For the first time, Portland’s transportation planners could get access to the powerful insights companies glean from tracking the location, moment by moment, of millions of cellphone users.
On Wednesday, after little debate, the Portland City Council voted unanimously to approve an agreement between the Portland Bureau of Transportation, TriMet and the regional government Metro to pay for a pilot test of a powerful new program called Replica.
Replica is software tool developed by Sidewalk Labs, the city-building subsidiary of Alphabet, Google’s parent company. Piloting it for a year will cost the three government agencies a total of $457,300. After that, agencies can choose to buy a one-year subscription for 12 cents per resident.
The tool is powered by the vast trove of precise location data collected by smartphones and smartphone apps that allow programmers to track the movements of their users. Sidewalk Labs will remix that data with information from the U.S. Census to create what it calls a “synthetic” version of the Portland metro area, populated by 2.3 million virtual people.
It’s SimCity, except the imaginary Sims are based on Portland’s real demographics and thousands of daily trips to work, school, shopping and the doctor captured by residents’ phones.
“It’s a full representation: all households, all people, all modes of transportation, the entire street network that people use, the sidewalk network — but all the people are synthetic,” said Nick Bowden, who leads the team developing Replica at Sidewalk Labs.
“It will be equivalent to how people move, but no single person could re-identify themselves in that web of movements.”
Portland’s transportation planners say the tool could dramatically improve their ability to understand and monitor how people use streets, sidewalks and transit.
“It is data that will augment, enhance, possibly replace PBOT’s existing data collection efforts,” said City Commissioner Chloe Eudaly, casting her vote to fund the project. Eudaly oversees the transportation bureau.
PBOT officials say the data Sidewalk Labs works with is “fully anonymized” and that the government agencies involved in the pilot will only be able to see travel patterns and will never collect or maintain individualized data.
But they also acknowledge that they don’t fully understand where Sidewalk Labs sources the mobile location data it uses in its model.
“Part of this entire effort is gaining a better understanding of how are they doing this, lining that up against our own citywide privacy standards, and making a comparison between the two,” said Michael Kerr, manager of the Office of Strategy and Innovation with PBOT.
A Powerful New Model
Sidewalk Labs is best known for its effort to build, from scratch, a new neighborhood called Quayside on 12 acres of riverfront in Toronto.
The Replica software is a separate initiative. Portland is one of eight U.S. cities, including Chicago and Kansas City, selected to test the tool.
The company says many cities can’t answer basic questions about how people use streets, sidewalks and public transit because their data isn’t sufficient.
“Since around the early ‘50s, regional planning agencies and cities have used the same kind of travel modeling technique that most of them still use today,” Bowden said. “They do a paper survey, typically; that’s a small sample of the population. They ask them to kind of diary out their movements for a particular week.”
Metro, for example, periodically conducts a travel behavior survey as part of its regional planning responsibilities. The most recent survey took place in 2011 and involved 6,450 households. The agency then extrapolated from that sample to model how people move around the region.
In addition, Portland collects some real-time data on how people travel the city using traffic cameras and loop detectors in the pavement that can count cars, bike sensors mounted on two of the city’s bridges, and volunteer “hand-counts” of pedestrians.
Much of that data isn’t gathered in a consistent, regular way.
Because its data is old or incomplete, simple questions are surprisingly difficult for PBOT to answer: How do Portlanders’ driving and walking habits change in different seasons? At different times of the day?
By contrast, the data that powers the Replica model will be updated every three months. Kerr says getting fresh data multiple times a year could help planners with “seeing the before and after” of the investments the city makes in its streets.
“If we install a new crosswalk, are we seeing more people walking as a result of that investment than would have walked before?” he said.
Portland’s population and housing costs have both skyrocketed in the past five years, pushing many people to the city’s outskirts.
Kerr says if Replica’s model works, it could help the city figure out how those displaced by the housing boom are commuting, and in turn make wiser decisions about where to invest in new public transit.
“What are their major forms of transportation?” he said. “And are there tremendous gaps in the level of service that they have available to them?”
Portland leaders agreed to take part in the Replica pilot quietly, after little public debate.
Eudaly placed the agreement on the Council’s consent agenda last week, meaning commissioners would have voted on it without discussion first. A member of the public requested the item be pulled, and so PBOT staff gave a brief presentation about the project.
The Council voted to fund the pilot project just days after a New York Times investigation revealed the extent of the data harvesting effort from personal smartphones and questioned whether users understand how closely they are tracked and how their data is resold, often to advertisers.
“Location information can reveal some of the most intimate details of a person’s life,” U.S. Sen. Ron Wyden, Oregon’s senior senator and a noted privacy advocate, told the Times. “It’s not right to have consumers kept in the dark about how their data is sold and shared and then leave them unable to do anything about it.”
Sidewalk Labs says Replica has been designed with privacy protections in mind, because the tool has been intentionally crafted to help public agencies that have been unable to use the datasets in the past, in part because of privacy problems.
“We believe pretty strongly that this data can be used in a privacy-sensitive way, first and foremost, and that the use of the data can have really great public good,” Bowden said.
Neither PBOT nor Metro officials have a complete understanding of where Sidewalk Labs obtains the mobile location data that it uses.
Documents filed with the City Council state the data comes from “Android Phones and Google apps,” but PBOT officials said the data might come from other sources too.
A Metro spokesman gave a vague answer as to where Sidewalk Labs sources its mobile location data but said Metro councilors expect more details before the deal is finalized.
“We expect to see full documentation of data collection sources before we sign a contract,” said Metro’s Nick Christensen.
Bowden said Sidewalk Labs works with different data providers, including telecommunications companies and companies that aggregate mobile location data from different apps.
“We only will receive data that has been entirely scraped of personal identifiable information,” he said.
A spokesman for Sidewalk Labs says Google is not a data source for Replica.
Bowden said Sidewalk Labs audits the companies that supply it with mobile location data and exclusively works with companies that give phone users a clear choice to opt out of the data collection.
“A big part of that checklist or audit that we go through is whether or not their sources of the data provides users with appropriate user consent,” he said.
Bowden says the company takes one other step to protect privacy: Any time one of the synthetic people populating the model has too unique of a travel pattern, that virtual person is removed from Replica.
Portland Bureau of Transportation officials say they believe Google follows a stringent standard when it comes to privacy protections.
The agency says that unlike the private companies that are mining cellphone location data in an effort to better tailor ads, PBOT isn’t trying to learn about individual households.
“We’re looking at this as a tool to make our roads safer, to make it more efficient,” said John Brady, PBOT’s spokesman. “For us, it’s all about trends and patterns.”