For AI to be embraced in the workplace, it must be developed and deployed in partnership with workers.
While consumers are readily embracing automation technologies built into device apps, smart speakers and connected-home systems, the mood in the workplace is decidedly more mixed. For members of the Service Employees International Union (SEIU), the trepidation about artificial intelligence is more a matter of exclusion than the prospect of having to adapt to the technology.
Fast-food workers, for instance, have had “zero access or opportunity to have their minds used in the design” of AI-powered systems, SEIU President Mary Kay Henry said recently at the second meeting of the California Future of Work Commission (of which she is co-chair), held at the Stanford d.School, formally known as the Hasso Plattner Institute of Design. “For working people, if there was a trust that government, employers and academics would welcome our minds into it, we could see a big explosion” in the adoption of AI.
The commission’s theme for the day, Technological Change and its Impact on Work, encompassed theoretical perspectives on AI and robotics and presentations of industry cases around how technology is changing specific types of work, such as trucking and advanced manufacturing enabled by automation and 3D printing, and what evolving organizational structures mean for the future of white-collar jobs. (The commission is a public-private partnership between the California Labor and Workforce Development Agency and Institute for the Future. See HAI’s first blog post on the Commission.)
Moving AI technology in the direction of complementing humans, rather than replacing them, is a key priority for Stanford’s Institute for Human-Centered Artificial Intelligence (Stanford HAI), where Susan Athey, the Economics of Technology Professor at Stanford Graduate School of Business (GSB), is associate director. (Fei-Fei Li, a member of the Future of Work Commission, is co-director of Stanford HAI.)
With AI still in the early stages of development, Athey told the commission, “our universities and research programs have outsize impact because if they do the right basic R&D, that lowers the cost of adoption across the board.
“So, I believe we can realistically change the trajectory of research to be more in this augmenting direction. That would make it more likely that companies would also move in that direction,” becoming more attentive to the human context in which AI-powered products and services are delivered.
With an eye toward social welfare, Athey recently founded the Golub Capital Social Impact Lab at Stanford GSB. The lab offers collaborators, including social impact organizations and government agencies, access to expertise in the areas of AI, machine learning, data and analytics, behavioral science, and incentive design. It aims to guide innovation and identify insights and best-practice methodologies that can be tailored to specific problems and challenges, such as personalized learning.
“If we can crack the nut of making training more personalized in a variety of ways,” Athey said, “that can be spread and duplicated and scaled” across diverse populations. Such advances would help people learn not just technical skills but also expectations for communication in work environments that might be less familiar to them.
The ubiquity of mobile devices means that “we can meet people where they are: on the bus, the train, while they’re sitting on the floor of a dark room while their children are not quite asleep. The education and training need to be delivered in a bite-sized, engaging way,” she added. “These are new opportunities that have been opened up for us.”
A more flexible labor market mediated by computation poses both an opportunity and a challenge. New digital business models, marketplaces and platforms can allow people to launch their own careers as small business owners — and not just in ecommerce, Athey observed. Electricians, interior designers and practitioners of other trades not usually associated with digitalization can access tools to overcome impediments to self-employment, whether it’s scheduling or learning how to advertise online, managing invoicing and financial reporting, or streamlining communications with customers and vendors.
But while these technologies enable a more flexible workforce, they also contribute to the fissuring or disintermediation of the labor economy. Technology makes it easy for corporations to hire temporary workers and outsource tasks once performed by employees. Subcontractors and freelancers in the gig economy, instead of working for organizations that provide health insurance and other benefits, are left to fend for themselves, relying on platforms for employment.
As Melissa Valentine explained to the commission, disintermediation is impacting white-collar workers because of how organizations are utilizing technology. While discussion of automation tends to focus on specific occupations or its impact on the broader labor economy, Valentine — an assistant professor in Stanford’s Management Science and Engineering Department and co-director of the university’s Center for Work, Technology and Organization — studies how advanced analytics, crowdsourcing models and other technologies are not only changing the nature of work and how decisions are made within organizations but how organizations themselves are structured.
Traditionally in corporations, she related, sets of decisions were divvied up and assigned to people based on their place within an organizational chart. Now, data science and Application Programming Interface (API) technologies are eclipsing organizational charts and automating decision making.
For data scientists, Valentine said, half in jest, having employees make decisions within organizational boxes is practically a source of “psychological pain.” Better, in their view, to gobble up gobs of data and have an algorithm explore the entire space of possible solutions across an organization. Meanwhile, instead of having white-collar workers perform administrative coordination, APIs connect and synchronize different components or systems within a company — automating, for example, the exchange of information between a “front end” that interacts with customers and a “back end” that manages inputs and stock.
The result: Not only are tasks becoming unbundled from jobs, often in ways that are hard to predict, but people are unbundling from organizational structures and traditional safety nets. They might be spun onto temporary projects, then spun off again, no longer part of a stable work collective.
From the perspective of a tech entrepreneur, this new paradigm of software enabling flexible work is exciting and, in theory, a net positive for workers in aggregate. But to help the commissioners understand the policy ramifications, Valentine shared a metaphor from sociologist Jerry Davis at the University of Michigan. He likens business components in the era of disintermediation to Lego blocks scattered about the landscape. One can snap them together, reconfigure them and scale-up an organization by adding different pieces on demand. But from the point of view of labor, if you are a Lego block, what if you no longer fit into a reconfigured structure or your color has gone out of fashion and you need new skills to do a different kind of labor?
Policymakers must grapple with this new reality as workers become untethered from traditional organizations and no longer have access to human resources departments and employer-provided benefits. Meantime, advised Valentine, speaking of the research community and those who develop and deploy new technology, “we need to keep close track of what the experience is like” for workers trying to navigate workplaces in which old verities no longer apply.