AI could trigger a ‘hollowing out’ of the middle class

April 17, 2019
As automation takes on more work done by humans, wages could flatten across the board, warns the University of Bath’s Joanna Bryson.

This article is an excerpt from MIT Technology Review Insights’ upcoming report “Asia’s AI agenda: AI and human capital.”  The research is based on a survey of nearly 900 senior business leaders, conducted in September and October 2018.

While the survey data presents a positive picture overall (almost 60 percent of survey respondents believe that job roles have been enhanced since the introduction of AI), the academic world is paying considerable attention to the downside scenarios. According to AI scientist and researcher Joanna Bryson, associate professor in the department of computer science at the University of Bath, “the disintermediation question is really interesting: people usually think of it as a negative because it implies disintermediating capital from the means of production”—in other words, reducing wages paid to laborers—“but technology-enabled disintermediation actually means getting rid of middle people, and connecting the people who provide a service directly to their customers,” potentially making the quality of services better, and their costs cheaper.

Yet, Bryson notes, the differentiation created by AI and automation technologies “also can lead to a net loss of wage differentiation, and the hollowing out of the middle class.” In the short-term, this could create “redistribution benefits,” such as allowing companies to implement a higher minimum wage. However, in the long term, it could amount to the flattening of wages across the board. She points to a general trend within retail banks. ATMs reduce the headcount per branch, making opening more branches cost-effective—provided that those that staff them are lower-level frontline staff. This allows banks to reduce their more expensive staff—the branch managers. “There are now fewer of them and those that remain are making lower wages because they have fewer people to manage,” she says. “This matters because these people were leaders in their communities, who were able to make, for example, philanthropic local contributions.”

And, says Bryson, in a world where machines are becoming good at “replicating every aspect of human intelligence at super-human levels, and can do a lot of ‘our’ work,” might organizations become less tolerant of real humans’ foibles? “In the quest for efficiency, this might have unintended and severe economic and social consequences: Are we going to micromanage populations and extract messy and difficult people from teams?” she says. In other words, could resource managers begin to view humans as fungible as machine inputs?

AI and automation technologies “can lead to a net loss of wage differentiation, and the hollowing out of the middle class.”

Joanna Bryson, Associate Professor, Department of Computer Science, University of Bath

Scott Park, president and CEO of Doosan Bobcat, a global market leader in compact construction equipment, doesn’t think so. “AI is just another tool in the innovation bucket,” he says. “Throughout the centuries, technology has destroyed jobs and created them. There will be more jobs in the future than there are today due to AI, and more than half of them do not yet exist.” Park sees AI as an opportunity to reshape jobs for the benefit of employees. “How ethical is it to have people doing menial tasks, such as three-way matches on invoices, when we have computers to do that more easily?” he asks.

The first part of this series, “Asia’s AI agenda: The ecosystem,” examines Asian governments’ AI plans. The second part, “AI for business,” explores how organizations are crafting strategies for deploying AI.