me. I didn’t love driving, but a lot of my drivers did like driving. And I think that was my first “aha” in the research. This is some sort of tension I need to look into. So, yes, the driver is telling me, “I’m earning more than I was at Walmart, or at the gas station. I can take care of my family. I like driving around town and showing people the sights of my city.”

But then if you zoom out of the workers, you’ve got to look at what’s the larger, legal, social, environmental influx. These people don’t have insurance coverage if they get into an accident. You might get coverage when the passenger is in the car. But what happens when you’re driving? You know, waiting to get that ride? You’re not covered on the same amount. You saw during COVID how there were questions about whether or not they would be covered to get the same sort of employment protections. There are issues with not being able to earn the minimum wage.

There’s a broader context that you’ve got to understand the work in, even if workers like it. And that’s why I call it the good-bad job. The thing is, it’s here to stay, and it’s not just the gig economy. Most of our work is becoming both good and bad, because these tensions are generative.

Loney: When you’re talking about working on these types of platforms, I guess you can consider it a bad job for a variety of different reasons, right?

Cameron: Yeah. There are reasons you can consider it a bad job. And I think just leaving it, like, “It’s a bad job, workers are exploited. They don’t have insurance, they don’t have minimum wage.” It’s true, but it’s only part of the story, because people are also enjoying this work. And for those that have been shut out of the labor market, particularly if you’re a first-generation immigrant — many of the people doing this work, they come over to the States and their credentials are not transferring, even though they have the skills. So they go into this type of work. Or they’re someone that’s been out of the labor market. Maybe they were incarcerated, or they were at home taking care of kids, or they were sick. It does provide an income-earning opportunity. So the fact they enjoy the work is real. In the same time, it exists in these larger conditions. And I think we’re finding more and more jobs, particularly as benefits for workers continue to be eroding, holding this tension.

Loney: When you look at where we are right now, in many cases, gig work is preferred to people over a traditional office.

Cameron: That’s a tricky one. I would say for my gig workers that I focused on — I’ve studied all the big platforms. Instacart, DoorDash, Amazon Flex, ride hailing. They typically have a lot of bad options. It’s warehouse, manufacturing, gas station, cater waiter. They see gig work as being the best option out of a sense of good options. But these were never people that had traditional office jobs. And so when you’re talking about those people who are preferring gig work over traditional office jobs, you’re talking about more high-skilled employees, or folks who are doing Upwork, or doing coding or copy editing. And that’s a different conversation, because you’re talking about a different type of laborer skill.

Loney: You talk in the paper about the element of consent, and the fact that that can be very important in terms of the perception of how good or bad a job may be. And I guess you have to look at it from those two perspectives — one being the human relationship, but two being the algorithmic relationship as well.

Cameron: That’s it. You have to have conversations with both. And particularly in this context, it’s conversations with the technology. But the classic question that paper looks at — and this question has been looked at over 100 years by different scholars — is, why do people keep on doing jobs […] that are kind of bad? Why do people enjoy work that qualitatively looks like it’s in bad condition?

And that’s the issue: the tension of the good-bad job that I’m hoping to unpack. And deeper in the paper, I talk about how people have small amounts of agency in their work. They might decline a whole bunch of rides in a row because they don’t want a shared ride, because you can get into fights with passengers if you have three or four people that are in the car. They might use geo-spoofing apps to try to get higher paid work.

Or, they can try to inflate surges [in pricing]. This is one of my funniest stories. It was the Wednesday before Thanksgiving and everybody’s heading out to the airport. I interviewed somebody who primarily drove in a college city, and he’s like, “They weren’t paying me enough to go to the airport. I just stayed in front of the dorms, and I clicked decline, decline, decline. And then I got a ride for $160 to go to the airport, and it’s usually $40.” He quadrupled his wage. There are ways they’re able to have small amounts of agency within this larger system of work that’s really not set up to give them any protection or benefits.

How Will Algorithms Impact the Future of Work?

Loney: How do you think the technology used in these types of jobs will potentially impact our larger workforce as we move into the future?

Cameron: You know, it’s a great question. And my first answer is I think it’s honestly scary. You think about those means of recourse. You know, I didn’t get paid for three hours. There are people that I interviewed who got kicked off the app for three days, couldn’t pay a utility bill, and supposedly, they said they were kicked out of the app because they had said something they shouldn’t have said to a woman. There was something that crossed gender norms. And then he got an apology message three days later, saying, “Oh, it wasn’t you. The algorithmic management system made the wrong reference. It was actually a different driver. You could come back on the app.” But for those three days, he had no income, and he had no way to navigate the system to get back on the app.

There is a concern that you keep on deskilling and de-splicing the work — one, it drives down wages, because people can’t really build skill. But two, algorithms make mistakes. And without a human in the loop, the human is lost.

Loney: Is the expectation that we probably still will have some of those conditions where algorithms do make errors from time to time, for a longer time? Because I think humans are putting a lot of the effort in behind the algorithm. And we certainly know that humans are not perfect. How can we expect the algorithms to be perfect if the humans are not?

Cameron: The fact that people hold these algorithms to god-like status, that [they] can’t make a mistake — I mean, that’s such a blind, naive view of technology. There’s always, always going to be a gap between how the technology is designed, and how it’s used and implemented by the workers. And in that space, you see agency, like what I talked about in the paper. But you also see mistakes the technical system is making. And I think the more we start getting blinded by techno-utopianism, that increases the risk. And the risk, particularly, for the most marginalized parts of our society. Because these algorithms, they’re refined and they’re tested on the most vulnerable parts of the population. You think about predictive policing. You think about whether or not I’m going to give somebody that has a low credit score a loan, or whether or not I’m going to let them go out for parole. These were tested on disenfranchised parts of the population, and once they were fine, bam. They come out for the masses.

And so, I think that’s a way to think about the gig economy and why it punches above its weight class, and why it’s having such a ripple effect. It started off with a group of more marginalized people who are shut out on the edges of the labor market. They’re refining these tools, and now they’re going after the middle-paying, the higher-paying, skilled jobs. And it’s just changing the landscape of work.

Loney: But does the concept of innovation really have the opportunity to grow and develop, because we’re adding in this component of technology? I mean, before, a lot of it was just innovation based on humans. Now we’re bringing in some other components to it.

Cameron: Maybe. If you want to be optimistic. I mean, I think it’s the question like, has Uber done more good for the world or more harm for the world? And it’s hard for me, as an academic, to see so black and white. I do think there are innovative things that they’ve done. In many ways, they’ve dismantled a taxi medallion system that wasn’t working for a lot of people. And at the same time, I think the feeling of what they accomplished — and I don’t want to pick on a ride hailing company — but when you have those easy wins, it can sort of blind you about the future, and then think that everything you’re doing is right. Like you have a mandate to go forward. And that’s where I get worried about the boundary of innovation. And what does it mean for people in human capital?

Loney: Where do you think you would like to take this path of research, and what’s that next step for you?

Cameron: I’m doing a lot of research in the Global South right now. Spent the summers in Brazil, in Nigeria, in Ghana. And back to that earlier point where I talked about how technology and these algorithmic skills are sort of honed on the most disenfranchised groups — I think the Global South is the future. And I see it in many ways, in thinking about how the gig economy and algorithmic management are evolving. I’m seeing the big changes happening in the Global South, and I see it refined when I’m looking in the U.S. And it’s easy to miss it, because it looks like noise within the data.

And so I’m looking at my research in a more global scale. I’m also thinking about, what are the boundaries of liability? I do not want to point my fingers at these companies and be like, “Tick tick, you’re wrong, big tech.” But I think there are questions about, what are the boundaries of the firm, and what responsibilities do you have if you’re trying to build a marketplace, and if you have independent contractors? We need to think about them in a more thoughtful way. Because this is not Walmart. But nor is it really a bunch of free-floating consultants that are meeting through a job board, through Craigslist. It’s a new organizational form, and it does have some responsibilities and liabilities in play. Accountability.