Rehoboth using AI to track visitors

Earlier this year, without public notice or discussion, the city installed a handful of Flock Safety license plate readers on State Road, Bayard Avenue and Rehoboth Avenue.

The license plate readers, which cost $12,000 to purchase and come with an annual maintenance contract of $12,000, use artificial intelligence-powered technology to identify vehicles based on make, model, color and other details.

At around the same time, early spring and with a similar level of transparency, the city executed a contract with Placer.ai, an analytics company that collects geolocation data using cellphone apps, for $21,000. City officials never discussed the city’s use of the technology in advance of its implementation, or the contract, but they have referenced it a handful of times over the past few months.

In September, City Manager Taylour Tedder touched on the technology briefly during his State of the City presentation. He described it as a visitor analytics and policy platform.

“People have never been able to say how many people are in Rehoboth Beach at a time. This gives us more insights and helps us gain better access to grants,” said Tedder.

In late October, as part of a statistics-related wrap-up of Sea Witch, the city said Placer.ai showed there were 93,700 visitors over the weekend.

The scope of data being collected was made clear more recently, as part of the discussion about what to do with the summer concert series at the Bandstand. Commissioner Craig Thier said the city has the ability to know how many people are in town at that specific time, where they came from, where in town they are and where in town they went. The only thing the city doesn’t know is how much someone spends in a store, he said.

The concept of using Placer.ai was brought to the city by Rehoboth Beach Main Street, said Brooke Thaler, city communications manager. City officials took a look at it and thought the information they provide could be helpful and allow the city to make more data-driven decisions.

Placer.ai collects location intelligence from mobile devices via third-party apps that have user permission to share location information, said Thaler. The platform combines this data with other sources such as public records, available demographic information and surveys, she said.

“It is important to note that the data collected and provided is anonymous and contains no personally identifying information. We are not tracking people with this service,” said Thaler. “We are looking at the bigger picture to help us make better-informed decisions. We are not spying on people.”

The city provided the Cape Gazette pages of information related to the type of data that’s being collected by Placer.ai. The data was specific to the Bandstand area during the summer concert times. Filters were set to only include data from a person who was in the Bandstand area for longer than five minutes, between 6 and 10 p.m., Friday through Sunday, from May 24 to Aug. 30.

Given the filters, the city knows there were 157,200 visitors over the course of the summer during the concert series and the average dwelling time was 45 minutes. The city knows, generally speaking, what percentage of visitors are coming from other parts of Delaware or where in surrounding states. The data provides the specific businesses people came from before going to the Bandstand and where they went afterward.

The technology could be used by the city’s police department, said Thaler. However, she said, it’s not currently being used by the police and there are currently no plans to.

Jerry Filbin, Rehoboth Beach Main Street board member, said the organization has explored the use of this technology as a way to stimulate off-season business, but it was too expensive. He said Main Street broached the subject and the possible benefits with the city and Thier, who is the commissioner liaison to Main Street.

Dan Slagle, Rehoboth Beach Main Street executive director, said the technology is discussed at industry conferences as being a useful tool to find out demographic data for visitors. Other Main Streets are having success with it, he said.

For Main Street’s purposes, Filbin said they’re still trying to figure out how to best use the technology.

Take Main Street’s annual Gumbo Crawl for example, said Filbin. Predictably, weather plays a big part in the day-of participation in an event like that, but the data collected shows how big of an influence. Maybe the crawl should be moved to when the weather is cooler, but dryer, he said.

“This is the type of information we need to continue to look at,” said Filbin.

Filbin said concerns related to tracking are valid, but people tacitly agree to be tracked when they download different apps. If people really don’t want to be tracked, they can turn that function off, he said.

Thier could not be reached for comment.

Expert explains technology

Phil Hagen is a cybersecurity expert who lives in Lewes. He’s the person who reached out to Rehoboth officials with transparency concerns about the city using Flock Safety. In response to his concerns, the city created a transparency portal related to the type of information Flock was collecting. Hagen sent the following response when the Cape Gazette asked for his insights into the Placer.ai technology:

“This general use case for AI technology is much more reasonable than some of the other, far less ethical ones. It appears the company is using their AI model(s) to analyze a wide variety of information into summarized form(s) that are shaped to a specific kind of question such as foot traffic in an area, use of public facilities, etc. Personally, it strikes me as a large-scale operations research service which isn’t covering a whole lot of net-new ground, but primarily attempting to improve the quality of the analysis with the sheer mass of data sources.

“As with any such AI implementation, though, those input data sources are a key to their value. From what I can tell, their data is primarily from mobile apps that have been configured to send telemetry to Placer – most notably, location data. While there is no reason to suggest they are not properly anonymizing and aggregating that telemetry to address the obvious privacy risk, there is no such thing as a perfect data protection scheme. There is some risk associated with that type of sensitive information being collected in the first place, so a large collection of that sensitive information compounds the risk.

“I feel this is why it’s so important to be cautious about what applications you share data with on a mobile device. While having hyper-localized context and time estimates for a food delivery is a great feature, it’s rarely easy to find out what other services the application developer is feeding that location data to. Periodically reviewing what applications have ‘always-on’ access to location and restricting or even revoking that permission is an important part of safely managing your data.”