DSCI 310 Data in Action S26 – Bryn Mawr College

Faith Somers, BMC ’27 & Alex Delgado, HC ’28

Praxis Course: Data in Action: Non-Profits and Data

Semester: Spring 2026

Faculty Advisor/Professor: Jennifer Spohrer

Community Partner: PHENND/CASE

Praxis Site Supervisor: Hillary Kane and Janine Wright

Praxis Poster:

DSCI Faith Somers_Summer Melt Praxis Project Poster Final

Further Context:
This semester, we had the opportunity to work with the Philadelphia Higher Education Network for Neighborhood Development (PHENND), a network of over 25 colleges and universities that focuses on “service-learning, civic engagement, and community partnership in Philadelphia” (phennd.org).

The organization has many diverse programs relating to achieving these goals in the Philadelphia region (College Success Network, PHENND Sustainability, Democratic Civic Engagement, etc.). Our project has focused on college completion and success strategies, specifically on reducing summer melt in the transition from high school to college. Another
organization we worked with is the College Access and Success Ecosystem (CASE), a collective impact forum focused on increasing high school and college graduation rates for Philadelphia students. Our two partners at PHENND serve as the co-chairs for CASE’s On Track to College Graduation Action Team.

To begin our project, we spent some time on a literature review, investigating the most relevant and recent literature on National Summer Melt, to plan out how to shrink the scope to the Philadelphia area. They had defined Summer melt as, “the phenomenon of college-intending
students who have applied to, been accepted by, and made a deposit to a college or university, but fail to matriculate to that college (or any other) in the fall following their high school graduation”(MCAN). We were unsure whether this definition fully captured every institution’s protocol when considering the various dates a student could be considered melted or if an
enrollment deposit was the determining factor.

We wanted to broaden our understanding of institution-based Summer Melt, so we applied our learnings in forming qualitative questions for our prospective institutions. We brought the valuable insights from involved faculty within our communities, as well as the insights from weekly meetings with our PHENND partners, to begin our outreach to Philadelphia colleges and universities.

The outreach we performed was incredibly helpful, both for our project and for our individual learning. We were able to create a set of questions that we asked each institution that explored the way they thought about and handled summer melt. At first, we kept each of these survey sets in a document, and then later transferred all the information into a spreadsheet, which allowed us to explore the ways different types of institutions handled summer melt through visualizations. We identified that some of the top risk indicators included that a student had not planned out their bill payments and had not attended orientation, despite signing up.

We defined Summer Melt using the institutions for which the definition was intended. In collecting the responses of each Philadelphia area institution we interviewed, we have come up with this definition:When a high school graduate demonstrates verified intent to enroll in a postsecondary institution but does not matriculate by a defined cutoff date (such as the first day of classes). We also thought to dig deeper into who does not fall into the category of what Summer Melt is, i.e., students who formally defer enrollment or students who do not demonstrate verified intent (enrollment deposit, for example).

On March 24th, we had an amazing opportunity to attend a CASE meeting where each of their action teams presented short summaries of their recent successes and plans. We were also able to hear from Katharine Meyer, an expert in evidence-based principles for reducing summer melt, who gave us incredibly useful information on some specific tactics like behavioral nudges and messaging, counselor-led summer outreach, peer or near-peer support, and summer bridge programs. We encountered many of these topics in our early research, and it was very helpful to learn more about the specifics from Katharine. Lastly, this meeting gave us critical insight into
the importance of highly communicative, cross-departmental work that is necessary to implement anti-melt strategies. The format of CASE and their meetings modeled the communication and commitment needed from colleges and universities looking to reduce their summer melt rates.

Another opportunity we were given was to speak with Jim Ramey at UPenn, who works closely with UPenn’s High School programs. We were able to have a meeting with him about the uses of Slate (online program for postsecondary onboarding/admissions processes) for reducing summer melt. He uses Slate to monitor student progress through post-deposit enrollment checkpoints. Ramey reads them as behavioral signals: students who stop engaging with checklist items after paying their deposit are flagged as high risk for melt and contacted directly. Our takeaways from this meeting were in line with the information we had gathered from all of
our other meetings with institutions. For example, Ramey emphasized UPenn’s previous issues with data disconnections across their different schools that have summer programs. There is often no centralized view of a student’s status. Additionally, he highlighted that effective melt prevention is labor-intensive, something we had also come across as a barrier for other institutions. Lastly, Ramey underscored the importance of human verification of checklist completion tasks. Students often accidentally turn in the wrong documents or miss a step of a task, meaning that automated completion indicators must be audited by staff. It was amazing to see the inside working of Slate, and this meeting helped us solidify the strengths and constraints of our project.

In the future, we would love to see more investigation into the summer melt practices at institutions beyond the Philadelphia area. With broader/more data from different institutions, it would be great to develop some practical data analytics that would solidify what works best to reduce summer melt. We were able to learn about an example of a model that identifies high-risk students based on their summer checklist completion progress. We will base a lot of our recommendations on how an institution might be best able to move towards a similar model. Lastly, we are hopeful that there could be some developments between relationships among high school counselors and college onboarding teams. The trust students have built with their
guidance counselors and their proximity to those individuals is very helpful in reducing summer melt, however, most counselors are not employed on a 12-month schedule, only a 10-month.

This project and Praxis course was incredibly helpful for our personal developments and skillsets. We both learned a lot about communicating with outside institutions and effectively gathering data. We are very grateful for this opportunity to work with PHENND and CASE.

Nina Hamilton, Nicole Huang, Krish Gupta, & Olivia Li

Praxis Course: DSCI 310: Data in Action   

Semester: Spring 2026   

Faculty Advisor/Professor: Jennifer Spohrer   

Community Partner:  The Barnes Foundation

Praxis Site Supervisor: Marie Edland and Liza Herzog

Praxis Poster:

DSCI 310 Nina_Hamilton_Barnes Poster

Further Context:

Hi! Our names are Nina, Nicole, Krish, and Olivia!

For our Praxis project, we are working with the Barnes Foundation as part of our DSCI B310: Data into Action course. The Barnes Foundation is an art institution based in Philadelphia, whose purpose is to support education in fine arts and horticulture.

For this project, we were given a large set of anonymized data from the Barnes Foundation and were told to conduct analysis of any kind. And so, our first step would be to design a study that:

1. Is impactful and useful for our partners
2. Can be completed within a semester (or less in our case)

We determined that our goal would be to analyze visitation trends and conduct additional cool analysis that would provide instructions to improve Barnes programming and engagement.

More specifically, we wanted to identify visitor-to-membership conversion trends, public programming and engagement, and visitation and membership trends with respect to location (zip codes).

One of our goals was to see which of the Barnes events were the most popular among members and non-members. While there were many events between 2022 and 2024, we focused on about 15 types of events present in the data (like group tours, free admission days, talks, etc.) that made up a majority of the data outside of general admission. We found that events which were free to the public (Free First Sundays and Barnes on the Block) had the biggest non-member visitors, which makes sense because it caters to groups that may not want to pay or afford to pay for regular tickets! This analysis was interesting because oftentimes members bought tickets for guests, and there were a lot of individual events that had to be grouped together in order to make these larger event types, so there was some nuance to what the number of people that attended each event meant. While it was difficult to organize the data into these groups, it was a good learning experience and provided interesting insights for our partners.

Another goal of this project was to understand where in the Philadelphia area Barnes visitors are coming from. The map visualizes visitation data across two dimensions, total visitor volume and membership share. Circle size reflects the number of total visitors from each zip code, while color indicates the percentage of those visitors who are members, ranging from warm red for low membership rates to deep blue for high. Unsurprisingly, the largest circles cluster around the zip codes in closest proximity to the museum. More revealing, however, is the color pattern. Several suburban zip codes show the darkest blue, indicating disproportionately high membership rates, while many zip codes within Philadelphia proper, particularly those with lower median incomes, tend to appear smaller and redder. This contrast suggests that Barnes has built strong loyalty among a suburban base but has room to grow both visitation and membership conversion closer to home. The aim of this analysis is twofold: to identify new areas the Barnes can target with outreach to drive first-time visits, and to spotlight communities where membership is already strong so the institution can invest in retention and deeper engagement.

We were interested in identifying patterns that lead to membership applications to the Barnes. And so, reflected by the top chart on the right, we analyzed the visitation count before a visitor decides to become a member, color-coded by the last event they visited before becoming a member. Surprisingly, most individuals who became members enrolled after their first visit.

To dig more deeply into specifics, we decided to study which programming event the members attended last, which led to their membership (chart on the bottom right). This analysis would help Barnes identify which events were most successful in promoting similar events in the future. The data is calculated by finding the percentage of individuals who became members among all those who attended the same program.

Like any project, our work has some limits. It is important to know them.

First, the data only showed if a visitor was a member on the day they visited. It did not show when they bought the membership. So when we say an event led to someone joining, this is just our best guess. A person may have bought a membership online days before. They may also have been moved by a visit many months earlier. We cannot know for sure. Our findings show useful patterns. But they do not prove what caused people to join.

Second, we did not have qualitative data to analyze. All of our data was numbers — ticket records, visit counts, and zip codes. We did not have interviews, survey answers, or visitor comments. For example, we can see which events came before someone joined. But we cannot hear in their own words what made them want to become a member. Qualitative data would help explain the story behind the numbers.

We want to be clear about these limits. This helps the Barnes team trust our findings in the right way. It also shows where the best next questions are.

Olivia Crolle (BMC ’27), Eden Raich (BMC ’27), Shalom Lencha (BMC ’27)

Praxis Course: Data in Action

Semester: Spring 2026

Faculty Advisor/Professor: Jennifer Sporer, Liv Raddatz

Community Partner: The Discovery Center

Praxis Site Supervisor: Bria Wimberly

Praxis Poster:

DSCI 310 Olivia_Crolle_Data in Action PRAXIS

Further Context:

The Discovery Center is part of the National Audubon Society through Audubon Mid-Atlantic, a non-profit dedicated to preserving and protecting birds across Maryland, Pennsylvania, and the District of Columbia through habitat maintenance. The Discovery Center is a space for visitors to engage with nature in a productive, educational way. It is all centered around the abandoned drinking water reservoir in the East Park. The East Park Reservoir is a shallow, closed-system body of water. Only 6-8 feet in depth, it is especially vulnerable and unable to cope with toxic algal blooms, which occur more frequently throughout the fall and summer months. These blooms can decrease the level of oxygen in the water and block out the sun, harming the over seven species of fish that reside in the reservoir. This has further harmful effects as birds that consume fish or even just have prolonged exposure to the blooms can be impacted by them. We were put in touch with Bria Wimberly to complete the coding on a water sensor that would monitor the conditions of the reservoir to predict and eventually prevent algal blooms in order to protect the park and the birds that the center is dedicated to.

Olivia’s work on this project involved keeping in communication with our contact at the Discovery Center: writing update emails, scheduling Zoom meetings, and delegating tasks. She acted as the go between for project progress, and ensured that the group met the proper deadlines. She was also in charge of part of the final deliverable: the infographic. She ensured that the product closely resembled the Discovery Center website, in order to maintain aesthetic consistency. She kept in mind what she learned in class about data visualization and presentation. The infographic will be displayed for the sake of visitors to the Discovery Center, who likely have very little or no exposure to the information. So, it had to be clear and informative in
general terms that could be easily understood. It also had to be visually engaging, which she achieved using the already contrasting colors of the Audubon Society website. She is happy with the clean, professional look that she achieved on Canva, and hopes that the infographic will be useful to the Discovery Center and its visitors. She learned a lot during this project, especially regarding the importance of frequent, transparent communication. Olivia enjoyed creating an infographic that was both informative and in line with the Discovery Center’s current graphic
design. She knows that she has come away with a lot of new and strengthened skills that will be valuable elsewhere, such as project management, graphic design, and communication.

For this project Eden was in charge of setting up the sensor coding, calibration and wiring. She worked on code provided from a previous intern, and made sure it successfully set up the sensors. The code was composed of four sections. The basic set up for the mayfly sensor and connection to Monitor my Watershed, and the set up for each of the three sensors. She had to rewrite most of the code, as there were some wiring issues which involved rebuilding the mayfly and changing the modes of connection. The coding involved using outlines provided by each of the sensor’s companies, and changing the variables to the actual sensor set up. She then had to write code to calibrate the PH and DO sensors and perform the calibration. She then compared the calibrated values to BMC Geology equipment to ensure they were providing accurate information. She also had to write a custom class for the DO sensor. There is no library provided by Mayfly to send this type of DO sensor data to Monitor my Watershed, so she had to make a
class which allowed it to do so.

Shalom focused on making the data from the Discovery Center’s sensors easier to access and use. Instead of relying on manually downloading files from Monitor My Watershed, she figured out how the platform retrieves its data behind the scenes and built a script that automatically pulls this information and converts it into a CSV file. This creates a foundation for
a live-updating dataset that can later be used for graphs, analysis, or integration into the website. One of the more interesting challenges was that the data wasn’t directly available through a simple link. Shalom explored the site’s network activity to understand how the data was being
requested and then replicate that process in Python. Since there wasn’t much real sensor data available at first, she tested the approach using sample datasets to make sure everything worked. Shalom also explored ways to display the data on the Discovery Center website. While
embedding graphs directly into Squarespace wasn’t possible, she found a workaround by linking to the live graphs hosted on Monitor My Watershed. This still allows visitors to view up-to-date information in a simple and accessible way. Working with her team also showed she the value of
collaboration and communication

Kripa Khatiwada, HC ’26 & Joao Hasse, HC ’28

Praxis Course: DSCI310: Data In Action

Semester: Spring 2026

Faculty Advisor/Professor: Jennifer Spohrer

Community Partner: Bryn Mawr Center for Career and Civic Engagement Center

Praxis Site Supervisor: Ellie Esmond

Praxis Poster:

DSCI310 Poster_Voter engagement

Further Context:

Since 2016, Bryn Mawr College has voluntarily shared its deidentified enrollment data with NSLVE (National Study of Learning, Voting, and Engagement), which matches it against publicly available voter files to understand voting patterns on college campuses. Our project, completed
in partnership with Bryn Mawr’s Career and Civic Engagement Center, set out to replicate this report, creating a locally hosted version that doesn’t depend on national funding or outside organizations. The demand for an independently run study grew from the recent scrutiny NSLVE and its parent organization, Tufts University’s Tisch College of Civic Life, has faced under the current federal administration over data privacy concerns. Our study responded by replicating the methods while addressing those concerns directly: conducting the study in-house rather than through an outside organization, communications and data shared solely among the team, and a public report to be generated later using fully anonymized data.

The idea was to use college-level demographic data and match it with publicly available data from Montgomery County across multiple election cycles. Soon after starting, we realized we couldn’t fully replicate the NSLVE methodology given our constraints. While NSLVE uses demographic and voting data submitted directly by colleges, we had access only to Montgomery County voter registration data and had to filter for registrants with Bryn Mawr campus box addresses to identify likely student voters. The enrollment records needed for a full replication were not available to us. Another limitation that NSLVE also acknowledges is that many students do a mail-in ballot in their home county when they are enrolled in Bryn Mawr, which leads to an undercount of engaged students. An additional constraint was that publicly available datasets pre-2026 were not clean enough to analyze—missing column headers, floating numbers, and inconsistent formatting meant we ultimately worked with the 2026 Montgomery
County data.

Our 2026 data came with significant limitations. Unlike the NSLVE, which includes variables like race, ethnicity, and education level, our Montgomery County dataset had none of those demographic breakdowns, leaving us with age and party registration as our primary analytical variables. Earlier datasets were also too inconsistent to use (missing column headers, floating numbers, and unclear formatting) and we had no way to match voter records directly to enrollment data. Critically, our dataset included only a single DateLastVoted field rather than separate columns for each election cycle, meaning we could not isolate who voted in which specific election. A student who voted in 2020 but not since would show 2020 as their last vote, while a student who voted in both 2020 and 2024 would show only 2024, making it impossible to cleanly attribute participation to any one cycle. Given this constraint, we used all available data to answer a broader question: had each student voted in at least one election since 2020? Working within those constraints, we calculated age as of Election Day from date of birth, binned students into age groups matching NSLVE’s categories, and built pivot tables tracking 2024 presidential and 2025 municipal election participation as best we could from the available data.

Of the 1,438 total Montgomery County registrants in the 19010 zip code, 1,206 (about 83.8%) had campus box addresses, suggesting they are likely students. Combining the 2024 and 2025 cycles, we identified 551 civically engaged students (45.7% of locally registered BMC students) who voted in at least one of those two elections. The 22-24 age group showed the highest 2024 presidential turnout at 25.9%, while the 2025 municipal election saw 37.3% participation overall. We found that 1,110 of 1,206 registered students (92%) had voted in at least one election cycle since 2020, though this figure should be interpreted cautiously, as it relies on a single DateLastVoted field rather than full voting history, meaning students who voted in 2020 or 2022 but not since would not appear in our count, and the figure may be somewhat inflated for recent cycles. All of these figures are known undercounts (students registered at out-of-county home addresses cannot be tracked) but the data demonstrates that students who do register locally are meaningfully and consistently engaged in democracy.

We documented every methodological decision and limitation so that the next team to take on this project will not face the same obstacles from scratch. The most important thing we learned is that missing and unclear data is not a dead end, it can be used to infer information and tell its own story about where gaps in civic infrastructure or data collection exist.

To complement the data analysis, we designed a survey on Qualtrics to understand voting behavior and preferences among Bryn Mawr students, with the goal of identifying areas where the Voting Education and Support office can better serve students and make the case for a polling station on campus. We used branching logic to tailor the survey experience to each student’s voting eligibility, status, and experience, and included international students to inform them about civic engagement beyond simply casting a ballot. Throughout the design process, we applied Data Action principles from class i.e. purpose, policy, biases, ground-truthing, and participatory ethics to ensure the survey was relevant to the entire Bryn Mawr population. The survey is set to be deployed on April 30, 2026 at the Montgomery County Voter Drive at Bryn Mawr College.

This project demonstrated that a locally produced voter engagement report is feasible, even with imperfect data. The biggest lesson we learned was that finding creative ways to work with the data you have, rather than waiting for ideal conditions, is itself a data science skill. Next steps in this project include using survey results to refine the Civic Engagement Center’s
outreach strategies and publishing a public-facing report in accordance with data privacy guidelines. We hope this locally produced mini-report serves as a foundation for Bryn Mawr to continue tracking student voter engagement independently, regardless of what happens to national data infrastructure, and that the groundwork we laid makes the next iteration of this project faster, cleaner, and more complete.

Abby Litchfield, HC ’26 & Clara Morton, HC ’26

Praxis Course: DSCI 310: Data in Action

Semester: Spring 2026

Faculty Advisor/Professor: Jennifer Spohrer

Community Partner: Philadelphia Higher Education Network for Neighborhood Development’s (PHENND) Climate Resilience Youth Council (CRYC)

Praxis Site Supervisor: Lane Frazee

Praxis Poster:

DSCI 310 CRYC

Further Context:

This semester, our team partnered with the Climate Resilience Youth Council (CRYC) supported by the Philadelphia Higher Education Network for Neighborhood Development (PHENND). CRYC is made up of 18 high school-aged youth from diverse communities and school types across Philadelphia. The program is focused on developing student’s climate change literacy and civic engagement. Another core piece of the curriculum is designing and supporting a climate resilience project focused within a student’s neighborhood.

This program has been running for several years however, until this point there has been no way to track and gather student feedback. Our partner wanted to develop a survey to understand student experiences within the program and implement feedback into future iterations of the program. Additionally, student feedback can help show the impact and reach of the program, ultimately helping CRYC obtain funding or grants.

We initially aimed for a mixed method approach in which we would design a survey to be given to the current cohort of CRYC students and combine this with a focus group to gain a nuanced understanding of students’ experiences. With the information gained from the survey and the contextualization from the focus group, we could revise the survey into a pre and post program assessment. Finally, we would provide CRYC with a plan for analyzing the results from these surveys so they could adjust the program as needed to future cohorts.

We focused on using Google forms to create a primarily likert scale survey for current students. With our partner we were able to isolate four categories that were most essential to understanding student’s experiences and takeaways. These categories were climate change literacy, civic engagement literacy, program experience, and leadership development. We chose to leave this survey completely anonymous to allow for respondents to feel comfortable giving positive and negative feedback.

We wanted to design a likert scale survey to present the current cohort of students with a way to express their growth in skills and knowledge that we were able to analyze clearly. We then included open ended questions to understand the reasons behind some of the trends found in the likert scale answers. Unfortunately, we were unable to conduct the focus group due to
funding constraints, so we ultimately incorporated some of the more vital questions into the survey that we plan to use for the beginning and end of the next cohort term.

This partnership experience emphasized for us the importance of clear communication and adaptability. Our collaboration and understanding of the needs of CRYC grew throughout the semester, and we feel confident that the surveys and resources we provide for them will sustain their data needs throughout future cohorts.