From AI recruiting tools to industrial automation and robotic assistants, new digital technologies are transforming the modern workplace. Many of these systems promise to improve efficiency, productivity, and well-being — but how are they actually affecting the people who interact with them every day?
It’s a complicated question with no cut-and-dry answers. But a growing body of research has begun to explore the nuanced ways in which technology is influencing the workplace and the workforce, shedding light on both its many benefits and substantial risks.
How Is AI Transforming Hiring?
One of the most significant areas in which technology has transformed the workplace is before new candidates even get in the door. AI tools can help recruiters sift through resumes, review cover letters, and even conduct virtual interviews. But these tools can also introduce new complexities and biases into the hiring process.
AI hiring tools can influence who applies: In one study, researchers asked more than 500 U.S.-based adults to imagine applying to a job through a system that used AI. They found that candidates who were already excited about the prospective employer and felt positively about AI in general were more likely to complete an application,. Candidates who were anxious or distrustful of AI, or who were less enthusiastic about the employer, were less likely to complete their applications if interaction with AI was required. This suggests that incorporating automated tools into the hiring process can affect different candidates’ experiences differently, influencing who ends up applying in potentially surprising ways.
Automated screening can perpetuate bias: While the potential for AI-based systems to perpetuate human biases is well-known, a new study found that even when explicitly gendered information (such as names or pronouns) is removed, today’s sophisticated machine learning models can still accurately determine a candidate’s gender. Furthermore, the study found that after controlling for job-relevant traits, when elements of a candidate’s resume did not line up with their gender — i.e., when a woman’s resume included traditionally masculine characteristics — they were less likely to get called back for an interview.
People are less offended by algorithmic than human discrimination: Given the prevalence of AI-driven bias, will companies feel pressure stop using these tools? At least one paper suggests they might not: Through a series of eight studies, researchers found that people tend to get a lot less mad when they learn that an algorithm discriminates than when a human makes the same discriminatory decision, meaning they’re less likely to blame an organization for discrimination if it’s perpetuated by an automated tool.
How Does Digital Monitoring Impact Employees?
Of course, the hiring process is hardly the last time that a new employee is likely to find themselves interacting with a digital system. The last several years have seen an explosion in employee monitoring tools, from keystroke tracking apps to wearable GSP monitors. And while proponents praise these tools’ potential to boost efficiency and transparency, recent research has painted a more nuanced picture.
Electronic monitoring can harm both workers and employers: A team of researchers conducted a meta-analysis of results from more than 50 different academic studies and found that being monitored electronically reduces employees’ job satisfaction and increases their stress levels. They also found that monitoring has no effect on performance, but that it does slightly increase the chances that an employee will engage in counterproductive behaviors, such as working less than expected, wasting resources, or mistreating coworkers and supervisors. This is in line with other recent research suggesting that monitoring workers makes them more likely to break rules, because it reduces their sense of responsibility for their own actions.
Being monitored may also boost engagement: That said, effective monitoring can also have a positive effect. A study that looked at data from more than 200 employees in higher education found that electronic performance monitoring could boost workers’ engagement. This is at least in part because digital tools are often perceived as more fair than traditional monitoring systems, leading workers to identify more strongly with their organizations and thus feel more invested and engaged in their work.
What Is It Like to Work Alongside a Robot?
Beyond simply being monitored by digital tools, employees are also increasingly likely to work with, take advice from, or even be managed by an automated system. On an individual level, research has identified a number of factors that can influence how people react to their new robotic colleagues.
People react better when automated systems feel authentic: When working with an automated tools such as chatbots or a recommendation engines, authenticity is key. Across a series of five studies, researchers found that people react much more positively when a tool is presented in an authentic manner, and in particular, when its human origins are highlighted. Conversely, anthropomorphizing autonomous technologies by giving them human-like qualities actually makes them seem less authentic, worsening people’s experience when interacting with them.
Another study found a similar effect in the context of algorithmic management: If workers are managed by an algorithm (such as the Uber algorithm, which automatically assigns work, gives performance feedback, and makes other supervisory decisions), they are more likely to react angrily to negative feedback if the robotic interface is anthropomorphized. This is because we subconsciously attribute more agency to human-like systems, and as a result, we’re more likely to feel they are “responsible” for giving us negative feedback.
People prefer to take advice from algorithms for certain kinds of decisions: Three other recent studies explored the contexts in which employees are more or less comfortable accepting advice from an automated tool. One paper found that for predictions or estimations, people prefer to take advice from algorithms than from humans — but when it comes to making decisions based on those predictions, people prefer to take advice from humans. Conversely, another series of experiments found that when it comes to delegating decisions over which people really want to retain control, they are often more willing to cede decision-making power to an AI than to a human.
In addition, research suggests that people want to understand why and how an AI makes its decisions. In a field study examining the use of AI diagnostic tools at a major hospital, medical professionals were less likely to incorporate input from an AI if it diverged from humans’ initial judgements without providing a clear reason. But when AI-generated diagnoses were accompanied by an explanation, doctors were much more likely to listen to them.
Workplace automation comes at a cost: Alongside the impact on individual workers’ experiences, rapid growth in automation has also had a substantial effect on macro-level social, political, and economic trends. An analysis of 14 years of U.S. census data cross-referenced with county-level growth in industrial robots found that automation of previously human jobs is associated with increases in drug overdose deaths, suicides, homicides, and cardiovascular mortality rates. Moreover, beyond direct health outcomes, workplace automation can foster negative sentiment in surprising ways. For example, data from more than 30,000 Americans and Europeans suggests that as people become more worried about automation threatening the security of their jobs, they tend to develop more anti-immigrant sentiment.
How Is Automation Changing the Composition of the U.S. Workforce?
While automation certainly has the potential to improve workers’ lives, this common fear — that automation poses a threat to job security — is far from baseless. Indeed, the data suggests that growing investment in AI and other automated technologies may have a substantial impact on the makeup of the workforce.
Automation boosts demand for educated workers and flattens org charts: On a firm-by-firm level, researchers have found that investment into AI tends to correlate with hiring more highly-educated workers. In addition, companies with more automation tend to have flatter organizations, with more junior workers and fewer mid-level and senior employees.
Automation reduces low-wage, non-service jobs: On an economy-wide level, an analysis of U.S. employment data found that increases in automation have led to a decrease in the availability of automatable, low-wage jobs. Interestingly, this shift has been accompanied by an increase in non-automatable low-wage jobs (i.e., service jobs in which humans cannot be replaced by robots). But this increase has not been sufficient to counteract the decrease in jobs that don’t have an interpersonal component. Moreover, the study found that job losses due to automation are largest among non-Asian people of color, shedding light on the complex interplay between racial equity, economic trends, and technological advances.
As with growth in any new technology, the recent explosion in digital workplace tools is neither all good nor all bad. Rather, optimism and enthusiasm for progress must be balanced with an acknowledgement of the very real — and not always positive — impact that these tools can have on the people who interact with them. As new research explores these varied effects, leaders much continuously check their assumptions, avoid oversimplification, and work to ensure that their decisions are driven not by knee-jerk reactions or gut feel alone, but by the latest data and evidence.