This tool helps you analyze the diversity of assigned authors on course syllabi
It takes a list of assigned readings, and estimates the proportions of authors by gender and race/ethnicity.
For details of the methodology and results for urban sustainability courses, see:
Millard-Ball, Adam; Desai, Garima; and Fahrney, Jessica (2022). "Diversifying Planning Education through Course Readings." Journal of Planning Education and Research, in press.
Note that the estimates in the published paper use ethnicolr 0.4.0. This current version uses 0.8.1, with slightly different results.
You can also run the tool online at www.syllabusdiversity.org.
Python 3.8 is required, along with the following Python packages. Other versions may also work, but have not been tested.
numpy 1.19.2
pandas 1.3.5
ethnicolr 0.8.1
gender_guesser 0.4.0
Flask 2.0.3
h5py 3.1.0
Keras 2.2.4
tensorflow 2.5.2
Install them with:
pip install numpy==1.19.2 pandas==1.3.5 ethnicolr==0.8.1 gender_guesser==0.4.0 Flask==2.0.3 h5py==3.1.0 Keras==2.2.4 tensorflow==2.5.2
Prepare your file with the list of authors or assigned readings using one of the two provided templates: template1.csv or template2.csv. Both .csv and Excel (.xlsx) formats are accepted, but for Excel, only the first sheet will be used.
In both cases, authors' names must be comma-separated, e.g. Agyeman, Julian. Authors without a comma will be ignored as they are assumed to be an institutional author (e.g. United Nations).
template1.csv has one row for each reading. The (optional) courseid column enables you to compare results for different courses. It can be numeric or the name of a course. If you just have one course, you can omit this. The reading column is a citation in the following format: a list of authors (last name, first name) separated by semicolons, with ** after the last author. For example:
Agyeman, Julian; Bullard, Robert; Evans, Bob** 2003. Just Sustainabilities: Development in an Unequal World. Cambridge, MA: MIT Press.
Everything after the ** is ignored
template2.csv has one row for each author. The readingid column is an identifier for each reading. This allows multi-authored publications to be weighted appropriately. The (optional) courseid column enables you to compare different courses. If you just have one course, you can omit this. The author column is the author's name in (lastname, firstname) format. For example:
Agyeman, Julian
Bullard, Robert
In the command line, type:
python analyze_readings.py inputfilename.csv [outputfilename.csv]
Summary statistics for each course will be displayed. If you include the optional outputfilename.csv, a new file will be created with the estimated gender and race/ethnicity for each author or reading.
Note that the estimates for race/ethnicity are probabilistic, i.e. the probability that an author falls into a given category. This works best with large numbers of authors.
We would love to hear about how this tool is being used. If you have any problems, please contact Adam Millard-Ball or open a GitHub issue.