Practical Example of Mastery Grading in a Data Science Class

In a paper from the 2022 Proceedings of Machine Learning Researchhttps://proceedings.mlr.press/v170/brown22a/brown22a.pdf, Sarah Brown discusses revisions she made in her Programming for Data Science course aimed at making it student centered. The course is designed as a programming intensive data science course and serves approximately 30 students. This paper summarizes the design overall and provides practical details about the instruction via participatory live coding and assessment with a competency-based grading scheme.

The syllabus for the course is listed here: https://rhodyprog4ds.github.io/BrownFall21/syllabus/

The learning outcomes are centered around the specific programming skills students are expected to master for the course:

Learning Outcomes

By the end of the semester

(process) Describe the process of data science, define each phase, and identify standard tools
(data) Access and combine data in multiple formats for analysis
(exploratory) Perform exploratory data analyses including descriptive statistics and visualization
(modeling) Select models for data by applying and evaluating mutiple models to a single dataset
(communicate) Communicate solutions to problems with data in common industry formats

We will build your skill in the process and communicate outcomes over the whole semester. The middle three skills will correspond roughly to the content taught for each of the first three portfolio checks.

Students are provided with several mappings that explain how the different levels of mastery are defined, as well as how they are mapped throughout the course based on the assignments and projects required in the class.

From the syllabus: https://rhodyprog4ds.github.io/BrownFall21/syllabus/achievements.html

Table of Achievement Definitions — The table below describes how your participation, assignments, and portfolios will be assessed to earn each achievement. The keyword for each skill is a short name that will be used to refer to skills throughout the course materials; the full description of the skill is in this table.

Table of Achievement Definitions continued — The table below describes how your participation, assignments, and portfolios will be assessed to earn each achievement. The keyword for each skill is a short name that will be used to refer to skills throughout the course materials; the full description of the skill is in this table.

Assignments and Skills — Using the keywords from the table above, this table shows which assignments you will be able to demonstrate which skills and the total number of assignments that assess each skill. This is the number of opportunities you have to earn Level 2 and still preserve 2 chances to earn Level 3 for each skill.

As well as how Mastery translates into a letter grade.

From the syllabus: https://rhodyprog4ds.github.io/BrownFall21/syllabus/grading.html

Detailed Mechanics of Grades — On Brightspace there are 45 Grade items that you will get a 0 or a 1 grade for. These will be revealed, so that you can view them as you have an opportunity to demonstrate each one. The table below shows the minimum number of skills at each level to earn each letter grade.

In terms of revision, Brown states: In my data science course, instead of having students revise a lot, I designed the assignments to have repetition of the skills (basically mini learning outcomes) employed and based the grading on the skills, not the assignments. Then students use feedback from assignment “n” on assignment “n+1” to achieve a given skill that is assessed in both [assignments]. This works because the content in the course builds and is very tightly coupled, so this [approach] may not work in every topic. All skills repeat at least twice, and some more. If a student needs more attempts at a given skill they can do a reflective revision in their portfolio, however, most do not.

Leave a Reply Cancel reply

Subscribe By Email