To achieve efficiencies and reduce costs, more and more companies are managing their employees by algorithm. In the article by Dr. Armin Granulo, Dr. Sara Caprioli, Professor Christoph Fuchs, former Professor at TUM School of Management, and Professor Stefano Puntoni, the authors present some of the first research findings concerning the effects of algorithmic management on workplace dynamics. One important finding is that employees managed by algorithms are less likely than colleagues managed by people to help others. The authors conclude with suggestions for how companies who want to use algorithmic management can mitigate its negative effects.

Algorithms are being deployed to automate managerial tasks in an increasingly wide variety of industries and settings. Amazon, Uber, and UPS, for example, use them to oversee the movements and performance of millions of drivers and warehouse workers, and 7-Eleven, IBM, and Uniqlo use them to track the sales performance of retail workers or assess employee skillsets.

This shift to “algorithmic management” obviously offers companies enhanced efficiency and profitability. But can it also have unintended consequences and alienating effects, particularly when it comes to workplace dynamics?

Surprisingly few researchers have methodically considered this question, which means we have very little data that can help us answer it. To address that gap, in our research we rigorously test whether the impact of algorithmic management extends beyond workers’ productivity. Most recently, we examined the effects that algorithmic management has on prosocial motivation, which is an important driver of creativity, productivity, social interaction, and overall well-being in the workplace. In doing so we identified one particularly interesting and important gap: Employees who are algorithmically managed turn out to be less inclined to help or support colleagues than employees managed by people.

Companies using algorithmic management need to be mindful of this problem and alert to other negative effects that algorithmic management can have on employee psychology and social dynamics. Fortunately, as we’ll discuss in this article, our research suggests that companies can mitigate these effects by creating a work environment that actively encourages social interactions.

The Problem

We started our research with a field survey of workers in the transportation, distribution, and logistics sectors, where algorithmic management is common. This was the first place we found that workers managed algorithmically are less inclined to help or support colleagues. The trend persisted even after we accounted for factors specific to their organizations (such as size or average employee tenure), their jobs (such as management satisfaction or overall satisfaction), and their personal characteristics (such as gender and income).

We next conducted a field experiment in collaboration with a German van-rental company to directly test the behavioral outcomes of algorithmic management. At the outset of this experiment, we paid approximately 1,000 gig workers from an online labor platform to create slogans for the van-rental company’s social-media marketing campaigns. The workers were randomly divided into two groups, one of which was guided and evaluated by an algorithm, and the other by a person. After the workers had completed the task, we asked them to offer advice to others on how to create effective marketing slogans, and we then measured their willingness to do so.

What we found was remarkable: The workers managed by the algorithm offered roughly 20% less advice to their peers than the workers managed by the person, and the quality of their advice was lower. (Interestingly, the quality of the actual slogans that the two groups came up with did not differ significantly, which suggests that algorithmic management doesn’t necessarily affect workers’ task-based performance.)

The Solution

When we conducted our field survey of workers in the transportation, distribution, and logistics sectors, we found that regular social interactions among workers act as a barrier to the negative effect of algorithmic management. This suggests that companies can actively mitigate adverse effects by promoting an environment where workers can connect and have meaningful exchanges. This could involve initiatives such as providing common break rooms, implementing team rotations, and organizing social events or joint leisure activities.

In another study, we randomly assigned participants to one of two conditions. In one condition, participants read about a work context where the management task described in the scenario was the evaluation of workers’ performance. In the other condition, they read instead about a work context where the management task described in the scenario was scheduling and work planning. We also concurrently manipulated whether the managerial task (evaluation vs. planning) was performed by a human manager alone or by a human manager using an algorithm (which is the way algorithmic management is often implemented).

Interestingly, we only observed a decrease in prosocial motivation with algorithmic management when the focal task was performance evaluation, which told us that algorithmic management doesn’t diminish prosocial motivation uniformly across all management tasks. The negative impact turns out to be particularly pronounced when algorithms are monitoring and evaluating employee performance. Companies need to be mindful of this effect. If they decide that they want to rely on algorithmic management in performance evaluation and other HR-related tasks, they should work to integrate human managers.

But even when human managers are involved, our research shows that the use of algorithms in performance evaluation still risks producing a negative effect on prosocial behavior. In the study just discussed and in another study that directly tested the effect of human involvement, negative effects on prosocial behavior persisted when human managers evaluated employee performance using algorithms.

Anticipating this, companies and managers need to proactively inform and involve employees in decisions regarding the use of algorithmic management. When employees are recognized and included in this way as stakeholders in the design and implementation of algorithmic management, they’re more likely to maintain prosocial behaviors — and less likely to feel objectified.

Companies such as Haier, one of the world’s largest appliance producers, have effectively implemented automated performance evaluation systems by empowering employees to establish their own performance benchmarks beyond the algorithmically determined minimum targets. Furthermore, companies need to ensure transparent and conscientious communication regarding how algorithms are used and who has the final say in the decision-making processes. For instance, IBM incorporates algorithms in their compensation decisions but also clearly communicates to employees that these algorithms provide recommendations that managers can decide to override.

There’s no denying that algorithmic management offers companies many new opportunities to improve how they get their work done. But we’re only beginning to understand the effects that the practice can have on personal well-being, collaborative behaviors, and team dynamics, so companies should exercise significant caution as they start to use it. In particular, they should work actively to mitigate the negative effects that algorithmic management can have on prosocial behavior, given how vital that behavior is more generally to success in the workplace at the individual and the collective levels. There’s a balance to be struck here, and companies will need to work diligently to find it.

How does your research findings contribute to advancing the current understanding or knowledge within your field?

To achieve efficiency gains and reduce costs, more and more companies are managing their employees by algorithms. In this work, we present some of the first research findings concerning the effects of algorithmic management on workplace dynamics. One important finding is that employees managed by algorithms are less likely than colleagues managed by people to help others. This work also shows that this negative effect of algorithmic management on prosocial motivation occurs because it increases the likelihood of objectifying co-workers, and suggests ways how companies can mitigate these negative effects

What are the potential implications or applications of your research for other researchers, practitioners, or policymakers in related fields?

On a practical level, our work shows that companies need to actively mitigate the negative effects of algorithmic management on prosocial motivation, which is an important driver of creativity, productivity, and overall well-being in the workplace. On a broader level, our research points to societal challenges arising from the increasing deployment of algorithms in management tasks.

Can you discuss any potential limitations or areas for further investigation identified in your study that could inspire future research directions within the academic community?

Our results demonstrate that the negative effects of deploying algorithms in management tasks also occur when human decision-makers rely on algorithms. However, our research did not systematically vary the ratio between human and algorithmic input in the evaluation process. Future research may seek to address this limitation through controlled experiments in different management decision contexts (e.g., rewarding vs. punishing employees based on their performance).

To the research paper: https://www.sciencedirect.com/science/article/pii/S0747563223004454 

• Armin Granulo is a postdoctoral researcher at the TUM School of Management in Germany. His work explores the impact of modern technology such as artificial intelligence on society, businesses, and the workforce.

• Sara Caprioli is a postdoctoral researcher at the TUM School of Management in Germany. Her work focuses on the effects of creativity and artificial intelligence on human behavior.

• Christoph Fuchs is a professor of marketing at the University of Vienna in Austria. His research is situated at the interface of marketing, technology, and human behavior.

• Stefano Puntoni is the Sebastian S. Kresge professor of marketing at the Wharton School, where he serves as the co-director of AI at Wharton.  

This Article was pulished in the Harvard Business Review (15 Feb 2024):

https://hbr.org/2024/02/the-social-cost-of-algorithmic-management