ESPLAT 2025 conference: Scientific thinking in challenging times

With implications for the design of adaptive learning tools, this study investigates whether confusion enhances learning performance and self-regulation in help-seeking interactions with ChatGPT. Two forms of help-seeking have been identified: instrumental help-seeking (e.g., asking for a clue), which is considered a self-regulated strategy, and executive help-seeking (e.g., asking for the answer), which is not. Confusion, as an epistemic emotion, can deepen cognitive processing when managed effectively. Interactive Digital Learning Environments (IDLEs) have shown potential in inducing and managing confusion effectively. However, unlike IDLEs, few studies, to our knowledge, have examined the role of confusion in self-regulated learning with ChatGPT.
Thirty-four university students participated in the study, randomly assigned to either an experimental group receiving ambiguous, confusion-inducing instructions or a control group with clear instructions. Initially, they read a short text and identified two lesser-known French figures of speech (préterition, hypallage). The experimental group received vague explanations, while the control group was given precise definitions and examples. Each participant could ask ChatGPT only one question from a predefined set of instrumental or executive questions. In the final task, they completed a multiple-choice test to determine whether the previously introduced figures of speech—or other figures—appeared in ten new sentences. Emotions were assessed before and after the task using a short version of the Epistemically-Related Emotion Scales (EES, Pekrun et al., 2016).
Surprisingly, there was no significant difference between the groups in self-regulated learning strategy use. However, as expected, the confusion group outperformed the control group in the final task and showed a significantly higher confusion score than the control group.
These findings suggest that while confusion may not directly influence self-regulation, it can enhance learning outcomes when properly leveraged. This research provides insights into optimizing confusion management in digital learning environments to improve learning effectiveness across various educational contexts.