Streetlight cameras in downtown San Diego / Photo by Megan Wood

San Diego’s smart streetlights were supposed to collect and produce anonymized data that planning professionals and app-makers could then use to help solve transit and mobility problems. Officials pitched the project to the City Council and then joined with a major corporation to market it globally as an example of civic innovation.

But more than three years into San Diego’s $30 million investment, the project is failing to live up to its hype and members of the public trying to work with the data are encountering problems that throw the project’s early promotional claims into question.

City officials have acknowledged a lack of expertise internally and said they recently hired a data scientist to help. They’ve also been simultaneously renegotiating their contract with GE Current, the outside company that processes the data, to bring down the city’s costs.

Chief sustainability officer Cody Hooven described the streetlights program — which was transferred to her department more than a year ago — as a living process.

“It’s new tech and either we as a city shy away from that forever, or we try it out, and don’t get it right the first time, but hopefully evolve and learn and try to improve,” she said.

The city’s Sustainability Department has been looking at the data to see how stay-at-home orders are changing pedestrian activity, but officials said they were unaware of any public-facing apps that had been developed yet.

Back in 2017, however, officials were promising to democratize the public process by letting developers track cars, people and bicycles in real-time, all from a single 40-foot pole.

The city’s language was tame compared to GE Current’s. The company was promising to fundamentally transform the way the public interacts with its government and its neighbors with data. In one video, a marketing executive provided an example of an app that draws from streetlight sensors to let drivers find open parking spots.

“When street lamps get a tech upgrade,” the company said in text accompanying the video, “you get innovative apps that improve public health, safety, and our quality of life.”

So far, that hasn’t happened.

Working separately at academic institutions and private companies, local researchers and developers eager to build the types of apps the city said it wanted are finding that they can’t because the data is either flawed or unavailable. For instance, only about a dozen of the more than 3,000 sensors attached to the streetlights are producing information about passing pedestrians. It’s limited to a few sites that officials chose as a select number of sensors were brought online last fall.

Most of the sensors are capturing information about vehicles, but not all of it is useful and reliable.

“The parking data is just wrong,” said Eric Busboom, a technologist and the director of the San Diego Regional Data Library.

That’s because the data doesn’t reveal which public parking spots on the street are filled and which are unoccupied, multiple developers said. Instead, the data show how many cars are coming in and out of a given area.

What San Diego is experiencing is not uncommon with a lot of big data projects, Busboom said.

“I’m not surprised the city has struggled to make the best use of it,” he said. “I think they just underestimated how much effort that’s going to be.”

Daniel Obodovski, the co-founder of ScaleSD, which tries to use data and technology to help cities solve transportation and other issues, said he thought San Diego’s streetlights program was well-intentioned, but oversold. Yet it could still prove to be a long-term source of economic value, he said.

One of his major hang-ups was that while San Diego owns the raw data collected on the sensors, it doesn’t own the data that’s been analyzed for passing cars, pedestrians and bicycles. GE Current does. If developers want to troubleshoot problems or propose any changes to the software system, they go to the Boston-based private company.

In other words, Obodovski blames the contract for externalizing what should be a democratic process.

“It gave too much power to GE,” he said. “When cities negotiate contracts like this going forward, they need to be in driver’s seat.” He called it “a great lesson learned” for similar smart city rollouts in other cities.

The company didn’t return a request for comment.

Clearly San Diego officials have also learned a thing or two since embarking on this public-private partnership.

In February, according to a memo obtained by NBC San Diego, deputy chief operating officer Erik Caldwell highlighted a number of internal issues related to costs and expertise, and proposed that the city “aggressively renegotiate” the GE Current contract. Officials are expected to present more details to the City Council later this year.

Colin Santulli, a program manager at the Sustainability Department, joined the project in mid-2019, and the city hired a data scientist of its own a couple months ago.

“Having staff like myself who are thinking about it more than we did in the past is gonna really pull more benefits from the project now,” he said.

Despite its shortcomings, there’s still plenty of interest in the data locally. Santulli said his department gets a request to access the data from an outside developer or researcher once a week.

Busboom, in fact, who’s been writing about the streetlights data on his blogs for more than a year, has been able to squeeze some uses out of it. He mapped, for instance, where San Diego goes to party.

At the same time, Cedric Whitney, a visiting scholar at UC San Diego’s Institute for Practical Ethics, has also been working with Ph.D. students to test the privacy implications of the city’s open data platform. They’ve been particularly interested in recent months in knowing if one could use foot-traffic flows to, say, gauge the commercial value of a storefront or tell where a homeless encampment was located.

That is, in fact, close to what San Diego officials said they wanted as the program was being launched and what they still offer on their website as a range of possible uses.

“They’ve been framing this as ‘the city of tomorrow, building apps off public data for planning and business endeavors,’ and I’m not sure how that vision doesn’t include pedestrian data,” Whitney said.

Santulli said the city is working to get more of the devices online and expects that by end of this summer there’ll be an even ratio of sensors collecting and producing vehicle and pedestrian data for the public. The city first needed to validate that each sensor was functioning properly and that its field of view was accurate — that what the machine was seeing the human eye was seeing also. That process was completed in October 2019.

Going forward, Santulli said, the goal is to bring five to 10 pedestrian and bicycle sensors online every week, but the city will still need some feedback from transit advocates on where data is most needed to stop fatalities and meet other mobility goals. (The city is also collecting environmental and atmospheric data that will inform climate policies.) Researchers and developers also said they were eager to provide the city with feedback so the system could be improved.

In the meantime, though, Hooven noted that the sensors can only process so much at a time, so officials have to choose what they want to prioritize — what bits and pieces of information the individual devices will select for.

The timetable on this part of the project is unclear.

“Everything has a big asterisk on it now,” Santulli said.

Criticism of the city’s smart streetlights program extends beyond the transit and mobility data. Officials rolled out the technology without getting buy-in from the public ahead of time, particularly from communities of color that have experienced disparate treatment from law enforcement and that might be uneasy about devices aimed at their roads and sidewalks.

The city’s Public Safety and Livable Neighborhoods Committee is now working on an ordinance that will govern the future use and acquisition of technologies capable of watching and listening to the public.

Over the last two years, the streetlights have become primarily useful to the San Diego Police Department, which wrote its own rules for retrieving the camera footage in the absence of strong policies from the city. Police initially said only very serious crimes would warrant tapping into the technology, but the definition of what’s serious enough to merit accessing the data has expanded.

Funding has also become a source of controversy. It initially came from General Electric, which offered the City Council a loan to expand a pilot program of its technologies in the East Village to the rest of San Diego. Citing LED lighting upgrades that would allow the streetlights to be brightened or dimmed from afar, officials said the loan could be paid back with its own energy savings.

City Councilwoman Vivian Moreno has also pushed back against the use of federal anti-poverty dollars to install hundreds of the devices in low- to middle-income neighborhoods.

The coronavirus is likely to shake up government finances for years to come, as the city’s sources of revenue dry up. The current budget deficit is hovering around $300 million. For the next fiscal year, which starts in July, Mayor Kevin Faulconer is proposing significant cuts to rec centers, libraries, parks and the arts. But he wants to maintain current levels of funding for the maintenance of the smart streetlights program infrastructure — approximately $1.3 million annually for data hosting, software and more.

Budget review hearings are slated to begin Thursday and run through early June.

Jesse Marx is a former Voice of San Diego associate editor.

Leave a comment

We expect all commenters to be constructive and civil. We reserve the right to delete comments without explanation. You are welcome to flag comments to us. You are welcome to submit an opinion piece for our editors to review.

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.