Teaching Responsible Data Science in Urban Spaces

Responsible Data Science in Urban Spaces

In July, Dr. Anat Caspi and I began a pilot course called Responsible Data Science in Urban Spaces. The intent of the class was to teach students the fundamentals of GIS, and how to use spatial technology ethically. Students would learn how to create spatial data driven applications to solve problems in civic analytics, especially around equity, transportation, and navigation. We wanted students to utilize data and processing from the AccessMap Walkshed generator, and show that this data could be used for a variety of applications. It was a priority to teach students how to use spatial data responsibly, mitigating bias and understanding the impact of big data. My role was to cover the GIS fundamentals via lectures and the open source GIS program QGIS. I would like to share some of the successful elements of my part of the course, and the projects our students worked on, as well as the lessons I learned teaching for the first time.

Course Design

My portion of the class consisted of 4 lectures on GIS fundamentals, and 3 lectures specifically on QGIS Most of the GIS lectures used QGIS to illustrate particular components. Most of the lectures had an accompanying QGIS lab on the material, which was adapted from existing teaching resources or written from scratch.

1. What is GIS: illustrates applications and definitions of GIS
2. Geodesy: taught the basics of coordinate reference systems, datums, and projections
3. QGIS Plugin/Connections: taught how to use the QGIS plugins and connect QGIS to databases
4. Attributes/Data Models: taught about raster/vector data models and attribute tables
5. GIS Operations: taught basic GIS operations such as overlays, joins, and buffers in QGIS
6. Project Specific Operations: students sent in questions on operations for their particular projects
7. QGIS Model Builder: using the QGIS model builder for complex and repetitive tasks.

Projects and Students

Students worked in groups to develop data-driven applications for civic analytics and engagement. Each project consisted of a proposal, an analytics plan, and a poster. The second quarter of the class was devoted to weekly student meetings to guide students through these projects. We tried two different methods of project selection. For the first cohort, students were allowed to create a project and question to work from. We decided it would be more efficient and directed to have students choose from a list of current lab project objectives for the second cohort. I have chosen two projects from each group as illustrative examples.

Cohort 1 (Highschool Level); Spatial Access to Healthcare: This project aimed to identify if different populations in Seattle and Bellevue have unequal spatial access to medical care, and if that variation is significant. Students found the best way to define ‘spatial access’ and measured this metric for neighborhoods in Seattle. They conducted a variety of statistical tests to determine the significance of the relationship. No significant relationship was found between median income and driving time to a hospital. This project was presented at a TCAT symposium in December 2023. These are the key visualizations:

Cohort 1 (Highschool Level); Frequent Transit and Income This project used GTFS data to identify frequent public transit stops in Seattle, and aimed to identify any relationship between income and access to these stops. Students successfully wrote scripts to classify transit stops, and used QGIS to visualize them. Students used TCAT walkshed scripts to identify the accessible reach from each station. The second question is still under development. This project was presented at a TCAT symposium in December 2023. The figures below show an example of a transit stop walkshed, and a map of frequent stop distribution.

Figure 1: QGIS Visualization of Frequent Transit Stops (indicated by purple circles)

Figure 4: Walkshed of One Stop

Cohort 2 (Undergraduate Level); What would it cost for Seattle to be fully accessible?: The project goal was to find and cost sidewalk issues and bottlenecks in Seattle. This would identify priority areas for improvements. Students in this group identified census tracts with greatest improvement needs, and created scripting to identify specific improvement locations on sidewalks. The figure shows how the cost function can be used to show sidewalk sections in need of the most urgent attention.

Figure 3: Using the new cost field to visualize sidewalk issues in the walksheds. Only sidewalks with observations included in the cost function are highlighted.

Cohort 2 (Undergraduate Level); Optimizing Seattle Curbside Disability Parking Spots: The project goal was to identify and analyze current disability parking placement and develop an optimization function for new placement in Seattle. Students identified the areas reachable from current parking spots using AccessMap scripts, and are using this information to deduce what locations are still inaccessible or under-accessible. This project made excellent use of the TCAT walksheds tools in a larger project application. The figure shows an example of how disability parking spots in downtown Seattle have access to ballot drop boxes.

Figure 2: Joint walksheds map with disability parking locations as start points

Teaching Lessons

Having not taught a defined college course before, I learned a great deal about what is required to teach effectively. Here are a few things I learned teaching CSE 495:

1. Make sure to explicitly define deadlines, grading, and the syllabus. Students were confused on due dates or how material was graded. A more explicit syllabus schedule and grading breakdown is easier for students and the teaching staff.
2. Make sure students feel confident reaching out about problems between classes. Students would sometimes get caught on issues resolvable with some basic research skills or a simple question for an entire inter-meeting period. This slows down project development.
3. Create in-class examples ahead of time and gather necessary information. There were several class sessions where I didn’t know what datasets to use for class examples on the spot, or the processing of said example failed in class. A lot of class time was wasted as a result; these should be planned and executed in advance.
4. Balance instruction slides with interactive examples. Slides are less helpful for QGIS material than in-software examples. It is difficult to structure a lecture linearly without instructive slides.
5. Create extensive documentation of project status, issues and objectives.