The Center for Data Innovation interviewed Karen Lightman, executive director of Metro21: Smart Cities Institute at Carnegie Mellon University, which researches, develops, and deploys technological solutions to solve the problems facing metro areas. Lightman discussed how cities can support open data initiatives and how Metro21 has helped Pittsburgh create safer intersections using sensors.
This interview has been edited.
Michael McLaughlin: Smart cities are often thought of as a wholly different way of doing things and using new technologies. However, cities are already collecting and using a lot of valuable data that can make them smarter. What types of data are many cities already collecting that they can better leverage to improve city operations?
Karen Lightman: Before identifying sources of data, it’s important for policymakers and decisionmakers to have a clear sense of the overall purpose of gathering data as well as how it will be used.
Open data is critical to ensuring transparency and building trust within a community, and when using this data, policymakers have to be cognizant of privacy and security. At Carnegie Mellon University, we work with our partners at the City of Pittsburgh which maintains Burgh’s Eye View. The database is populated with real-time information of 311 calls, crime reports from the police, and parking data. Through our partnerships and related data-sharing agreements, we also have access to non-public information that helps us create tools to improve citizen quality of life.
When implementing policies based on this information, decisionmakers must recognize that the technology is always evolving and that a policy may become outdated based on new innovations. Having a relationship with a university like Carnegie Mellon can help policymakers stay current on recent technology developments and better understand the intended and unintended consequences when using data.
But the data is truly endless. One exciting example is the use of smart garbage cans to help public works teams determine when trash needs to be removed from certain locations. Through effective use of data, cities can optimize resources, improve efficiency, and build a better quality of life for members of the community.
McLaughlin: How should cities plan to support open data initiatives as they collect more kinds of data for smart city programs? What kind of guidelines should they use to determine what data should be open?
Lightman: Having a neutral third party that can serve as a warehouse and portal for open data, like the Western Pennsylvania Regional Data Center, is a great first step. This entity can become a trusted resource—offering training, hosting workshops, and conducting outreach to the local community. From a tech enthusiast hacking on a weekend to a PhD student working on a thesis, this database should be available to the public 24/7 and provide clean, robust and reliable information to all audiences. The City of Pittsburgh and Allegheny County have done a tremendous job creating a user-friendly website that is relatively easy to navigate and help people understand how to use the data. And the information is robust—pulling from several resources including data collected via Burgh’s Eye View. All of this information helps researchers and the city better understand what is happening in each neighborhood and leads to ideas and decisions that can improve daily lives.
Before city officials consider creating such a resource, it’s important that they engage in conversations with social scientists and ethicists to better understand possible unintended consequences. What is the worst thing that could happen if inaccurate data gets out? What needs to be built in the process to avoid such issues? Asking these types of questions will ensure a strong data resource that can influence effective policy decisions.
The only way to build trust is through collaboration. Open data gives researchers the ability to focus on deploying technology and doing it in a way that is iterative and collaborative.
A city that is a leader in this space is Boston. They created a smart city playbook that serves as a guideline for how members of the community can engage with Boston on smart city projects. The guideline outlines important criteria, rules of engagement, best practices, next steps, and other important components. When it comes to smart cities, citizen engagement is paramount. Every city should look to this effort as a model.
McLaughlin: How is Metro21 using image sensors in its Platform Pittsburgh project? How can the project lead to safer intersections?
Lightman: The goal of Platform Pittsburgh is to understand why near misses—instances where pedestrians are almost hit by cars—happen at intersections. To understand these instances, the project team posted eight high-resolution cameras at one intersection to capture data from a variety of angles. Using tagging and machine learning on those images, data is processed in real-time and an initial analysis is conducted in an equipment box at the intersection, processing data at the edge—a technique referred to as edge computing. This edge computing technology then sends the already analyzed data to a bigger computer on campus, where a more sophisticated analysis is conducted to see blind spots, evaluate signage, and identify better designs for intersections to avoid near misses.
Another component of safer intersections is understanding the environmental impact of idling cars. The team used machine learning to train an algorithm to identify the amount of particulate matter in emissions from trucks versus electric vehicles. By studying particulate matter in tailpipe emissions, the team could understand other factors impacting air pollution and public health.
McLaughlin: Metro21 is also using data to assess a structure’s fire risk. What data is Metro21 using in its model? How has the model helped the relevant authorities prioritize which structures to inspect?
Lightman: Using information from 311 calls and past inspection reports, Metro21 used data to predict the likelihood of fires in commercial buildings. By digitizing information and using predictive analytics, researchers were able to determine the likelihood of structure fires in a building during one year—improving the accuracy of fire prediction from zero to 85 percent. There are more than 20,000 commercial buildings in Pittsburgh that need to be inspected and only six inspectors. This model has helped authorities take a much more calculated and methodical approach, helping them to prioritize which buildings to inspect.
McLaughlin: Pittsburgh police are evaluating a predictive analytics program Metro21 developed to identify areas that will likely have crime flare-ups. Can you discuss how Metro21 made sure the model did not encode any existing bias in police data?
Lightman: Researchers at Carnegie Mellon University partnered with the Pittsburgh Bureau of Police, the Pittsburgh Department of Innovation and Performance, and the Pittsburgh Public Safety Department to reduce crime through a new hot spot policing program. The team used artificial intelligence methods to analyze information from a database of more than 300,000 crime incidents and more than 1.5 million 911 calls in Pittsburgh to develop a tool that predicted violent crime at specific locations, but not specific individuals, across the entire city.
One unique component that differentiates this program from others is that it is the first in the United States to predict and manage temporary hot spots for proactive police patrol activity. Through these efforts, the program successfully decreased the number of serious violent crimes by more than 34 percent in some areas of the city during a one-year span.
Whether it’s predicting fires in commercial buildings or helping police reduce crime, the university plays an important role in showing municipal partners what is possible. We leverage data and technology to develop the tools, but it is ultimately up to policymakers to determine what to do with the information that is available.