Anthony F. Botelho, Ashvini Varatharaj, Eric G. Van Inwegen, and Neil T. Heffernan. 2019. Refusing to Try: Characterizing Early Stopout on Student Assignments. In The 9th International Learning Analytics & Knowledge Conference (LAK19), March 4–8, 2019, Tempe, AZ, USA. ACM, New York, NY, USA, 10 pages. https://doi.org/10.1145/3303772.3303806

Introduction

Stopout

• a micro version of drop out
• students stopping early on in an assignment
• aka refusal to try
• Stopout is juxtaposed to persistence/grit

Research objectives:

(1) Student stopout after the first problem can be stochastically modeled as an exponential decay, but that this model fails to account for roughly half of the stopout that occurs on the first problem.

(2) Specific actions (immediately prior to stopout) by students correlate with different patterns of stopout over time.

(3) High stopout on the first problem correlates to low levels of self-reported confidence.

Data

ASSITSments 438 students who exhibit stopout

Refusal to try

• fit two exponential curves to the data
  • one on all the data
  • the second on just the 2nd through 10th problem
• the second model showed an almost perfect model fit to the data (.991) for the 2nd through 10th problem, but there were over twice as many stopout students as expected.

Categorizing Stopout Behavior

Possible previous behaviors:

• none, this their first problem
• Correct attempt
• incorrect attempt
• request help

__I appreciate that this project addresses such a common behavior that is so underresearched.

Cluster 2 exhibits a mismatch between confidence (low) and performance (high). These students also stopout before starting. Why do you think this combination exists? Perhaps their high performance is driven by only attempting fewer and/or easier problems. Should prior correctness be weighted by problem difficulty and the number of problems attempted to account for this possibility?

The paper concludes that future research should attempt to create causal links between behavior and stopout. How would you operationalize that? It seems like it would be difficult (or impossible) to run an experiment where both the dependent and independent variables are behaviors that are most likely the result of some latent mental state (lack of motivation, feelings of frustration, etc.). Do you have any ideas on how this might be studied?_