Why Remote Work Could Lead to Less Innovation
Kanebridge News
Share Button

Why Remote Work Could Lead to Less Innovation

A new study suggests that when employees from one company run into employees from another company, creative sparks fly

Wed, May 17, 2023 8:36amGrey Clock 3 min

Do chance encounters among employees of different Silicon Valley companies in coffee shops, restaurants and other public places lead to innovation? The answer is yes, say researchers who examined such “knowledge spillovers” in a study that may have implications for today’s work-from-home culture.

The researchers—Keith Chen of the University of California, Los Angeles, and David Atkin and Anton Popov of the Massachusetts Institute of Technology—tracked the locations of 425,000 phones using commercially available cellphone-location data. Though the data is anonymous and linked only to the unique ID number of each phone, the researchers surmised where the phone owners worked by looking at where the phones spent large parts of the workday, using a map of buildings occupied by Silicon Valley companies that have filed patents.

Examining instances where phone owners went outside the office and ended up near someone from another Silicon Valley company, they found 218 million episodes in which two workers from different companies were in the same place between September 2016 and November 2017.

For their study, they considered only situations in which both people were near each other for at least a half-hour, and used a probability technique to eliminate meetings that might have been arranged in advance. They also assumed that many of these people bumped into someone they already knew, such as a former colleague.

Sharing knowledge

Such chance meetings “may spark a conversation that leads to a transfer of knowledge or a collaboration,” the researchers wrote.

Next, the research team pulled up patent applications filed by the companies of the employees. Such applications list relevant patents from other companies in so-called patent citations. Patent citations are “one measure of which firms are influencing each other and how firms are sharing ideas,” says Prof. Chen, who studies behavioral economics and strategy at UCLA’s Anderson School of Management.

The researchers then worked backward in time. They looked for places where employees of a patent-filing company may have crossed paths with workers from companies cited in the patent application.

“We rewind the clock to a year before when they would have been developing this technology,” says Prof. Chen. “What school were they dropping their kids off at, what mall were they shopping at, what bar do they frequent. And you infer who was at that bar when they were there,” based on the phone-location data.

The goal, Prof. Chen says, is “to connect workers of the firm that is going to file the patent, at the establishment where we infer that patent was innovated, with what other workers they were interacting with.”

Next, the researchers calculated the overall number of such citations that appear to have been linked to unplanned encounters. The upshot: The researchers say that without these encounters, there would have been about 8% fewer cross-firm patent citations in the period covered by the phone-location data.

“There is a tremendous correlation between my workers’ meeting a lot with your workers, and my workers’ citing your workers’ patent,” says Prof. Chen.

The innovation boost from the encounters, by the team’s calculations, is about twice as large as a similar effect found by other research that looked for knowledge transfer based on whether two companies’ offices are near each other, Prof. Chen says.

Their study comes with some caveats. The researchers don’t know whether these employees actually spoke when they were in the same location, or, if they spoke, what they talked about. And they don’t know whether the workers’ jobs would have facilitated a tech discussion—they might have involved a Google HR staffer and an Apple maintenance person.

Still, the report shines a light on what some experts have long suspected: that random conversations involving people in similar industries can increase innovation.

Enrico Moretti, an economics professor at the University of California, Berkeley, says the study “significantly advances our understanding of knowledge spillovers and how they shape the geography of innovation.” Prof. Moretti, who says he has been working on the topic for 25 years, says, “I find this paper to be one of the most direct and convincing pieces of evidence on this question. It provides important insights into why Silicon Valley-style clusters of innovation exist.”

Remote work’s impact

Though the study involved cellphone data from before Covid, the researchers say it has implications for an era when many people work all or part of the time from home.

The researchers looked at people who occasionally worked from home in the study period, based on where their phones were located during daytime hours, and then at how that affected their probability of attending planned or serendipitous meetings with someone from another company who didn’t work from home, Prof. Chen says.

Looking at two hypothetical companies, the researchers extrapolated that if one-half of employees at each business work from home, their meetings of all types—serendipitous and planned—would fall 35% and patent citations between the companies would decline almost 12%.

“We think this means information exchange between firms is decreasing,” Prof. Chen says. “It is worrying. These businesses co-locate for a reason. If they can’t learn from each other, we think that is a big deal.”

“Presumably,” he adds, “an even bigger effect is the harm that it does to serendipity and flow of information and innovation within the firm.”


This stylish family home combines a classic palette and finishes with a flexible floorplan

35 North Street Windsor

Just 55 minutes from Sydney, make this your creative getaway located in the majestic Hawkesbury region.

Related Stories
Should AI Have Access to Your Medical Records? What if It Can Save Many Lives?
How an Ex-Teacher Turned a Tiny Pension Into a Giant-Killer
By MATT WIRZ 27/05/2024
The Problem With Behavioural Nudges
By Evan Polman and Sam J. Maglio 27/05/2024
Should AI Have Access to Your Medical Records? What if It Can Save Many Lives?

We asked readers: Is it worth giving up some potential privacy if the public benefit could be great? Here’s what they said.

Tue, May 28, 2024 4 min

We’re constantly told that one of the potentially biggest benefits of artificial intelligence is in the area of health. By collecting large amounts of data, AI can create all sorts of drugs for diseases that have been resistant to treatment.

But the price of that could be that we have to share more of our medical information. After all, researchers can’t collect large amounts of data if people aren’t willing to part with that data.

We wanted to see where our readers stand on the balance of privacy versus public-health gains as part of our series on ethical dilemmas created by the advent of AI.

Here are the questions we posed…

AI may be able to discover new medical treatments if it can scan large volumes of health records. Should our personal health records be made available for this purpose, if it has the potential to improve or save millions of lives? How would we guard privacy in that case?

…and some of the answers we received. undefined

Rely on nonpartisan overseers

While my own recent experience with a data breach highlights the importance of robust data security, I recognise the potential for AI to revolutionise healthcare. To ensure privacy, I would be more comfortable if an independent, nonpartisan body—overseen by medical professionals, data-security experts, and citizen representatives—managed a secure database.

Anonymity cuts both ways

Yes. Simply sanitise the health records of any identifying information, which is quite doable. Although there is an argument to be made that AI may discover something that an individual needs or wants to know.

Executive-level oversight

I think we can make AI scanning of health records available with strict privacy controls. Create an AI-CEO position at medical facilities with extreme vetting of that individual before hiring them.

Well worth it

This actually sounds like a very GOOD use of AI. There are several methods for anonymising data which would allow for studies over massive cross-sections of the population without compromising individuals’ privacy. The AI would just be doing the same things meta-studies do now, only faster and maybe better.

Human touch

My concern is that the next generations of doctors will rely more heavily, maybe exclusively, on AI and lose the ability or even the desire to respect the art of medicine which demands one-on-one interaction with a patient for discussion and examination (already a dying skill).


People should be able to sign over rights to their complete “anonymised” health record upon death just as they can sign over rights to their organs. Waiting for death for such access does temporarily slow down the pace of such research, but ultimately will make the research better. Data sets will be more complete, too. Before signing over such rights, however, a person would have to be fully informed on how their relatives’ privacy may also be affected.

Pay me or make it free for all

As long as this is open-source and free, they can use my records. I have a problem with people using my data to make a profit without compensation.

Privacy above all

As a free society, we value freedoms and privacy, often over greater utilitarian benefits that could come. AI does not get any greater right to infringe on that liberty than anything else does.

Opt-in only

You should be able to opt in and choose a plan that protects your privacy.

Privacy doesn’t exist anyway

If it is decided to extend human lives indefinitely, then by all means, scan all health records. As for privacy, there is no such thing. All databases, once established, will eventually, if not immediately, be accessed or hacked by both the good and bad guys.

The data’s already out there

I think it should be made available. We already sign our rights for information over to large insurance companies. Making health records in the aggregate available for helping AI spot potential ways to improve medical care makes sense to me.

Overarching benefit

Of course they should be made available. Privacy is no serious concern when the benefits are so huge for so many.

Compensation for breakthroughs

We should be given the choice to release our records and compensated if our particular genome creates a pathway to treatment and medications.

Too risky

I like the idea of improving healthcare by accessing health records. However, as great as that potential is, the risks outweigh it. Access to the information would not be controlled. Too many would see personal opportunity in it for personal gain.

Nothing personal

The personal info should never be available to anyone who is not specifically authorised by the patient to have it. Medical information can be used to deny people employment or licenses!

No guarantee, but go ahead

This should be allowed on an anonymous basis, without question. But how to provide that anonymity?

Anonymously isolating the information is probably easy, but that information probably contains enough information to identify you if someone had access to the data and was strongly motivated. So the answer lies in restricting access to the raw data to trusted individuals.

Take my records, please

As a person with multiple medical conditions taking 28 medications a day, I highly endorse the use of my records. It is an area where I have found AI particularly valuable. With no medical educational background, I find it very helpful when AI describes in layman’s terms both my conditions and medications. In one instance, while interpreting a CT scan, AI noted a growth on my kidney that looked suspiciously like cancer and had not been disclosed to me by any of the four doctors examining the chart.


This stylish family home combines a classic palette and finishes with a flexible floorplan

35 North Street Windsor

Just 55 minutes from Sydney, make this your creative getaway located in the majestic Hawkesbury region.

Related Stories
The Problem With Behavioural Nudges
By Evan Polman and Sam J. Maglio 27/05/2024
Boost for World Economy as U.S., Eurozone Accelerate in Tandem
By JOSHUA KIRBY 25/05/2024
The suburbs where we’re building the most new homes
By Bronwyn Allen 28/05/2024
    Your Cart
    Your cart is emptyReturn to Shop