Key points are not available for this paper at this time.
Abstract Research on AI in language education has not yet addressed how cognitive–symbolic reasoning can support L2 poetic interpretation, leaving a gap in the literature at the intersection of neurocognitive meaning construction and AI-assisted literary learning. This study examines the perceptions of AI-mediated poetry interpretation held by L1 Arabic students of L2 English. Specifically, we investigate whether students differentiate between AI’s computational pattern-matching capabilities and the embodied, culturally grounded interpretive processes required for cross-linguistic poetic comprehension. A quantitative survey was administered to such students in Saudi Arabia and Kuwait using a Likert-scale instrument that comprised six-constructs. The results indicated that the students perceived AI as minimally effective in phonotactic–prosodic detection, figurative mapping and structural–stylistic tracking. Its interpretive consistency was thus described as either hallucinatory, partially useful or only conditionally acceptable. Categorising the perceptions of usefulness revealed differences between learner types: strategic, metacognitively regulated learners maintained interpretive depth, whereas heavily AI-reliant learners demonstrated conceptual borrowing, shallow synthesis and less originality. The paper proposes integrating Jacobs’ neurocognitive poetics with frame semantics can support layered, self-regulated interpretive reasoning. It also explains how poetic interpretation may be distributed between human cognition and AI architectures. Therefore, L2 poetry instruction must explicitly train students to distinguish computational affordances (pattern detection, structural analysis) from interpretive limitations (cultural grounding, embodied resonance, intertextual authentication). The validated measurement model, quantifying student perceptions across distinct analysis dimensions, pedagogically, defines how AI assisting requires supplementation.
Alqarawi et al. (Sun,) studied this question.