Digital Employment Technologies, AI Under Scrutiny
How we find jobs and how we work at them are just a two areas of life totally transformed by AI and digital platforms since the pandemic. Experts debate the best — and worst-case — scenarios.
By Peter Krass
Weak ties still matter for job seekers — at least, in some industries. Artificial intelligence needs to be designed much more deliberately. And remote work requires re-imagined forms of management. Most of all, current social media and AI are at a tipping point where business and society have to make sense of the super-fast pace of change.
These were some of the insights shared at ThinkerFest 2023 on February 23, virtually attended by more than 850 people worldwide.
ThinkerFest 2023 was a unique collaboration by MIT Initiative on the Digital Economy (IDE), which examines the digital economy, and Thinkers50, a UK-based group that identifies, ranks, and shares management ideas that can make a positive difference in the world. (Read the related story here.)
Fixing the ‘Splinternet’
The day’s first session examined the way online misinformation and fake news has splintered society into numerous, often hostile factions — even though the internet was originally designed to bring us together. What can be done to fix the “splinternet?
The internet’s ubiquity is creating extreme, conspiracy-theory communities, warned Rahaf Harfoush, executive director of the Red Thread Institute of Digital Culture. In earlier times, she explained, these groups were few, separate, and isolated. Now they are forming larger ecosystems with what Harfoush called “super-mutant conspiracies” that have more clout and influence, especially when amplified on social media.
Panelist Martin Lindstrom raised the alarm even higher. The founder and chairman of Lindstrom Co., which helps clients reconnect their brands with customers, Lindstrom pointed out that the consequences of social media already have been “pretty severe,” including fake news and challenges to democracy and mental health.
“I’m concerned,” Lindstrom said. “We should have asked fundamental questions when social media kicked off.” It’s not a minute too soon to ask those questions now about the metaverse and our online future, he said.
Boston University professor, Marshall Van Alstyne, offered a more optimistic view. He argued that empowering users in a decentralized environment could help fix the “splinternet” without infringing on constitutional rights. One right would be to receive information of your own choosing, and the second is the right to be heard and to influence the decisions that affect you.
To balance contradictions between these two rights, Van Alstyne proposed one more; an incentive for posting truths: “You can express yourself in any way you wish, facts or opinions. But if you express yourself with facts,” your message will be inserted into the information stream of people that don’t want to hear it as a way of breaking down information silos. “In addition, you’ll be able to warrant that it’s true.”
Weak Ties: A Mixed Bag for Job Seekers
In the event’s keynote, MIT IDE director Sinan Aral, offered insights from his team’s latest research updating a seminal 1973 report that has dramatically changed how people think about social networks. That paper, The Strength of Weak Ties, was written by sociologist Mark Granovetter when analog systems ruled.
One of Granovetter’s most surprising ideas was that job seekers should turn for help not to their closest friends and colleagues, but instead to their “weak ties” — people they know, but not all that well.
In the 50 years since, labor and employment — in particular, job recruiting and job search — have been completely transformed. Paper resumes, word-of-mouth and personal interviews have been largely replaced by digital resume screeners and their algorithms. “It’s a completely different process,” Aral said. “It’s extremely digitally mediated.” That matters, he said, because
who you know drives your job opportunities as well as your social mobility. And increasingly, who you know is influenced by algorithms.
There are broader implications, as well. Aral pointed out that labor turbulence has wracked the recent economy. The year 2020 saw the biggest unemployment spike in nearly 100 years, a shock that was felt worldwide (See figure 1 below):
Just as disruptive as the shocking level of unemployment was the surge of employment post-pandemic in 2022, and the strong market so far this year. (See figure 2 below).
Aral — along with co-authors from MIT, Harvard, Stanford, and LinkedIn — conducted an experimental study involving some 20 million people worldwide and the connections they made over a period of five years to see how well the original Weak Ties theory holds up. Their findings, as reported in a recent Science journal article, concludes that it all depends.
“Weak ties are more important for the most digital industries,” Aral explained during his ThinkerFest keynote. “But in more analog industries, people now benefit more from strong ties.” Moreover, they found that by designing algorithms properly, developers could actually drive employment and equality.
Better design of AI is top-of-mind for others at the IDE, too. Human-First AI was the theme in a chat between Renée Richardson Gosline, senior lecturer at the MIT Sloan School of Management and a research group lead at MIT IDE, and Sanjeev Vohra, senior managing director & global lead, applied intelligence, at Accenture.
Gosline and Vohra discussed the implications of ChatGPT, the AI-powered chatbot launched by Open AI late last year. Both speakers said they’ve tried the technology, though for Sanjeev and his clients, using data for the right use cases is the priority.
“How do I get value out of AI, while keeping humans at the center?” Sanjeev asked. “As a company, we’re always asking how technology and human engineering can create value together.”
Gosline had additional concerns. For one, while ChatGPT relies on information published online, it has no way of verifying whether the information it uses is accurate and true. “I ask my students to find the gaps between what they know and what the AI says,” Gosline told the audience.
She also sees that in the quest for greater efficiency, AI may be reducing friction too much. “ChatGPT is collecting content frictionlessly,” Gosline said, echoing points she made recently at the CMO Summit@MIT. “But friction itself is not bad. Too much is bad, but so is too little.”
If that sounds obvious, it’s not. Sanjeev and his consulting colleagues at Accenture recently polled C-suite executives in nearly 20 industries on AI topics. One question was, “Have you thought about using data responsibly?” They found that most do not.
Managing Remote Work
Another megatrend discussed at the event was the huge shift to working from home and other remote locations since the pandemic. What started as a temporary response to COVID now appears to be a permanent change.
Jerry Carter, a VP of engineering at Dell Technologies, offered his perspective as someone who’s been working remotely for decades. He said the quality of remote work has actually improved in the last two years.
If anything, working from the company office has become the exception. “There’s no critical mass [of employees] there,” Carter said.
Another panelist, John Horton, agreed, saying that remote collaboration requires new ways of getting work done. Horton, an associate professor of information technologies at the MIT Sloan School of Management, as well as a research group lead at MIT IDE, called this “a more asynchronous way of production. How do you get things done without bottlenecks?”
Meanwhile, Dartmouth professor and MIT IDE visiting scholar Geoff Parker, cautioned that while standardized tasks are relatively easy to manage remotely, more creative work requires more team collaboration. “How well codified is the work?” he asked. “Is it standardized or ambiguous?”
Looking ahead, Carter predicted that the best remote leaders will be cross-functional: one part manager, one part organizational psychologist, and one part engineer. And, he added, they’ll need a willingness to spend more time on planes.
Wrapping up the day in a closing session, Aral said that digital platforms are an opportunity to study human relationships and how economies evolve, but they’re also their own socio-technical systems.
“We are designers of them, we are students of them, and we are participants in them,” he said. Importantly, “we also have a responsibility to understand how they shape the digital economy, and how they shape digital society.”