Skip to main content

Study Ranks Jobs Threatened by Robots—and Offers Robot-Safe Options

worker in meat packing plant
The most resilient alternative job a meat packer could switch to is a textile machine operator, according to the new study. | U.S. Government Accountability Office

First, it was werewolves in the Middle Ages. Then, it was vampires during the Enlightenment. Now, sentient robots are society's boogeymen of choice. But nefarious robots or artificial intelligence (AI) taking over the world is a less pressing concern for economists and policymakers, who find themselves facing a more immediate issue: how to handle robotic automation of jobs and what to do with displaced workers.

A new method featured in the April 13 issue of Science Robotics tackles this problem, presenting a way to evaluate how likely 967 jobs are of being automated. Going beyond previous assessments, the analysis also offers a way to help workers in at-risk jobs find new robot-safe positions – ones that require minimal retraining and use skill sets that robots won't have for a long time.

"Even the simplest job requires a mix of different abilities at different levels, and at least some of them will be beyond reach of machines for quite some time," the Science Robotics authors said in an email.

While physicists are the safest from automation, meat packers are the most in danger. The most resilient alternative job a meat packer could switch to is a textile machine operator, which uses a similar enough skill set that is harder for robots to pick up.

"If we can identify the right skills to develop, humans can remain one step ahead in the race with robots," the authors said.

Calculating Job Risk Little by Little, Byte by Byte

Fear of robotic automation is nothing new. Robots have been replacing jobs in industries that rely heavily on physical labor, such as agriculture and manufacturing, for a while. Now though, AI and robots are becoming better equipped to take on higher-skilled jobs. This complicates job resiliency across diverse sectors including corporate ones .

"Many studies suggest that the next wave of automation by intelligent machines will impact a large range of jobs and will leave many people unemployed," the authors said. "But those studies focused mainly on artificial intelligence software, not on physical robots, and did not provide solutions for the unemployed."

In the new study, Antonio Paolillo, a researcher at Istituto Dalle Molle di studi sull'intelligenza artificiale who was formerly at École Polytechnique Fédérale de Lausanne, and his colleagues decided to address risk from both AI and physical robots as well as provide solutions for those at risk. First, they devised an automation risk index to measure how in danger jobs from across many industries are of being automated. While creating the algorithm, the team used the U.S. Department of Labor's O*Net database and gathered a list of requirements for 967 jobs. They also employed the European 2020 Robotics Multiannual Roadmap to build a list of the tasks that robots can currently do and how well they do them.

Following this, the scientists examined how many requirements in each job description could be performed by a robot versus a human, ranked the importance of these abilities, and accounted for how ready robots are to fill those needs. "This method allowed us to produce for any job listed in the O*NET database an automation risk," the authors said.

Next, they created a resilience index to optimally shift workers from positions deemed higher-risk by the automation risk index to lower-risk options with minimal retraining. They quantified retraining effort by assembling every possible pairing of jobs in the automation risk index, calculating the difference between each pairing's required skills, and determining the importance of each skill. The smaller the difference, the less retraining the worker would need. When tested on data from the 2018 U.S. workforce, the resilience index successfully shifted workers around to place them in their most resilient options. In doing so, it significantly decreased average automation risk across the job market.

"We use 2018 U.S. workforce data because the U.S. has been affected strongly by automation," the authors said. "Naturally, our approach can be extended to any demographic on the planet. The most recent data, from May 2020, shows similar patterns as those from 2018."

To make the results more accessible, the researchers have designed a website where people can look up their position's automation risk and find three robot-safe alternatives for a career shift. For example, the site suggests municipal firefighters looking to be safer from automation could best repurpose their skill set by becoming clinical research coordinators, compliance managers, or fire inspectors in that order. It might seem counterintuitive that fire inspector is the third choice. However, since the position is the least safe from automation of the three, it goes last despite having the most similar skillset.

"Economists could incorporate our indices into more sophisticated models of occupations to generate evidence-based, replicable predictions of the impact of automation," the authors said. "Policymakers could use the automation risk index to identify workers who are more at risk and tailor appropriate welfare policies and could use the resilience index to formulate effective retraining and upskilling programs. Individual workers may use the resilience calculator to assess their own risks and make feasible career transitions with minimal retraining."