Friend of the Lab: Andrew Eland
Claire Fram sat down with Andrew Eland to talk about data, the built environment, and trying to build software responsibly.
Claire Fram: You have worked in ‘big data’ since before ‘big data’ was cool- and then an overused buzz phrase. Leading the development of Google Maps before we all had smartphones in our pockets and you had to use a desktop computer to access Google Maps. You then worked at Google.org before joining DeepMind in 2016. At DeepMind you were an engineering director of their health team — a major program for DeepMind, and a minefield of data ethics questions.
In 2019 you left DeepMind to forge a new project. Your focus has been on sustainability and finding imaginative ways to make data use in the built environment transparent. As part of your current work you have been collaborating with the small but influential, London-based data ethics design firm, Projects by If and the global engineering and design firm, Arup. We have been humbled to work with you as part of the City Modelling Lab.
Could you tell us more about why the built environment is an important area of focus for you? Are there particular experiences you that led to focus on the built environment?
Andrew Eland:
Cities are a complex, elegant, magical ballet
As a software engineer you spend an awful lot of time trying to make sense of complex systems. You’re either designing something to interact with a complex system, or more likely, both the system that you’re building and the team behind it has evolved into its own complex system that you’re trying to maintain and bring a semblance of order to. When approaching how to design software, and those systems around software, I’ve always drawn inspiration from the complex systems of cities. When you sit in a well designed public space, you can see that the design actually encourages an intricate ballet of many, diverse groups of people using space in very, very different ways. It’s hard not to be inspired by the elegance of spaces like that.
I also love the sense of adventure and exploration that cities bring. I guess that’s why I loved working on Google Maps. In some ways, it felt like I was working in that physical space of cities. So using the skills that I have, and as a software engineer, I try to find ways that I can contribute back to the magic that is the city.
Cities are part of the solution
Cities are an important lever to help alleviate a lot of the challenges society faces. If you look at health, for example, we know that access to fresh produce and safe public space is linked closely with health outcomes. I am very proud of the work we did while I was at DeepMind, creating Streams. We made a measurable impact on the quality of care people received when progressing through The Royal Free [the health care system where Streams was in use.] But when you step back and look at the problem systemically, an app is not the solution for better health outcomes. You need to keep people healthy, and keep people out of hospitals to begin with. For many people dealing with chronic conditions, and multiple comorbidities, it’s too late when they end up in hospital. We have let cities grow in ways that have contributed to significant public health issues. For example, when the only spaces teenagers feel safe in are fast food restaurants, that has important health consequences for our communities.
Our technology industry is at a critical juncture
Society is, rightly, asking about the power that the technology sector currently holds. I see really striking parallels between where our technology industry is today, and the rise of post-war modernism, where people were left out of the process. Le Corbusier during WWII convened a small group of privileged white men together to define their utopia: the Athens Charter. Their vision was of elegant, concrete monoliths and rolling parkland. This vision had a huge, lasting influence on Brutalist housing that you still see to this day. But people hate living in these spaces. The “utopia” for the authors of the Athens Charter didn’t work for people. It took activists like Jane Jacobs and Jan Gehl to bring people, and even just human scale, back into urban design and the built environment.
As we see the technology sector grow in power and influence, we not only run the risk of technology setting out a vision of utopia that doesn’t provide people with services that they actually want or need- but there could be serious consequences for all of the industries that intersect with technology. Planners could find themselves in a position where really important decisions are delegated to technology and tools, rather than the people they are trying to serve.
CF: As we think about this juncture for the technology industry, what frameworks, people, or ideas do you return to to guide your process?
AE: The two people who have influenced my thinking most in this space are Richard Pope, especially his talk software is politics. And Bianca Wylie, who brings our attention to what happens when we don’t acknowledge how we delegate decisions to technology.
I also think about where we are with user centred design in technology. We have seen a tremendous shift in the past few decades, from early user interfaces (UIs) to what we have today. But I think we have gone too far.
User centred design is encouraging us to focus on individualism, resulting in filter bubbles. We need to be stepping back and focusing on how to design for groups of people, rather than just the individual. Especially when we are talking about data in the built environment. For example, if I’m sharing data about my travel, then I’m also sharing data about my children when I bring them to school. My friends at IF have been thinking a lot about this, and created the Manifesto for Society Centered Design. Working towards better systems and designs that redistribute power is really exciting to me.
CF: The UI of technology has become a very polished barrier between the user and what is under the hood. As we build machine learning models, and increasingly rely on data for decision making, our interaction with the model outputs may be via those polished UIs. That can make it difficult to understand how the model was constructed and what data the model uses. Whose work do you follow, when thinking about how we make our models more explainable and transparent?
AE: There has been some really nice work in this space. Google’s People + AI have been thinking about ways to explain models, operations, and biases. IF have a great design catalogue of data patterns for how to bring transparency into a system. Sidewalk Labs have done some really nice work around advancing digital transparency in the public realm.
There is lots of encouraging effort in explaining how these models and data systems work. But I also feel that we are barely scratching the surface of what we need to do. If you think of these systems as a hammer, our current efforts are just describing how the hammer works. I think society is not interested in how the hammer works, but what happens when you hit something with that hammer?
If we look at explainability in the built environment, we might learn from the example of early highway and traffic models: decades before machine learning models would have been considered, traffic modellers realised that highway traffic could be modeled like a fluid. This was tremendously useful for planning highways as a result. It was also very easy to explain. But the consequence of easy-to-use, easy to understand highway planning models was that highways were also very easy to build.
Today, as cities are trying to encourage more walking and cycling- it is considered more difficult to understand the consequences of investing in new walking or cycling infrastructure because we don’t have a comparably clean way of understanding those consequences yet. It is more complex to model the impact of cycling and walking infrastructure, compared with modelling a fluid. And that might mean that the models are, in turn, more complex to explain.
I wonder if we need more tools that allow us to step outside of common understandings for how a system is supposed to work. Tools that help show us what else is possible.
CF: You are a friend and advisor to the City Modelling Lab. What made you interested in working with the Lab?
We need more exploration in our approaches to helping people understand the ramifications of changes to complex systems in the built environment. The tools that we currently use are not particularly accessible to most people. It is critical that we find better ways to explain the complex consequences of proposed changes in the built environment, so that important decisions reflect the concerns and priorities of society.
Agent based modelling presents one approach that I think is promising. There are particular questions that ABMs may be very helpful in answering. Like, how do we move away from a world that is purely car centric? How do we redress the balance in terms of evidence and planning? How do we explore and explain the impact of changes, in a more informed way? Allowing people to participate in decisions about the kind of future they want for the areas that they live in.
We need a lot of tools in this space of participatory decision making and transparent processes. Agent based modelling is one category of tool that has a lot of promise.
CF: What’s your advice for the City Modelling Lab?
AE: You have to start somewhere. There is no project that I’ve worked on that ended up in exactly the same place as you envisioned at the start. The way you get to a point of having impact is not by creating a perfect, crystalline vision. You have to launch. It is a long, painful, and torturous journey of trying out things that don’t work and then learning from those things. Pick a place to start, and move forward in a responsible way. That’s the only way I know how to build or grow software.
CF: Thank you Andrew. Looking forward to working with you along this long, painful, tortuous journey.
