Robots at the Helm: The Hidden Cost of Algorithmic Management

How would you feel about being managed by a machine?

Advances in AI and machine learning algorithms that can collect, interpret and generate data are happening fast and with ubiquity. What happens when organizations automate managerial processes in pursuit of productivity gains?  

AI management systems are already well-established in the gig economy. Uber and Lyft have operational processes that are essentially human-free, with algorithms in the place of managers — assigning rides, determining routes, coordinating payment and even providing feedback on performance.

As human-mimicking technologies increasingly pervade other parts of the economy, taking over similar managerial roles, it bears asking: How will the broader workforce experience this change? How might having an AI “boss” make people feel about themselves, their jobs and their status within the organizations that employ them?

Robots and Human Costs

Not positively, says Darden Professor Roshni Raveendhran.

In a recent study, she and colleagues ran experiments to gauge how AI-based algorithmic management might influence people’s perceptions and forecasts of status. What they found across the board is that under AI management (relative to prototypical human management), people feel worse about their own status. They feel that their standing is diminished and that their work will be seen as “simple,” rote or less demanding. They also confer lower status to others under algorithmic management. And this happens even when the actual experience of being managed by AI can in itself be positive.

These are insights with critical ramifications for organizations.

People are less likely to trust the “machine,” and by extension, the organization, says Raveendhran. Interestingly, however, after actually working with a competent algorithm, their trust in the algorithm increased.

“The productivity benefits of AI are huge, and it’s inevitable that organizations will look to machines to automate certain roles, including administrative responsibilities,” Raveendhran says. “It’s important to anticipate how human workers are likely to respond emotionally and psychologically if machines start taking over managerial tasks. So, we wanted to identify the psychological implications for the workforce that organizations are going to have to navigate in the future.”

Funky Coffee and Working for the Machine

Raveendhran and her co-authors ran five experiments with participants from a range of professional backgrounds. Participants were surveyed to understand how they would feel about being managed by algorithms and how this would compare to being managed by humans. They were also asked how they believed others might judge their work under algorithmic management and — critically — if they would see other colleagues’ work as being less complex or rote, knowing that they had an AI “boss.”

In one experiment, Raveendhran and her colleagues simulated a work task. They built their own proprietary algorithm to manage and assess a group marketing activity, and then invited participants to reflect on the experience.

“For our fifth study, we created our own algorithm — this was actually before ChatGPT came out — and had it instruct people to brainstorm and come up with a funky name for a coffee shop. The algorithm then gave each person a score based on their idea,” says Raveendhran. “One group was told they’d be managed by a human being, the other that they would be managed by AI.”

Although the process for participants was identical, people’s responses were markedly different. Asked about status, those who were told they’d been managed by the AI felt significantly less positive about themselves and their work. This group also felt they were treated with less politeness, dignity and respect, as well as felt less trusting both of the algorithm and the organization deploying it than did the group that believed they had interacted with a human manager.

“What’s interesting is that both groups actually reported having an overall positive experience with the algorithm, so reality doesn’t match the psychological effects we are seeing,” says Raveendhran.

This could yield more insights for organizations looking to preempt or attenuate human pushback to new technologies in the workplace.

Focusing on the ‘Why’ of AI in the Workplace

Countering negative perceptions around status and job complexity comes down to articulating why AI is used to perform managerial functions, say Raveendhran. Being upfront with people about why and how you intend to integrate autonomous systems can help. But importantly, Raveendhran and colleagues emphasize that dispelling the notion that AI is only for managing basic, low-skill tasks is key to reducing feelings of status loss.

“Effectively leveraging AI to perform social roles like managing others — without harming the well-being and motivation of the workforce — will boil down to thinking carefully about mitigating people’s negative perceptions about their own status and empowering them to feel good about themselves and their jobs,” says Raveendhran.

“AI tools are used in industries with highly complex jobs,” she says. “They can serve as tools to reduce workload and free up people to do more creative and complex work, wherever they sit within the organizational hierarchy. So, it’s about flipping perceptions and really emphasizing that this is also a tool for highly complex jobs. ”

Organizational productivity and human well-being don’t have to be mutually exclusive. In fact, the changes machine management will bring can be beneficial overall — with careful communication from leadership.

Roshni Raveendhran co-authored “Algorithmic Management Diminishes Status: An Unintended Consequence of Using Machines to Perform Social Roles,” which appeared in the Journal of Experimental Social Psychology, with Arthur S. Jago of the University of Washington at Tacoma and Nathanael Fast and Jonathan Grach, both of the University of Southern California.