Standards-Based Grading in Undergraduate Courses for Technology Majors

To better support students and enhance learning outcomes, the authors began exploring various alternative grading methods at their small private liberal arts university. Their goal was to identify strategies that would allow students to demonstrate mastery of course content while improving student retention – all while maintaining academic rigor and promoting fairness and accuracy in grading. Alternative grading offers several mechanisms to achieve these aims, including decoupling learning from grading, encouraging students to engage with instructor feedback, promoting learning through mistakes, and providing opportunities for reassessment without penalty.

This paper aims to have broad applicability stemming from the diversity of the author’s approaches. Over the past two years, the authors independently transitioned from traditional grading systems to a standards-based approach. Standards Based Grading (SBG) emphasizes small, re-takeable formative assessments aligned with specific learning objectives, as opposed to large, summative evaluations. The transitions varied among instructors: some fully adopted the new model immediately, while others introduced it gradually. In some cases, the revised courses were new preparations for the instructors, whereas in others, the same courses had previously been taught using traditional grading methods. The range of courses included skills-based introductory computer science classes, several “information-based” courses in computer science and cybersecurity, theoretical upper-level computer science courses, and hands-on data science courses.

Traditionally, most departmental courses assessed students with approximately 50% of the grade based on “out-of-classroom” work (such as homework and projects), and the remaining portion based on “in-classroom” assessments (such as quizzes and exams). However, with the emergence of Large Language Models (LLMs) and Generative AI tools like ChatGPT, the authors began to reconsider the emphasis placed on out-of-classroom assessments. Recognizing that these tools are easily accessible and may influence how students engage with assignments, the authors sought to ensure that students were developing a deep understanding of the material. This led to a shift in grading weight toward in-class assessments, where instructors have greater control over the evaluation environment. Standards-Based Grading (SBG) offered an effective solution, enabling the use of frequent, low-stakes, in-class assessments while moving away from high-stakes, one-time exams.

Although the weight of out-of-classroom work was reduced, homework and projects remained integral to the curriculum. These assignments continued to support learning outside the classroom and allowed for the assessment of more complex or integrated skills that do not fit neatly into a single learning objective. In addition, homework provided students with early exposure to the types of questions used in standards-based assessments, incentivizing engagement with the course material. Importantly, the authors emphasize that standards-based grading does not need to be applied uniformly across all components of a course. Certain aspects such as exams or skill-based assessments can follow the SBG model, while others, like projects or homework, may retain traditional grading structures or even a different alternative grading approach. This hybrid approach allows instructors to tailor assessment strategies to the goals of the course while also allowing them to experiment with ways to better support student learning.

Mastery-Based Grading is an assessment system that provides students with a clear set of objectives and evaluates them based on their eventual mastery of the material [5]. Unlike traditional grading models, Mastery-Based Grading allows students to submit work, receive feedback, revise or retry the assignment, and resubmit it without penalty. A specific implementation of this approach is Standards-Based Grading (SBG), which typically involves low-stakes formative assessments aligned with clearly defined course objectives [19, 11]. Rather than focusing on earning a particular percentage, students aim to demonstrate mastery of specific learning goals. As a result, assignments are often evaluated on a “pass/fail” basis or using a standards-based rubric that emphasizes levels of mastery (e.g., emerging, proficient, exemplary) instead of traditional point values, with opportunities for revision and resubmission until mastery is achieved.

A key feature of SBG is its shift away from infrequent, high-stakes assessments, such as major exams that offer only a single chance to demonstrate learning, toward more frequent, lower-stakes evaluations. These ongoing assessments provide students with multiple opportunities to show understanding over time. Consequently, this model is often associated with reductions in student stress and test anxiety [14, 17, 12, 4]. Furthermore, the ability to revise and reattempt assessments encourages a growth mindset [2], as students are empowered to view intelligence as something that can be developed through effort and persistence.

Student reflection on feedback is another cornerstone of the SBG approach. In traditional grading systems, students often glance at their score and overlook the instructor’s feedback, missing valuable opportunities for learning and improvement. When the emphasis is solely on the grade, feedback is frequently ignored, which hinders growth [13]. In contrast, SBG requires students seeking to improve their performance to actively engage with instructor feedback. This reflection is necessary before they revise or resubmit their work, reinforcing the learning process.

Although numerous studies have explored the implementation of SBG in mathematics [5], relatively few have examined its use in computer science education. One such study by McCane et. al. [15] compared students in a Java course who had previously taken either a mastery-based or traditional Python course. The findings indicated that students from the mastery-based course performed better, particularly those who had previously struggled. Notably, high-performing students were unaffected by the grading model. However, a longitudinal follow-up by the same authors [16] found that the absence of deadlines for mastery-based assessments led to increased procrastination. Additional research supports the claim that mastery- and standards-based grading better supports lower- and middle-performing students in computer science contexts [18].

The CS Equitable Grading Practices group has also contributed to the field by creating a community of practice that shares resources relevant to a variety of grading models, including specifications grading, which shares similarities with SBG [7]. Moreover, a study by Fine [9] that explored the integration of competency-based grading with equity-focused strategies and Universal Design for Learning found that this hybrid model fostered increased student engagement and deeper understanding through more frequent interactions.

Recent scholarship also underscores the need to revisit assessment practices in light of the growing influence of AI tools. Educators are increasingly concerned with issues related to identity verification, plagiarism, and the assurance of authentic learning [10]. They must remain vigilant in upholding the core objectives and principles of education [6]. One study concluded that “Assessment was seen as needing to shift towards more in-person and invigilated tasks that involved greater authenticity, personalisation, higher-order thinking, and disclosure of sources” [3]. This affirmation of the need for reform has further validated the authors’ ongoing application and study of standards-based grading.

In general, the instructors implemented SBG in courses where learning objectives could be clearly articulated and broken down into specific, measurable standards. The instructors who adopted this approach teach a diverse range of courses across the department, including mathematics, computer science, cybersecurity, and data science. This paper primarily focuses on the computer science and data science courses in which SBG has been applied. The courses and standards included for this work are:

Lower-Division Courses (1st and 2nd Year):

• $\mathrm{■}$

  Introduction to Computer Science I and II (Python and Java respectively)

• $\mathrm{■}$

  Data Structures and Algorithms

• $\mathrm{■}$

  Principles of Software Development

• $\mathrm{■}$

  Introduction to Data Science

• $\mathrm{■}$

  Machine Learning

Upper-Division Courses (3rd and 4th Year):

• $\mathrm{■}$

  Computer Security

• $\mathrm{■}$

  Operating Systems

• $\mathrm{■}$

  Database Management Systems

• $\mathrm{■}$

  Algorithms

• $\mathrm{■}$

  Theory of Computation

Introductory computer science courses typically run five sections per year, each enrolling 20–25 students. Since different instructors teach these courses each semester, maintaining consistency in course content is essential for ensuring students’ continued success in subsequent classes. Creating a shared set of standards that all instructors can use – while still allowing flexibility in how they implement them – strikes a balance between academic freedom and curricular consistency. This approach offers a clear record of the material covered, which is valuable for instructors teaching later courses in the sequence.

A small set of examples of the standards and some assessment questions used for the courses in this paper can be found in Appendices A and B. The authors have created a repository of the standards from courses they have taught that can be found online at figshare – SBG in Technology Courses.

An important avenue of evaluating student understanding and mastery in a standards-based course is with low-stakes re-takeable assessments, which are referred to as standards. The design of the standards aims to target one specific learning objective or a lesson. The format varies according to the material. Within the department they have ranged from interpreting code/results, having the student program by hand, demonstrating understanding of theoretical concepts, or coding a solution up on the computer. These standards were usually based on the material covered in the previous week’s lecture and homework assignments, allowing students time to interact with the content, practice, and then be assessed on the material.

The implementation and set-up of alternative grading can be tailored to suit the preferences and style of each instructor. In the following sections, the authors discuss different approaches to administering, grading, and weighting standards in the final grade, allowing instructors to combine various methods to create the most effective system for them. Different combinations of these approaches have been used in all of the classes taught by the authors.

With each course having 20+ standards that students can re-attempt multiple times, it is important to have an organized management system to track student progress. Most of the instructors used Excel sheets to determine which students had passed (or attempted) each assessment and how many attempts each one took. This was beneficial as instructors could update the learning management system once per week instead of updating it every time a student attempted a standard.

After two years of implementing standards-based grading in multiple sections, the authors have found several promising results as reflected in their retake statistics, student feedback, and classroom experiences.

One concern with allowing retakes is grade inflation. However, the authors did not observe this in their transition to SBG. The pass/fail nature of SBG also raises concerns of artificially lowering the grades. However, this was also not their experience. In fact, these two concerns tend to balance out. For example, in the Computer Security class, the same instructor has continuously taught 1-2 sections of the course for last seven years. The median final exam grade, shown in Table 4, has remained fairly consistent between the course prior to the implementation of standards and the course since standards have been added.

Looking at things more closely, pre-pandemic exam averages were 75 and 77. During the pandemic, the instructor switched to take-home, open-notes exams, with an attendant rise is scores. In 2022, exams were back in person, and there was a dramatic drop in scores. The instructor saw this drop across other courses as well. For the last two years, exams have been in-person and the course has used SBG. Scores recovered to their pre-pandemic levels. The instructor has noticed a drop in the preparedness of students pre-pandemic to post-pandemic. SBG may have offset this drop.

This analysis provides confidence that using standards is not hindering students, nor is it inflating grades. But the authors also see this as evidence that supports greater confidence that students are mastering the learning outcomes since a large portion of the final grade is based on passing standards.

One instructor had standards in all three of their courses (an introductory CS course, an upper-level CS course, and an upper-level math course) and noted that 36.2% of the first attempts to pass a standard were correct and an eventual pass rate of 88.7%. Students averaged 2.3 attempts before getting it right or giving up.

In Fall 2024, one instructor recorded 63 students making 4684 individual standard attempts, of which 1735 (37%) were correct attempts. There was a 34.5% initial pass rate with an eventual pass rate of 88.1%. Students averaged 2.2 attempts until correct or giving up. The maximum number of attempts for one standard by one student was 11.

End of course evaluations, administered by the university, include the option for students to provide written feedback on the course. There were many comments related to the use of standards. Many students appreciated the use of standards, with comments such as “I liked how the standards allowed for an opportunity to better learn and grow” or “having these little moments to evaluate that add up are great”. Some students did not like standards commented that they “disliked the use of standards seeing it is difficult for me to code on paper”. Seeing this feedback from past semesters, the authors have made a point in current courses that this is an opportunity to practice a skill they are not strong at, and that allowance is made for simple mistakes that a compiler would note and students would be able to quickly correct. Additional feedback from students can be found in the Appendix.

Overall, students appreciated the opportunity to retake standards as many times as they needed to show that they had mastered the material. However, some students disliked the amount of assessments that were given, their pass/fail nature, and the fact that the failed assessments piled up for them in multiple classes.

Some instructors found that the pass/fail system was discouraging for beginning students who demonstrated that they knew most of the material but had not completely mastered it. Because of this, some instructors adopted a “half-pass” designation. Grading everything with this system allowed instructors to provide more nuanced feedback and encourage students who showed partial understanding to attempt retakes and improve their grades.

Some authors found that, while students don’t seem to study for the standards, especially in the entry-level courses, they do request to do retakes which eventually gets them to study. There is a high need to encourage students, sometimes in every class, to retake standards, announcing when those opportunities are to remind them to plan for the next chance and avoid procrastinating their retakes. There can be a negative impact on commuters, athletes, and other students whose schedules don’t align with the instructor’s availability. Consideration should be given to find a way to allow students the time to retake rather than being penalized for having a different schedule. One option to accommodate such schedules is to find class time where students can do retakes, such as if there is time after completing new standards or the odd day in the semester when it is not conducive (such as around breaks) to introduce or test on new material.

The authors acknowledge that standards-based grading may not be suitable for all courses, teaching styles, or learning environments. Alternative grading approaches can be equally effective depending on the course objectives and instructional context. For example, some of the authors have successfully implemented specifications grading in a general education math course, yielding positive outcomes. Another promising approach is ungrading or cooperative grading, which may be particularly well-suited for seminar-style or discussion-based courses. Exploring the effectiveness of these and other alternative grading models remains an important area for future research.

Overall, the authors appreciate this grading approach because it reduces the pressure on students to master all the material by a specific deadline – such as a major midterm – instead allowing them to demonstrate understanding over time. It also encourages students to engage with instructor feedback and revisit material they initially struggled with in order to earn a passing grade on an assignment.

Generative AI is having a profound impact on the way students approach their coursework. This impact may be positive as informal tutors, or even as formal tutors as in the case of GPT tutors for calculus courses among others [8, 1]. Many teachers are, however, familiar with the negative impacts of AI disincentivizing study and mental work. When the learning objectives in a course can be expressed as a set of problem-solving skills, such as in many technical courses, the authors highly recommend the use of in-class assessments to combat this.

While a strict mastery or standards-based grading model typically excludes homework, the authors believe that eliminating homework entirely would not serve most of their students well. Despite the possibility of students finding ways to cheat, homework remains a valuable form of practice that provides opportunities for feedback. However, it must carry enough weight in the overall grade to incentivize completion; if the value is too low, many students simply won’t do it.

This grading model tends to benefit mid-range students the most. High-performing students generally succeed regardless of the grading system, while students who typically struggle to earn a C may now be able to achieve a B or even an A. Lower-achieving students will still face challenges, though more of them may ultimately pass under this system. Additionally, it offers meaningful support for students who experience significant disruptions during the semester. For a student with a solid academic foundation, this approach allows them to recover and still perform well, rather than being penalized for circumstances beyond their control. It also encourages self-regulation, as students are responsible for monitoring their progress toward mastery and deciding when to reassess. The authors have found SBG instills a growth mindset, prevents cheating, rewards study, and, most importantly, has increased student learning as discussed in the results. Future work will focus on collecting more comprehensive data to evaluate the long-term impact of standards-based grading on student outcomes, as well as gathering student perceptions to better understand their experiences with this model.