Making cities smarter How citizens’ collective intelligence can guide better decision making
January 26, 2017
As massive, complex organisms, cities somehow manage to deliver services, move products and goods, protect us, educate us, inspire us. But how can they do it smarter? Jim Guszcza and Bill Eggers talk about how urban planners can harness the wisdom of the crowd to deliver better services.
If you're a smart leader you know you don't know the next [person] that it's going to come from. It can come from the most junior person in the organization. But the more voices you can listen to the better chance you are going to have of getting the estimate right or getting the new idea that no one else thought of.
Explore Deloitte Review Issue 20
TANYA OTT: Where do good ideas come from, and how can we reorient our thinking to find them?
I’m Tanya Ott, and this is the Press Room—Deloitte University Press’s podcast on issues and ideas that matter to your business. And today, we’re tackling cities—those massive, complex organisms that somehow manage to deliver services, move products and goods, protect us, educate us, inspire us. But how do they do it—and how can they do it smarter?
We’ll go there in a minute, but first, let’s rewind more than a century to a guy named Francis Galton. He lived in England, and he loved to count and measure things. He was a keen observer. And, in 1906, he made a curious observation while visiting a livestock fair.
JIM GUSZCZA: He said he witnessed a game at the County Fair, which is if you can guess the weight of an ox, you win the game.
TANYA OTT: That’s Deloitte Consulting’s Chief Data Scientist Jim Guszcza.
JIM GUSZCZA: So Galton, he collected all this data . . .
TANYA OTT: All the guesses that people made about the ox’s weight . . .
JIM GUSZCZA: And he wanted to analyze the data and prove a point: that people actually were very bad at processing information and weren't very good at making decisions. But when he analyzed the data, what he found was that the median guess, the bell curve of guesses, came within one pound of the truth to the ox.
TANYA OTT: So basically, Galton discovered the wisdom of the crowd: the idea that a smart team is smarter than the smartest person on the team. Galton, by the way, was pretty smart himself. He basically invented the idea of regression toward the mean, and regression analysis. He’s not quite as famous as his cousin, Charles Darwin. But he was a noted mathematician and anthropologist, and he was knighted a few years after the livestock fair epiphany.
The wisdom of the crowd is everywhere. You just have to know how to find it. Let’s take a modern example.
Imagine you're an urban planner. You're charged with putting walkways in a newly built park. How would you decide where to put those paths? Maybe you survey likely park users. Maybe you copy best practices from other parks in the city. Maybe you build a sophisticated computer simulation that can help you predict foot traffic.
Those are traditional ways of tackling the problem. But what if you started by just . . . looking?
BILL EGGERS: Open up the park. Observe the different paths in the dirt that are naturally created by foot traffic. You're looking at how are they actually getting from one place to the other. What are the places [where] foot traffic [is very heavy]? Which is the most natural for them to do. Then basically you're building the path that way, as opposed to trying to preplan ahead of time how people should walk from one place to another.
TANYA OTT: That’s Bill Eggers. He’s executive director of Deloitte’s Center for Government and Higher Education. He and Jim Guszcza have been thinking a lot about how these kinds of decisions get made. They write about it in their article Making cities smarter: How citizens’ collective intelligence can guide better decision making.
The park path example is very organic—and low tech. But Bill and Jim are really interested in how cities can marry that discovery model with technology—especially now that we’ve got millions of apps and Internet of Things sensors and other ways of gathering data.
JIM GUSZCZA: Now we’re leaving digital breadcrumbs behind as we go about our day-to-day activities because we're all carrying these little smartphones with us. They capture our locations or the buses we're riding in, or the cars we're driving are connected the sensors. Now it is much more possible to gather information about people's day-to-day activities and use this behavioral data to actually figure out, what do people want? What are the services, what are the bits of infrastructure improvements that would really make their lives easier and safer and more prosperous? It’s kind of like tapping into the decentralized knowledge and preferences and behaviors of the multitude. Use that to inform the decisions rather than relying more on theory or ideas of central planning.
BILL EGGERS: The fact of the matter is cities are some of the smartest, most intelligent things ever invented by humankind. Think about how cities work, and how we are able to get all sorts of groceries and everything else that goes on in city life, and that complexity. Cities are already really, really smart. What the smart city movement is all about is [to] take that wisdom that's already there and accelerate it vis-à-vis technology and behavioral economics, tapping the wisdom of the crowd and some other new tools that we have today.
TANYA OTT: You've got some really interesting examples. The New York City Department of building, for instance, for years chose which properties to inspect for problems based on complaints. Which makes a whole lot of sense—except in 2011 alone, the city received almost 25,000 complaints about just one very specific type of problem.
BILL EGGERS: But they only had 200 inspectors to cover the workload. So how do you decide which places to inspect? It's a really, really difficult issue. And if you end up inspecting a lot that are fine, and you miss the ones that are bad, you could end up with people dying because of fires and unsafe conditions. So that's the issue that the city was facing, and what they did was that they started applying analytics. They talked to the building inspectors, and they found out the inspectors knew a lot of things. They actually built a guide into a predictive data model to help understand what are the triggers that are most important, and what do we know that will let us get to these properties that are the most likely to be unsafe, and fire hazards due to these illegal conversions.
JIM GUSZCZA: They fed this into a predictive model that resulted in the risk score. They were able to literally come up with a prediction score of each building in New York City, each building site, for its riskiness, its need of a vacate order. And so by rank order, all the buildings from most risky to least risky, they were able to target their inspection activities accordingly.
BILL EGGERS: Before they were doing it, they had about a 70 percent false-positive rate, or even higher than that. Essentially, previously only 13 percent of the complaints ended up requiring vacate orders—which meant that all but 13 percent of the time, when they would go into one of these properties, it was actually fine. So they were missing a lot of other ones.
Now after they filtered out a lot of these false positives, the share of complaints leading to vacate orders escalated up to 70 percent. They didn't have to re-engineer anything. They just had a better predictive model, again, using a little bit of the wisdom of the crowds. The wonderful thing that happened was that you actually saw a lowered risk for firefighters in fires in illegal conversions—they went down dramatically. We saw New York City experience zero fire deaths for the first time since 1916, in June 2015, using this analytics-driven approach.
TANYA OTT: Cities are using technology and data in other interesting ways. In some cities, they’re encouraging residents to use their cellphones to snap photos of potholes and broken water mains. Using geolocation, they can map the problems and then deploy resources accordingly.
But there can be problems. Several years ago, there was a media company in a major metropolitan area that decided to crowdsource reports of closed roads and excessive snow and ice during a blizzard. They encouraged people to report problems, and then they mapped the locations. But a funny thing happened. The map showed problems in all but one quarter of the metro area. It was clear. Clean. No snow—or, actually, just no respondents from that area. Maybe they weren’t big fans of the media company. Maybe they couldn’t afford a subscription.
Jim Guszcza says cities have to be really sure they’re actually capturing the data they think they’re capturing—otherwise they might fall victim to algorithmic bias.
JIM GUSZCZA: As a data scientist, I can tell you that whenever we build predictive models or artificial intelligence apps or what have you, those models or those apps are going to be dependable and reliable and accurate only to the extent that the data that they were built on is representative and complete of the world going forward in which you’re going to implement the models. You know the world has changed, then your model might not be accurate. We've seen this in elections, right? If the data you're feeding into a predictive model or forecasting model is not representative of who's actually going to vote, you might get a wrong indication or a misleading indication. Or if you're not smart enough to adjust for those discrepancies.
So that's a general point about predictive models, that the data has to be suitably representative in a way that helps you create better indications going forward. In the case of cities, you’re exactly right. If only one segment of the population owns the mobile phones or has the cars connected to sensors or whatever it is, then you're going to get a lot of blanked-out parts of your map. Metaphorically speaking, maybe literally speaking.
TANYA OTT: One of the other things that I found really interesting is you write about how smart cities are making their data publicly available, and you talk about an example in Boston where they're using Yelp restaurant reviews as raw data for predictive algorithms. What are they trying to do there?
JIM GUSZCZA: Yeah, that was an ingenious example. Again, it's this idea of digital breadcrumbs. It’s very analogous to this New York City building inspection example in some ways. In both cases, you've got health inspectors or hygiene inspectors issuing labels: this restaurant is very clean and very safe. This other restaurant maybe has a health violation. It’s very similar to the idea of, this building is unsafe so let’s issue a vacate order. So the logic is very similar, and in both cases, we can use a predictive model to identify, flag ahead of time, restaurants that might need a health violation. We can build a predictive model to send out our health inspectors to the restaurants in most need of attention.
And the data they're using to do that, what was particularly thought-provoking about this example, is they're using Yelp restaurant review data. It's unstructured text, so from a technical point of view, it’s interesting. But it's also interesting from this perspective that we were saying earlier: Again, these are examples of digital breadcrumbs.
You know in the old days, if you went to a restaurant, and you had a bad experience, there it would be word of mouth. You might just tell your friends. You might write a letter to somebody. You might just verbally tell somebody, oh yeah, I felt queasy after eating at this restaurant last night.
Now a lot of this is on Yelp reviews or other social media reviews. So why not just take that data and use that as inputs to predict ahead of time which restaurants are most likely to need a health citation? The logic is very similar to the New York City building inspection example, but the data being used was particularly thought-provoking, and it really does exemplify this idea of people leaving behind digital breadcrumbs they go about the day to day lives.
TANYA OTT: Some of the examples Jim Guszcza and Bill Eggers studied were pretty high tech. But some others were pretty basic. Maybe not obvious—at least to someone who’s not a behavioral economist—but basic.
Take for instance, an interesting little tweak that made a big difference along Lake Shore Drive in Chicago.
BILL EGGERS: It was a really, really dangerous stretch of road when I was growing up, because the curve came so suddenly and people were going very quickly. So they had to figure out how to reduce the number of accidents, especially near this one area of the curve. What they did is they created a visual illusion by painting a sequence of white lines from the pavement, and each much narrower than the previous one on the approach to the dangerous curve. So what we had was this succession of shrinking lines, and it gave drivers the feeling that they were actually speeding up, prompting them to slow down and take the curve at a safer speed, even though it was no different than it was before, but it was that illusion that it created. That ended up reducing crashes within the first six months by over a third, just from that very, very simple change. And that's using the science of behavioral economics and human psychology to improve public infrastructure, in this case, and dramatically reduce traffic accidents.
TANYA OTT: I think that's really interesting. We were just in Chicago a couple of months ago. It's a fascinating example.
BILL EGGERS: Yeah, and there's a million of those sort of examples. How do you get people to walk more up the stairs and get better exercise? What they did in Stockholm was they partnered with a company who then made it [so] that music would play as you walked up the stairs. And the people loved it. They enjoyed it, and the amount that they walked up the stairs increased dramatically because they changed what the experience was. They made it a great experience to actually walk up the stairs rather than ride up the escalator.
Or how do you get people not to sit in the subway seats for the seniors and disabled? Well, make it make it a fun thing. Put a happy face on, and say, I know you're a good person and be sweet to your seniors. There's a whole bunch of different ways where you can change that physical environment and tweak it in different ways where we know human psychology responds in a more positive way. That's going to be more positive for the general civic population.
TANYA OTT: The city of Philadelphia used behavioral economics and the idea of nudges to get more people to pay their property taxes. Originally their letters to taxpayers were stuffed with jargon and legal terms and threats about penalties. About 10 percent of property owners were delinquent. That’s a pretty high number. But when they rewrote the letters to appeal to residents’ civic duty . . .
BILL EGGERS: Saying, here are specific public services like garbage collection and schools and so forth the property taxes paid for . . .
TANYA OTT: More people paid their bills.
BILL EGGERS: They are able to reduce unpaid taxes by one third in just two years, just by changing a lot of the language and how they set that up—which is a pretty huge number for seemingly small tweaks.
TANYA OTT: We have a saying down here in the South, where I live: You catch more flies with honey than with vinegar. Kind of a truism—like that one about death and taxes.
Do we have you thinking about how you might apply this kind of discovery thinking to crowdsource solutions for your city or your company? Bill Eggers says one of the most important things to remember is to be flexible. Try things out, and, if they don’t work, adjust.
BILL EGGERS: One of the big problems with government is [it] will pass the bill and say, this is how you should do this, and then never test and iterate and test. I talk about beta government. Government should be more like technology, where we're constantly testing programs to see what works, what doesn't work and iterating on it. The problem right now is we have what I call a waterfall model, essentially, where we go down this path, and we're never testing it with real people. Then we end up falling into what I call the design-free design trap, where you've got these designs that you [have seen written] about, you figured it out in a room somewhere, and then you put it out to real people, and they didn't work because you hadn't tested them constantly. Anything that's digital today can be tested on a daily basis to see what's working and what's not working. Government needs to move more in that very, very iterative way. That's [what] making smart [government] is all about, right? It's basically being an all-learning organization. Think about government as an applied learning organization that's able to be very nimble, very agile, and constantly changing based on what they're learning and interactions with real citizens.
TANYA OTT: What are the biggest obstacles that governments face in trying to do this?
BILL EGGERS: One of the challenges certainly is within legislative policymaking, where it gets written up into a law, into statute, and they get very prescriptive about how civil servants need to administer something that doesn't allow them to iterate. I mean, look at with Obamacare—we found lots of problems throughout it that the administration wanted to fix, but at that point they had a Congress that was not willing to make a lot of those changes. So I think the answer to that is we need to be much less prescriptive from a legislative front, just focus much more on the outcomes that we want to achieve, and also build into that the fact that there's going to have to be a lot of testing and iterating all throughout the process, because we never know in advance what is going to work and what's not going to work. You only know that when you're testing with real people and real situations, you always are surprised. That's why we talk about this discovery process, and how that relates to being a real smart city, and using the wisdom of thousands, if not millions, of citizens and aggregating that up.
TANYA OTT: Bill Eggers and Jim Guszcza’s article on smarter cities has many more case studies and advice for applying these principles. You can find it at dupress.com. That’s also where you’ll find our show archive—and where you can subscribe so you don’t have to miss anything.
We love to hear what you think about what we’re doing. You can tweet us at @du_press and email us at firstname.lastname@example.org. I’m Tanya Ott for the Press Room. Thanks for listening, and have a great day!
This podcast is provided by Deloitte LLP and its subsidiaries and is intended to provide general information only. This podcast is not intended to constitute advice or services of any kind. For additional information about Deloitte LLP and its subsidiaries, go to Deloitte.com/about.