Abstract:

Many large, introductory computer science classes rely on high-stakes summative assessments, often during a university-designated examination period. Doing so is known to increase anxiety and negatively impact students’ mental health. On the other hand, frequent testing is known to aid in retention, a necessary process for introductory classes designed to familiarize major and non-major students with foundational knowledge. The semester-long “Introduction to Computer Science and Programming using Python” (6.100A) class at the Massachusetts Institute of Technology, designed to promote students’ well-being and facilitate memory formation, relies primarily on low-stakes formative assessments. In addition to biweekly problem sets and exercises associated with lectures, assessments include 30-minute microquizzes, which are usually delivered with a weekly cadence. However, by depending on formative assessments and allowing students to drop up to two (Fall) to three (Spring) of their worst performing quizzes without consequence, the course does not incentivize students to revisit previous trouble spots.
As such, this paper outlines the implementation of the “Mastery Quiz”: a culminating, low-stakes 45-minute final microquiz that is tailored to each student’s weak areas, determined by performance on previous microquizzes. This summative assessment adheres to class goals of promoting student well-being by using familiar question types, maintaining a similar length to previous microquizzes, and contributing the same amount to the students’ final grades as previous microquizzes.
To reduce instructor workload and increase feasibility for scaling, a large-language model is used to generate a question bank for the Mastery Quiz. The bank serves to standardize assessment difficulty and balance scalability with intentional use of artificial intelligence.
The initial findings from the study revealed that 91% of the students who took the Mastery Quiz believed that it contributed to their learning. 64% indicated that this final microquiz made them "feel proud" of their growth over the semester, while 51% felt that the assessment increased their confidence in their coding abilities. 90% of the students performed better on the Mastery Quiz than on previous microquizzes that tested the same content.

Keywords:

Technology, education, educational technology, personalization, computer science, artificial intelligence, higher education, reinforced learning, spaced learning.