THE FUTURE OF DATA SCIENCEIN THE CITY OF BOSTON

Andrew Therriault Chief Data Officer, City of Boston

andrew.therriault@boston.govtwitter: @therriaultphd

17,000+ employees

60+ departments, offices, and agencies

17 cabinet chiefs

1 aesthetically-questionable City Hall (see next slide)

Measured as a portion of a $4.9B total budget:*

31% education18% streets & infrastructure

15% police & fire5% parks & environment

4% libraries3% health & human services

2% technology

*these are all approximate figures - refer to the official budget for actual amounts!

Chief Data Officer for the City of Boston, leading Citywide

Analytics Team

Past:PhD in poli-sci from NYU (2011)Post-doc at Vanderbilt (2011-12)Analytics consultant (2012-14)Director of Data Science at the

Democratic National Committee (2014-16)

INTRODUCTION

MY BACKGROUND WHAT THE CITY DOESTHE CITY OF BOSTON

BOSTON’S CITYWIDE ANALYTICS TEAMA brief introduction

“Innovation is transforming City Hall as well. This year we are creating a Citywide Analytics Team, to bring the power of data to everything we do.”

Mayor Martin J. WalshState of the City Address, January 2015

DEC 2012Boston About Results (BAR) launches

MAR 2014Mayor’s Dashboard launches

JAN 2016CityScore launches

MILESTONES

OCT 2012Open Data Portal launches

JAN 2015Citywide Analytics Team formally announced

FEB 2017Analyze Boston (aka Open Data Portal 2.0) goes beta. Official release this spring.

PERFORMANCE MANAGEMENT

How do we improve the quality of government services and make their delivery more efficient?

MOTIVATING QUESTIONS

ENTERPRISE ANALYTICS

How do we empower others in City government to apply data and analytics to their own work?

OPEN DATA

How do we share information to promote transparency, accessibility, and collaboration?

DATA SCIENCE

How do we use data to develop new insights and identify new ways to make life in Boston better?

WHAT WE’VE BEEN WORKING ONSelected projects from the first 2 years

LOTS (!) OF DASHBOARDS

AND MAPS!

Custom applications for public information in areas such as:

● Tracking plows during snowstorms

● Information about properties for renters

● Custom reports for parents on school options & quality

MORE EXAMPLES

APPLICATIONS PERFORMANCEOPEN DATA

Collaborations with departments to optimize performance in areas like:

● Responses to 311 service requests

● Ambulance allocation and prioritization

● Firefighter scheduling and overtime use

Open Data Portal: http://data.cityofboston.gov

Boston Maps (on ArcGIS): http://cityofboston.gov/maps

Analyze Boston (in beta): http://data.boston.gov

EXPANDING THE ROLE OF DATA SCIENCEWhere we’re going in 2017 and beyond

HOMELESSNESS

Can we identify those at risk of becoming homeless?

What is the most efficient way to allocate shelter beds?

DATA-DRIVEN JUSTICE

How can we improve mental health or addiction treatment?

How do we keep at-risk youth away from gang violence?

EDUCATION

How do we find students at risk of dropping out?

How can we help students affected by fires or crime?

HEALTH & SAFETY

Which buildings are most likely to have a fire?

Which restaurants are most likely to cause illnesses?

PLENTY OF OPPORTUNITIES

CITY SERVICES

How can we route 311 requests to the right people?

How do we make sure we’re serving all of our residents?

URBAN MOBILITY

What areas are underserved by existing public transit?

How do we make the city safer for cyclists and pedestrians?

Building the technical infrastructure for data science in a city context is much the same as in a private company, but at a much larger scale (hundreds of data sources) and with more limited resources.

AND PLENTY OF CHALLENGES

INFRASTRUCTURE TECH EXPERTISEGOVERNANCE

The skillsets our team brings are much more varied than on a typical data science team. Because of the wide range of our responsibilities, we have more experts in specific subjects and tools, but fewer “full-stack” data scientists.

We have to understand the implications of public records law and uphold our duty to use information about the public in a responsible way, to respect citizens’ privacy and maintain the public’s trust.

PREVENTING FOODBORNE ILLNESS AND OTHER RISKS

Since July 2016, the City of Boston has used predictive models of critical health code violations to prioritize restaurant inspections.

Using inspection histories, citizen complaints, property records, and more, these models have helped to improve inspection efficiency by more than 20%, so problems are addressed earlier and more often.

A similar approach can also be applied to other risks such as students dropping out and individuals becoming homeless, where early intervention is key.

IMPROVING SUBSTANCE ABUSE AND MENTAL HEALTH TREATMENT

We’re working to expand Boston’s “311 for Recovery Services” system to coordinate care and keep patients on track by integrating data and making it accessible to providers.

We’ve also piloted a system to help identify patients in our public health systems who are likely to overdose, in order to provide proactive interventions before that happens.

If successful, these program could be expanded to include a broader range of mental health issues, in order to better connect those in need with specialists who can help.

PREDICTING CRASHES ON OUR STREETS

We’ve started a project to build crash prediction models for Boston’s streets to address structural problems and prevent tragedies.

These models will combine data on traffic patterns and hazards from a wide variety of sources, including traffic sensors, GPS data, historical crash records, traffic incident reports, and residents’ reports of safety concerns across the city.

We’re doing this as a collaborative effort with volunteers from Data for Democracy, and building on open source so it can be adopted by other cities & states as well.

LOOKING BEYOND BOSTONWhy we need more civic data scientists

Civic data science: the application of data science methods* to support the work of governments, nonprofits, advocacy groups, political campaigns, and other organizations which exist to deliver positive social impacts

* broadly defined—everything from fundamental data management, analysis, and reporting to predictive

models, experimental tests, and complex visualizations

Two core principles: efficiency and effectiveness.

With scarce resources, this kind of optimization can be the difference between success and failure. And even if not, marginal gains are still very meaningful.

Beyond just being meaningful, civic data science projects can also be really interesting challenges to work on.

Subjects are varied and unique, people are smart and motivated, and opportunities for creativity and innovation are unparalleled.

The tech industry talks about making the world better (and sometimes does), but the ultimate motives are still financial.

Civic orgs use money as a tool to promote social goals, and their direct impacts are much larger.

THE CASE FOR CIVIC DATA SCIENCE

CIVIC ORGS MATTER EXCITING WORKBIG POTENTIAL IMPACTS

My final pitch: Whatever your long-term career plans, spending a few years working on civic data will make you a much better data scientist.

You’ll work on more varied projects, learn more new skills, advance more quickly, have more autonomy, and get more opportunities to share your work with the world.

Even if your main goal in doing data science is to make money, the experience will be much more valuable than the modestly higher salary you’d otherwise get now.

THANK YOU!

For more information about the City of Boston’s Analytics Team, go to boston.gov/analytics.

And if you want to talk more:andrew.therriault@boston.gov (work)

andrew.therriault@gmail.com (personal)@therriaultphd (twitter)