In the summer of 2020, Daniel Beunza, a voluble Spanish social scientist who taught at Cass business school in London, organised a stream of video calls with a dozen senior bankers in the US and Europe. Beunza wanted to know how they had run a trading desk while working from home. Did finance require flesh-and-blood humans?
Beunza had studied bank trading floors for two decades, and had noticed a paradox. Digital technologies had entered finance in the late 20th century, pushing markets into cyberspace and enabling most financial work to be done outside the office – in theory. “For $1,400 a month you can have the [Bloomberg] machine at home. You can have the best information, all the data at your disposal,” Beunza was told in 2000 by the head of one Wall Street trading desk, whom he called “Bob”. But the digital revolution had not caused banks’ offices and trading rooms to disappear. “The tendency is the reverse,” Bob said. “Banks are building bigger and bigger trading rooms.”
Why? Beunza had spent years watching financiers like Bob to find the answer. Now, during lockdown, many executives and HR departments found themselves dealing with the same issue: what is gained and what is lost when everyone is working from home? But while most finance companies focused on immediate questions such as whether employees working remotely would have still access to information, feel part of a team and be able to communicate with colleagues, Beunza thought more attention should be paid to different kinds of questions: how do people act as groups? How do they use rituals and symbols to forge a common worldview? To address practical concerns about the costs and benefits of remote working, we first need to understand these deeper issues.
Office workers make decisions not just by using models and manuals or rational, sequential logic – but by pulling in information, as groups, from multiple sources. That is why the rituals, symbols and space matter. “What we do in offices is not usually what people think we do,” Beunza told me. “It is about how we navigate the world.” And these navigation practices are poorly understood by participants like financiers – especially in a digital age.
The engineers who created the internet have always recognised that people and their rituals matter. Since it was founded in 1986, the Internet Engineering Task Force (IETF) has provided a place for people to meet and collectively design the architecture of the web. Its members wanted to make design decisions using “rough consensus”, since they believed the internet should be an egalitarian community where anybody could participate, without hierarchies or coercion. “We reject: kings, presidents and voting. We believe in: rough consensus and running code” was, and still is, one of its key mantras.
To cultivate “rough consensus”, IETF members devised a distinctive ritual: humming. When they needed to make a crucial decision, the group asked everyone to hum to indicate “yay” or “nay” – and proceeded on the basis of which was loudest. The engineers considered this less divisive than voting.
Some of the biggest decisions about how the internet works have been made using this ritual. In March 2018, in a bland room of the Hilton Metropole on London’s Edgware Road, representatives from Google, Intel, Amazon, Qualcomm and others were gathered for an IETF meeting. They were debating a controversial issue: whether or not to adopt the “draft-rhrd-tls-tls13-visibility-01” protocol. To anybody outside the room, it might sound like gobbledegook, but this protocol was important. Measures were being introduced to make it harder for hackers to attack crucial infrastructure such as utility networks, healthcare systems and retail groups. This was a mounting concern at the time – a year or so earlier, hackers seemingly from Russia had shut down the Ukrainian power system. The proposed “visibility” protocol would signal to internet users whether or not anti-hacking tools had been installed.
For an hour the engineers debated the protocol. Some opposed telling users the tools had been installed; others insisted on it. “There are privacy issues,” one said. “It’s about nation states,” another argued. “We cannot do this without consensus.” So a man named Sean Turner – who looked like a garden gnome, with a long, snowy-white beard, bald head, glasses and checked lumberjack shirt – invoked the IETF ritual.
“We are going to hum,” he said. “Please hum now if you support adoption.” A moan rose up, akin to a Tibetan chant, bouncing off the walls of the Metropole. “Thanks. Please hum now if you oppose.” There was a much louder collective hum. “So at this point there is no consensus to adopt this,” Turner declared. The protocol was put on ice.
Most people do not even know that the IETF exists, much less that computer engineers design the web by humming. That is not because the IETF hides its work. On the contrary, its meetings are open to anyone and posted online. But phrases like “draft-rhrd-tls-tls1.3” mean most people instinctively look away, just as they did with derivatives before the 2008 financial crisis. And, as with finance, this lack of external scrutiny – and understanding – is alarming, particularly given the accelerating effects of innovations such as AI. Many of the engineers who build the technologies on which we rely are well-meaning. But they – like financiers – are prone to tunnel vision, and often fail to see that others may not share their mentality. “In a community of technological producers, the very process of designing, crafting, manufacturing and maintaining technology acts as a template and makes technology itself the lens through which the world is seen and defined,” observes Jan English-Lueck, an anthropologist who has studied Silicon Valley.
When the IETF members use humming, they are reflecting and reinforcing a distinctive worldview – their desperate hope that the internet should remain egalitarian and inclusive. That is their creation myth. But they are also signalling that human contact and context matter deeply, even in a world of computing. Humming enables them to collectively demonstrate the power of that idea. It also helps them navigate the currents of shifting opinion in their tribe and make decisions by reading a range of signals.
Humming does not sit easily with the way we imagine technology, but it highlights a crucial truth about how humans navigate the world of work, in offices, online or anywhere else: even if we think we are rational, logical creatures, we make decisions in social groups by absorbing a wide range of signals. And perhaps the best way to understand this is to employ an idea popularised by anthropologists working at companies such as Xerox during the late 20th century, and since used by Beunza and others on Wall Street: “Sense-making”.
One of the first thinkers to develop the concept of sense-making was a man named John Seely Brown. JSB, as he was usually known, was not trained as an anthropologist. He studied maths and physics in the early 60s, and finished a PhD in computer science in 1970, just as the idea of the internet was emerging, and then taught advanced computing science at the University of California, with a particular interest in AI. Around this time, after meeting some sociologists and anthropologists, he became fascinated by the question of how social patterns influence the development of digital tools, too.
He applied for a research post at Xerox’s Palo Alto Research Center (Parc), a research arm that the Connecticut-based company set up in Silicon Valley in 1969. Xerox was famous for developing the photocopier, but it also produced many other digital innovations. The authors of Fumbling the Future, a book about the history of the company, credits it with inventing “the first computer ever designed and built for the dedicated use of a single person … the first graphics-oriented monitor, the first handheld ‘mouse’ simple enough for a child, the first word-processing programme for non-expert users, the first local area communications network … and the first laser printer.”
During his application process to Parc, JSB met Jack Goldman, its chief scientist. The two men discussed Xerox’s research and development work, and its pioneering experiments with AI. Then JSB pointed to Goldman’s desk. “Jack, why two phones?” he asked. The desk contained both a “simple” phone and a newer, more sophisticated model.
“Oh my God, who the hell can use this phone?” Goldman said, referring to the new phone. “I have it on my desk because everyone has to have one, but when real work gets done I’ve got to use a regular one.”
That was exactly the kind of thing, Seely Brown said, that scientists at Xerox should also be researching: how humans were (or were not) using the dazzling innovations that Silicon Valley companies kept creating. Having started steeped in “hard” computing science, JSB realised that it paid to be a “softie” when looking at social science, or – to employ the buzzwords that were later popularised in Silicon Valley by the writer Scott Hartley – to be a techie and a “fuzzy”.
JSB joined Parc and put his new theories to work. Although the research centre had initially been dominated by scientists, by the time JSB arrived, a collection of anthropologists, psychologists and sociologists were also there. One of these anthropologists was a man named Julian Orr, who was studying the “tribe” of technical repair teams at Xerox.
By the late 20th century, copy machines were ubiquitous in offices. Work could collapse if one of these machines broke down. Xerox employed numerous people whose only job was to travel between offices, servicing and fixing machines. These technicians were routinely ignored, partly because the managers assumed that they knew what they did. But Orr and JSB suspected this was a big mistake, and that the technicians did not always think or behave as their bosses thought they should.
JSB first noticed it early in his time at Xerox, when he met a repairman known as “Mr Troubleshooter”, who said to him: “Well, Mr PhD, suppose this photocopier sitting here had an intermittent image quality fault, how would you go about troubleshooting it?”
JSB knew there was an “official” answer in the office handbook: technicians were supposed to “print out 1,000 copies, sort through the output, find a few bad ones, and compare them to the diagnostic”. It sounded logical – to an engineer.
“Here is what I do,” Mr Troubleshooter told JSB, with a “disgusted” look on his face. “I walk to the trash can, tip it upside down, and look at all the copies that have been thrown away. The trash can is a filter – people keep the good copies and throw the bad ones away. So just go to the trash can … and from scanning all the bad ones, interpret what connects them all.” In short, the engineers were ignoring protocols and using a solution that worked – but one that was “invisible … and outside [the] cognitive modelling lens” of the people running the company, JSB ruefully concluded.
How common was this kind of subversive approach? Orr set off to find out. He first enrolled in technical training school. Then he shadowed the repair teams out on service calls, at the parts depot, eating lunch and just hanging out when there was not much work to do. The fact that Orr had once worked as a technician himself helped in some respects: the repair crews welcomed him in. But it also created a trap: he sometimes had the same blind spots as the people he was studying. “I had a tendency to regard certain phenomena as unremarkable which are not really so to outsiders,” he later wrote in a report. He had to perform mental gymnastics to make “familiar” seem “strange”. So, like many other anthropologists before him, he tried to get that sense of distance by looking at the group rituals, symbols and spatial patterns that the technicians used in their everyday life.
Orr quickly realised that many of the most important interactions took place in diners. “I drive to meet the members of the customer support team for breakfast at a chain restaurant in a small city on the east side,” Orr observed in one of his field notes. “Alice has a problem: her machine reports a self-test error, but she suspects there is some other problem … [so] we are going to lunch at a restaurant where many of [Alice’s] colleagues eat, to try to persuade Fred, the most experienced [technician], to go to look at the machine with her … Fred tells her there is another component that she needs to change, according to his interpretation of the logs.” The repair teams were doing collective problem solving over coffee in those diners, using a rich body of shared narrative about the Xerox machines, and almost every other part of their lives. Their “gossip” was weaving a wide tapestry of group knowledge, and tapping into the collective views of the group – like the IETF humming.
This knowledge mattered. The company protocols assumed that “the work of technicians was the rote repair of identical broken machines,” as Lucy Suchman, another anthropologist at Parc, noted. But that was a fallacy: even if the machines seemed identical when they emerged from the Xerox factory, by the time repairmen encountered the machines they had histories shaped by humans. What engineers shared at the diner was this history and context. “Diagnosis is a narrative process,” Orr said.
The Xerox scientists eventually listened to the anthropologists – to some degree. After Orr issued his report on the technicians, the company introduced systems to make it easier for repair people to talk to one another in the field and share knowledge – even outside diners. A two-way radio system allowed tech reps in different regions to call on each other’s expertise. Xerox later supplemented these radios with a rudimentary messaging platform on the internet known as Eureka, where technicians could share tips. JSB viewed this as “an early model for social media platforms”.
Other Silicon Valley entrepreneurs became increasingly fascinated by what Parc was doing, and tried to emulate its ideas. Steve Jobs, a co-founder of Apple, toured Parc in 1979, saw the group’s efforts to build a personal computer, and then developed something similar at Apple, hiring away a key Parc researcher. Other Parc ideas were echoed at Apple and other Silicon Valley companies. But Xerox’s managers were not nearly as adept as Jobs in terms of turning brilliant ideas into lucrative gadgets, and in subsequent decades Xerox’s fortunes ailed. That was partly because the company culture was conservative and slow-moving, but also because Parc was based on the west coast, while the main headquarters and manufacturing centres were on the other side of the country. Good ideas often fell between the cracks, to the frustration of Parc staff.
Still, as the years passed, Parc’s ideas had a big impact on social science and Silicon Valley. Their work helped to spawn the development of the “user experience” (UX) movement, prodding companies such as Microsoft and Intel to create similar teams. Their ideas about “sense-making” spread into the consumer goods world, and from there to an unlikely sphere: Wall Street.
A social scientist named Patricia Ensworth was one of the first to use sense-making in finance. Starting in the 80s, she decided to use social science to help explain why IT issues tended to generate such angst in finance. Her research quickly showed that the issues were social and cultural as much as technical. In one early project she found that American software coders were completely baffled as to why their internally developed software programmes kept malfunctioning – until she explained that office customs in other locations were different. In the early 90s, Ensworth joined Moody’s Investors Service, and eventually became director of quality assurance for its IT systems. It sounded like a technical job. However, her key role was pulling together different tribes – software coders, IT infrastructure technicians, analysts, salespeople and external customers. Then she formed a consultancy to advise on “project management, risk analysis, quality assurance and other business issues”, combining cultural awareness with engineering.
In 2005, Ensworth received an urgent message from a managing director at a major investment bank. “We need a consultant to help us get some projects back on track!” the manager said. Ensworth was used to such appeals: she had spent more than a decade using techniques pioneered by the likes of Orr and Seely Brown in order to study how finance and tech intersected with humans.
The investment bank project was typical. Like many of its rivals, this bank had been racing to move its operations online. But by 2005 it was facing a crisis. Before 2000 it had outsourced much of its trading IT platform to India, since it was cheaper than hiring IT experts in the US. But while the Indian coders and testers were skilled at handling traditional investment products, they struggled to cope with a new derivatives business that the bank was building, since the Indian coders had formal, bureaucratic engineering methods. So the bank started to use other suppliers in Ukraine and Canada who had a more flexible style and were used to collaborating with creative mathematicians. But this made the problems even worse: deadlines were missed, defects emerged and expensive disputes erupted.
“In the New York office, tensions were running high between the onsite employees of rival outsourcing vendors,” Ensworth later wrote. “The pivot point occurred when a fight broke out: a male Canadian tester insulted a female Indian tester with X-rated profanity and she threw hot coffee in his face. Since this legally constituted a workplace assault, the female tester was immediately fired and deported. Debates about the fairness of the punishment divided the office … [and] at the same time auditors uncovered some serious operational and security violations in the outsourced IT infrastructures and processes.”
Many employees blamed the issues on inter-ethnic clashes. But Ensworth suspected another, more subtle problem. Almost all the coders at the bank, whether they were in India, Manhattan, Kyiv or Toronto, had been trained to think in one-directional sequences, driven by sequential logic, without much lateral vision. The binary nature of the software they developed also meant that they tended to have an “I’m-right-you’re-wrong” mentality. Although the coders could produce algorithms to solve specific problems, they struggled to see the whole picture or collaborate to adapt as conditions changed. “The [coders] document their research in the form of use cases, flowcharts and system architecture designs,” Ensworth observed. “These documents work well enough for version 1.0, because the cyberspace model matches the user community’s lived experience. But over time, the model and the reality increasingly diverge.”
The coders often seemed unaware of the gap between their initial plan and subsequent reality. Ensworth persuaded the suppliers in India to provide training about American office rules and customs, and tried to teach the suppliers in Ukraine and Canada about the dangers of taking an excessively freewheeling approach to IT. She showed coders videos of the noisy and chaotic conditions on bank trading floors; that was a shock, since coders typically toiled in library-like silence and calm. She explained to managers at the bank that coders felt angry that they could not access important proprietary databases and tools. The goal was to teach all “sides” to copy the most basic precept of anthropology: seeing the world from another point of view.
Ensworth did not harbour any illusions about changing the bank’s overall culture. When the financial crisis erupted in 2008, the project was wound down and she moved on. However, she was thrilled to see that during the 18 months that she worked at the bank, some of the anthropology lessons stuck. “Delivery schedules and error rates were occasionally troublesome, but no longer a constant, pervasive worry,” she later wrote. Better still, the workers stopped throwing coffee around the office.
But what would happen to the business of sense-making at work if humans were suddenly prevented from working face to face? As he hovered like a fly on the wall of trading rooms on Wall Street and in the City of London in the early 2000s, Beunza often asked himself that question. Then, in the spring of 2020, he was unexpectedly presented with a natural experiment. As Covid-19 spread, financial institutions suddenly did what Bob had said they never would – they sent traders home with their Bloomberg terminals. So, over the course of the summer, Beunza contacted his old Wall Street contacts to ask a key question: what happened?
It was not easy to do the research. Anthropology is a discipline that prizes first-hand observations. Conducting research via video calls seemed to fly in the face of that. “A lot of my work depends on speaking to people face to face, understanding how they live their lives on their own terms and in their own spaces,” said Chloe Evans, an anthropologist at Spotify, to a conference convened in 2020 to discuss the challenge. “Being in the same space is vital for us to understand how people use products and services for the companies we work for.”
However, ethnographers realised there were benefits to the new world, too: they could reach people around the world on a more equal footing, and sometimes with more intimacy. “We see people in contexts not available to us in lab situations,” observed an ethnographer named Stuart Henshall, who was doing research among poor communities in India. Before the pandemic, most of the Indian people he interviewed were so ashamed of their domestic spaces that they preferred to meet in a research office, he explained. But after lockdown, his interviewees started talking to him via video calls from their homes and rickshaws, which enabled him to gain insight into a whole new aspect of their lives. “Participants are simply more comfortable at home in their environment. They feel more in control,” he observed. It was a new of type of ethnography.
When Beunza interviewed bankers remotely, he found echoes of this pattern: respondents were more eager to engage with him from home than in the office, and it felt more intimate. The financiers told him that they had found it relatively simple to do some parts of their job remotely, at least in the short term: working from home was easy if you were writing computer code or scanning legal documents. Teams that had already been working together for a long time also could interact well through video links.
The really big problem was incidental information exchange. “The bit that’s very hard to replicate is the information you didn’t know you needed,” observed Charles Bristow, a senior trader at JP Morgan. “[It’s] where you hear some noise from a desk a corridor away, or you hear a word that triggers a thought. If you’re working from home, you don’t know that you need that information.” Working from home also made it hard to teach younger bankers how to think and behave; physical experiences were crucial for conveying the habits of finance or being an apprentice.
Beunza was not surprised to hear that the financiers were eager to get traders back to the office as soon as they could; nor that most had quietly kept some teams working in the office throughout the crisis. Nor was he surprised that when banks such as JPMorgan started to bring some people back in – initially at 50% capacity – they spent a huge amount of time devising systems to “rotate” people; the trick seemed not to be bringing in entire teams, but people from different groups. This was the best way to get that all-important incidental information exchange when the office was half-full.
But one of the most revealing details from Beunza’s interviews concerned performance. When he asked the financiers at the biggest American and European banks how they had fared during the wild market turmoil of spring 2020, “the bankers said that their trading teams in the office did much, much better than those at home,” Beunza told me in the autumn of 2020. “The Wall Street banks kept more teams in the office, so they seem to have done a lot better than Europeans.” That may have been due to malfunctions on home-based tech platforms. But Beunza attributed it to something else: in-person teams had more incidental information exchange and sense-making, and at times of stress this seemed doubly important.
The bankers that Beunza observed were not the only ones to realise the value of being together in the same physical space. The same pattern was playing out at the IETF. When the pandemic hit, the IETF organisers decided to replace in-person conventions with virtual summits. A few months later they polled about 600 members to see how they felt about this switch. More than half said they considered online meetings less productive than in-person, and only 7% preferred meeting online. Again, they missed the peripheral vision and incidental information exchange that happened with in-person meetings. “[Online] doesn’t work. In person is NOT just about the meeting sessions – it is about meeting people outside the meetings, at social events,” complained one member. “The lack of serendipitous meetings and chats is a significant difference,” said another. Or as one of them put it: “We need to meet in person to get meaningful work done.”
They also missed their humming rituals. As the meetings moved online, two-thirds of the respondents said they wanted to explore new ways to create rough consensus. “We need to figure out how to ‘hum’ online,” said one member. So the IETF organisers experimented with holding online polls. But members complained that virtual polls were too crude and one-dimensional; they crave a more nuanced, three-dimensional way to judge the mood of their tribe. “The most important thing to me about a hum is some idea of how many people present hummed at all, or how loudly. Exact numbers don’t matter, proportionality does,” said one.
A Silicon Valley veteran might have described this as a case of techies craving some “fuzzy” connections. Beunza, Orr and Ensworth would have depicted this as a search for “sense-making” tools. Either way, it was a sign that even in a digital world, humans need those physical connections that we all too often ignore.
Adapted from Anthro-Vision: How Anthropology Can Explain Business and Life by Gillian Tett, which will be published by Random House Business on 8 June. To order a copy, go to the Guardian bookshop.