Not all who avoid help are lost: The relations between help avoidance and performance in a digital learning platform

Abstract

Although digital learning platforms (DLPs) provide students with corrective feedback, hints, and scaffolds, learners differ substantially in whether and when they use these supports. While help seeking has been extensively studied, there is little consensus on how to define and operationalize help avoidance (i.e., when students bypass available supports like hints) in online settings, hindering efforts to reliably identify, measure, and address it. This study revisits an established operationalization of help avoidance using fine-grained interaction logs from middle school students (N= 348) in ASSISTments, a digital math learning platform. Building on prior work, we refine the measure by leveraging detailed temporal interaction data to characterize help avoidance as unassisted answer attempts by students with low prior knowledge before accessing available support. We then examine how these behavioral patterns relate to proximal outcomes (i.e., within ASSISTments) and distal math achievement (i.e., on a standardized math test). Our analyses identify distinct subgroups of students classified as help avoidant, suggesting that existing operationalizations combine qualitatively different behaviors, including patterns consistent with productive struggle and unproductive persistence. These findings demonstrate that help avoidance is not a homogeneous construct and suggest that more behaviorally nuanced measures are necessary to differentiate productive struggle from a genuine need for intervention, thus guiding adaptive support in DLPs.

Materials

• Full BibTeXapi.crossref.org

Cite

@article{martin2026not,
  author = {Martin, Natalia S. and Vest, Nicholas A. and Closser, Avery H. and Botelho, Anthony F.},
  title = {Not all who avoid help are lost: The relations between help avoidance and performance in a digital learning platform},
  journal = {Contemporary Educational Psychology},
  year = {2026},
  volume = {87},
  articleno = {102505},
  publisher = {Elsevier BV},
  doi = {10.1016/j.cedpsych.2026.102505},
  url = {https://doi.org/10.1016/j.cedpsych.2026.102505}
}