Why AI is making us treat people worse at work

Will artificial intelligence take over our jobs? Maybe. But what is for certain is that AI isn’t just changing the way we work, but how we work with each other, too.

A new study from the London School of Economics, published in the Journal of Consumer Psychology, suggests that the more human-like AI becomes, the more likely we are to treat other people as if they were machines.

Across a series of psychological studies, researchers found that when people perceived AI-powered robots and virtual assistants as having human qualities, they began to see humans through the same lens they used for AI. The consequence was greater acceptance of dehumanising attitudes and workplace practices.

In one study, people exposed to emotionally intelligent AI were less likely to avoid companies with a reputation for mistreating employees, suggesting they had become more accepting of poor treatment of workers.

Another experiment found that participants who watched robots display emotional behaviours, like dancing, were more supportive of toxic workplace behaviours. This included replacing meals with nutritional shakes, or housing workers in capsule dormitories.

Of course, meal replacements aren't inherently problematic – but the mindset they reflect can be. If efficiency is all that matters, why stop for a rest and lunch with a colleague when you can drink a bottle of flavoured paste at your desk?

“As AI seems emotionally and socially humanlike, people may start treating real humans more like machines, less deserving of care and respect,” says lead researcher Dr Hye-young Kim, assistant professor of marketing at the London School of Economics. “As AI becomes more humanlike through seemingly neutral technological advances, it can quietly reshape how we see and treat one another.”

The social costs of generative AI

The problem is that the way we interact with AI chatbots, like ChatGPT, doesn’t necessarily stay with AI, says Fabian Stephany, a research lecturer in AI & work at the Oxford Internet Institute, University of Oxford.

“It can spill over into how we treat the people around us,” he says. “One possible explanation is that AI makes work interactions more transactional. When we ask a chatbot for help, we don't make small talk, read facial expressions or acknowledge another person's time and emotions, we simply issue a request and receive a response. Those seemingly trivial social rituals are actually what build trust, empathy and respect between colleagues.”

(Pexels)

AI becoming more human, or rather imitating humans more realistically, is one thing. But another emerging problem is humans begin interacting with each other more like machines. “While the evidence is still emerging, it suggests AI may not only change how we perform work, but also how we relate to the people we work with,” says Stephany.

Many of us say ‘please’ and ‘thank you’ to ChatGPT because it feels polite.Yet in our interactions with real people, especially online, those courtesies can be the first thing to disappear. According to a survey of 900 business leaders, 31% admitted to being less polite to humans since using AI. 

Part of the problem is that AI is intensifying work and stressing us out. In an eight-month study of generative AI usage at a US technology company, the 200 employees were found to work faster, take on more tasks and work longer hours. But the results were not what the employer had hoped for: employees burned out more quickly, and the resulting fatigue led to poorer decisions.

“People can send more emails, produce more reports and respond more quickly, but that also creates more demands and higher expectations,” says Stephany. “Under that pressure, the informal conversations, mentoring and relationship-building that make organisations healthy can easily disappear.”

"As AI seems emotionally and socially humanlike, people may start treating real humans more like machines, less deserving of care and respect"

Dr Hye-young Kim, assistant professor of marketing at the London School of Economics

Women could end up paying the social price of AI at work

Additionally, women may bear a disproportionate share of the social costs as generative AI reshapes workplace communication. Research has consistently shown that men are more likely than women to use generative AI tools at work. They are also more likely to report that using these tools has reduced their interactions with colleagues: while 32% of workers overall say they speak to co-workers less since adopting generative AI, the figure rises to 36% for men, compared with 28% for women.

This matters because workplace communication is collective. If the group that uses generative AI most also withdraws more from day-to-day interactions, it changes communication norms for everyone. Women may therefore experience a workplace that is less collaborative, less connected and more impersonal, even though they are less likely to be driving that shift themselves.

“Our recent research finds that women tend to have greater concerns about AI's impacts on health, privacy, climate and employment,” says Stephany.

“If women are both more hesitant to adopt AI because of legitimate concerns and judged more negatively when they do use it, these disadvantages can reinforce one another. The goal shouldn't simply be to get everyone to use more AI. It should be to build AI that deserves people's trust.”

It’s not all doom and gloom, though. The more embedded AI becomes in our lives, the more we value innately-human social skills, like teamwork, leadership and ethical judgement. “Our research suggests a more optimistic story,” says Stephany.

“Rather than accepting increasingly transactional workplaces, organisations should actively invest in these social capabilities. They are becoming a competitive advantage in the age of AI.”