What Is Best for Students, Numerical Scores or Letter Grades?

Evi Micha, Shreyas Sekar, Nisarg Shah

Introduction

Traditionally, grades are viewed as a final measurement of learning—a static report card. However, educational research suggests that grades also function as a critical input: the feedback received on one assessment directly influences a student's motivation and effort for the next. This paper introduces a formal model to investigate how the precision of grading schemes—specifically numerical scores versus coarse letter grades—impacts long-term student performance through psychological feedback loops.

The Motivational Model

The authors' framework is built on two psychological pillars:

  1. Anchoring: Students maintain an internal sense of their "intrinsic quality" ($q$) and expect their performance to align with this self-perception.
  2. (De)Motivation: Deviations from expected performance trigger emotional responses. A grade higher than intrinsic quality provides a motivational boost ($\alpha_m$), while a lower grade leads to demotivation ($\alpha_d$).

The study compares Numerical Scoring (NS), which provides precise feedback, to Uniform Letter Grading (ULG), where scores are grouped into buckets. To model student perception, the authors use a "midpoint scheme": a student receiving a letter grade representing a range (e.g., 80-90) interprets it as the midpoint value (85).

Short-Term vs. Long-Term Dynamics

The research reveals a striking "horizon effect" where the optimal grading system depends on the number of evaluations.

The Short-Term Case (2-4 Evaluations)

In scenarios with few evaluations, such as a midterm and a final, numerical scoring is generally superior. This is particularly true under the realistic psychological assumption that "bad is stronger than good" ($\alpha_d > \alpha_m$). In these cases, numerical scores minimize "unearned" demotivation that can occur when a score is rounded down to a lower letter grade midpoint.

The Long-Term Case (25 Evaluations)

The most surprising finding emerges in long-term scenarios, such as weekly quizzes. Over 25 evaluations, numerical scoring can lead to a steady decline in student quality if demotivation outweighs motivation. Conversely, coarse letter grades (e.g., a 4-bucket system) exert a "stabilizing effect." Once a student's performance reaches the midpoint of a grade interval, the coarseness of the scale acts as a buffer, preventing minor fluctuations from triggering a downward spiral of demotivation.

Key Results

Performance comparison between Numerical Scoring (NS) and Uniform Letter Grading (ULG) for different evaluation counts and motivation levels. Source, p. 7

Performance comparison between Numerical Scoring (NS) and Uniform Letter Grading (ULG) for different evaluation counts and motivation levels.

Theoretical analysis and simulations of 5,000 students demonstrate that:

  • Joint Symmetry: In a two-test setup, numerical scoring outperforms any uniform letter grade when $\alpha_d > \alpha_m$.
  • The Buffer Effect: In long sequences with high demotivation sensitivity ($\alpha_m = 0.2$), students under a 4-grade system ($ULG_4$) maintained significantly higher final quality than those under numerical scoring, whose performance eroded over time.
  • Motivation Sensitivity: When motivation is high ($\alpha_m = 0.8$), numerical scoring remains the dominant strategy as it maximizes the positive feedback loop.

Conclusion and Future Directions

While numerical scores offer precision, they also expose students to constant, potentially demotivating fluctuations. Coarse letter grades, by contrast, provide a stable anchor that can protect students from long-term performance degradation.

Future research could explore non-uniform grading buckets (e.g., smaller ranges for high grades) or personalized grading schemes tailored to an individual student's sensitivity to feedback. The current model also assumes a linear update rule for quality, which could be refined using Prospect Theory to account for the non-linear psychological impact of crossing grade boundaries.

Summary updated

Abstract

We study letter grading schemes, which are routinely employed for evaluating student performance. Typically, a numerical score obtained via one or more evaluations is converted into a letter grade (e.g., A+, B-, etc.) by associating a disjoint interval of numerical scores to each letter grade. We propose the first model for studying the (de)motivational effects of such grading on the students and, consequently, on their performance in future evaluations. We use the model to compare uniform letter grading schemes, in which the range of scores is divided into equal-length parts that are mapped to the letter grades, to numerical scoring, in which the score is not converted to any letter grade (equivalently, every score is its own letter grade). Theoretically, we identify realistic conditions under which numerical scoring is better than any uniform letter grading scheme. Our experiments confirm that this holds under even weaker conditions, but also find cases where the converse occurs.
BibTeX
@inproceedings{micha2024what,
  author = {Evi Micha and Shreyas Sekar and Nisarg Shah},
  title = {{What Is Best for Students, Numerical Scores or Letter Grades?}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2024},
}
Figure