Reconceptualizing Human-AI Language Interaction: A Psycholinguistic Account of Predictive Processing, Cognitive Load, and L2 Language Processing in Generative AI-Mediated EFL Learning

ABSTRACT : While generative artificial intelligence offers substantial opportunities for interactive secondlanguage learning, the psycholinguistic processes underlying learner interaction with AI-generated language remain underexplored. This study examines predictive processing, cognitive load, and L2 language processing in university-level EFL learning mediated by generative AI among 20 students at TED University. The available dataset consists of 20 questionnaire records containing self-reported English-language knowledge, perceived usefulness, ease of use, satisfaction, and interaction-related perceptions. Descriptive statistics were calculated for the available variables. Mean self-rated English knowledge was 4.05 for speaking, 4.25 for listening, 4.45 for reading, and 4.15 for writing on the five-point response scale. The questionnaire responses also showed generally positive perceptions of usefulness, ease of use, and satisfaction. Because the spreadsheet does not contain chronologically ordered learner–AI turns, validated cognitive-load measures, or experimental group labels, inferential comparisons and sequential-pattern statistics are not reported. The findings therefore provide a descriptive empirical profile that can support, but does not by itself establish, claims about predictive processing or cognitive mechanisms in GenAI-mediated EFL learning. The study highlights the value of integrating psycholinguistic theory with human-AI interaction analysis and identifies the additional turn-level data required for direct process-sensitive testing.

KEYWORDS : generative artificial intelligence; predictive processing; cognitive load; L2 processing; EFL learning; human-AI interaction; psycholinguistics