Do Deep Neural Nets Display Human-like Attention in Short Answer Scoring?

Abstract

Deep Learning (DL) techniques have been increasingly adopted for Automatic Text Scoring in education. However, these techniques often suffer from their inabilities to explain and justify how a prediction is made, which, unavoidably, decreases their trustworthiness and hinders educators from embracing them in practice. This study aimed to investigate whether (and to what extent) DL-based graders align with human graders regarding the important words they identify when marking short answer questions. To this end, we first conducted a user study to ask human graders to manually annotate important words in assessing answer quality and then measured the overlap between these human-annotated words and those identified by DL-based graders (i.e., those receiving large attention weights). Furthermore, we ran a randomized controlled experiment to explore the impact of highlighting important words detected by DL-based graders on human grading. The results showed that: (i) DL-based graders, to a certain degree, displayed alignment with human graders no matter whether DL-based graders and human graders agreed on the quality of an answer; and (ii) it is possible to facilitate human grading by highlighting those DL-detected important words, though further investigations are necessary to understand how human graders exploit such highlighted words.

Automatic Short Answer Grading for Finnish with ChatGPT

“…ASAG research, spanning three decades (Burrows, Gurevych, and Stein 2015), remains aligned with current NLP developments, as many ASAG systems now employ deep learning techniques (Bonthu, Sree, and Prasad 2021; Haller et al 2022). Additionally, recent advancements include generalizable ASAG, wherein models are trained to generalize to target domains that do not overlap with their training domain (Zeng et al 2023). Aside from systems primarily centered on score prediction, other research directions encompass using machines as a second grader (Kulkarni et al 2014), resorting to human evaluators for more challenging questions (Li et al 2023), exploring adversarial attacks on grading systems (Filighera, Steuer, and Rensing 2020;Filighera et al 2022), and generating explainable predictions (Tornqvist et al 2023), among other areas of investigation.…”

The future of standardised assessment: Validity and trust in algorithms for assessment and scoring

“…These new tools have enabled researchers to probe black‐box systems by comparing machine and human explanations. For instance, Zeng et al (2022) used a dataset of 60 answers provided by US Grade 10 pupils in subjects like English and Science to compare human and machine annotations. The researchers analysed whether several high‐performing ML systems agreed with a sample of 20 academics on the most important words in the pupil responses that determined the final mark (which could go from zero to three).…”