AV研究所

Your research lies at the intersection of machine learning, natural language processing, and large language models. What drew you to applying those fields to social impact?

I received my PhD in computer science, and much of my research has focused on data science and machine learning for social science applications. In particular, my dissertation centered on rewiring police officer training networks to reduce forecasted use of force, so it just felt like a natural transition from that to my current position.

How would you describe the work you鈥檙e doing at the Boston College School of Social Work?

It鈥檚 mostly applied data science. I was hired for the Gateway Initiative, the School鈥檚 long-term commitment to building capacity across Massachusetts鈥� 26 Gateway Cities鈥攎idsize urban centers that anchor regional economies around the state. A lot of the work that I do is using publicly available data and providing insights from that. Sometimes it鈥檚 also providing nonprofits with the help that they need鈥攄ata collection, analysis, visualization.聽

What project best illustrates how you鈥檙e using data science through the Gateway Initiative?

Most of the impact that we have had through the Gateway Initiative is providing information in an accessible way. There鈥檚 so much data out there, but it鈥檚 hard to analyze, visualize, and understand, so that鈥檚 where we come in.

In particular, we鈥檙e working on a data-driven initiative aimed at better understanding police incidents through the analysis of publicly available police data. The Lawrence Police Department publishes every call for service it receives, so we downloaded the entire dataset, which includes information from 2018 through 2024, and now we鈥檙e working on updating it every day.

Using this data, we designed an interactive dashboard for people to better understand incidents in their community. Where are these incidents happening? Are there specific types of incidents happening in specific areas of their city? What neighborhoods are more prone to motor vehicle incidents of certain types?

This information is also helpful for policymakers. Let鈥檚 say they notice a hotspot for motor vehicle incidents. They might examine whether road conditions are contributing to those incidents and determine whether certain roads need to be fixed鈥攇oing to the root cause of the problem.

What kinds of information can people explore through the dashboard?

The dashboard has 14 or 15 different incident categories. We started with motor vehicle incidents because they are the most common, but it also includes violent incidents, drug-related incidents, and public disturbances.

We also have a heat map鈥攜ou can click and see exactly where incidents are happening. And you can select different years: Do you want to see how incidents have changed from 2018 up until 2024? You can do that.

You can also overlay different types of data, including census data, unemployment data, and education data. We also have points of interest. So you can overlay where, for example, schools are, where grocery stores are, and where nonprofits are.

How has feedback from the Lawrence community shaped the work?

This past summer, we met with nonprofit staff and community leaders, who provided feedback on the tools that we鈥檝e developed, which was very helpful.聽

Our next step is training community members to use these tools and asking them to help us understand what else would be useful for them. We aren鈥檛 from Lawrence, and we recognize that we are not the community. We need to understand what residents鈥� needs are and how we can be helpful in meeting them.

You鈥檝e incorporated AI into the public safety project. How does that work?

We recently created a chatbot so that community members can ask questions in plain English. The good thing is that it鈥檚 grounded in this incident data. So if it doesn鈥檛 know anything about the question you ask, it will tell you, 鈥淚 don鈥檛 know anything about it.鈥�

Let鈥檚 say you ask: 鈥淚n 2024, how many drug-related incidents happened?鈥� On the back end, the chatbot will convert your question into code and provide a clear answer in plain English.聽

What impact do you hope the dashboard will have in Lawrence?

The dashboard is a learning tool. We want residents to use it to gather information to take to their representatives and present it as evidence for where change might be needed. If community members are getting some value out of it, I think that鈥檚 good enough for us.

A lot of people might not immediately see the connection between AI, data science, and social work. Where do you see that connection?

I think there鈥檚 a big connection. I鈥檓 a huge proponent of making things accessible for people, and I think AI is making things easier for people to understand.

For the dashboards that we鈥檝e created, community members have asked whether we can design chatbots to go with them. They鈥檝e told us that they don鈥檛 necessarily understand what the numbers on our dashboards mean, so having a chatbot answer questions about these figures helps people understand what they鈥檙e seeing.

As long as we鈥檙e using reliable data and not providing access to people鈥檚 private information, AI can help people who aren鈥檛 technical learn more about their community and have a bigger social impact.

AI and data science raise concerns about bias, privacy, and fairness. How do you approach those issues?

I鈥檓 a huge proponent of keeping humans in the loop. We can鈥檛 have AI do all of our work鈥攚e need to have people evaluating its output as we go.

Here鈥檚 an example: We have a dashboard that鈥檚 pulling data from a range of websites for nonprofit organizations in Lawrence. On the back end, we鈥檙e using Gemini to summarize the information.

But we don鈥檛 trust everything that Gemini is gathering, so we have a student who鈥檚 evaluating 15 percent of its output. We identify where the tool is performing poorly and design better prompts to improve its performance. I think there needs to be a human involved. We can鈥檛 leave everything up to the machine.

What excites you most about the future of data science, social work, and community-engaged research?

Oh, I鈥檓 so excited. I see the world as an oyster. People are now able to use AI tools in their daily lives to better understand data, ask more questions, and question things that they otherwise might have taken for granted.聽

How do you see these tools shaping social work practice and research over the next decade?

There are so many avenues that haven鈥檛 been looked into yet. But these AI tools are providing an inroad for us to explore data in new ways. There鈥檚 so much information out there鈥攁nd only so much that people can do.

AI tools have made our lives so much easier. A task that once took a month now takes a week. Things have gotten faster鈥攁nd I鈥檓 very excited about how these tools will continue to evolve.

What gives you confidence about where these technologies are headed?

I鈥檓 very excited for the future, and I鈥檓 an optimist. I feel like there鈥檚 so much good that can come from using these technologies. And I can鈥檛 wait for people to keep exploring them and use them for good.

This story is part of our series highlighting how students, faculty, and staff at AV研究所SSW are putting AI into practice across teaching, research, and clinical training.