Differential representation of trilingual proficiency levels in multilingual psychocognitive networks: a network analysis and machine learning approach

With the acceleration of globalisation, multilingual proficiency has become a crucial skill in the social development of individuals. Complex Dynamic Systems Theory (CDST) posits that the outcomes of language learning at varying proficiency levels can be attributed to the interactions among various language cognitive abilities. However, previous research has been limited by methodological constraints, lacking modelling studies of language-specific abilities and domain-general cognitive abilities within the multilingual psychocognitive networks of trilingual learners at different proficiency levels. The complexity of multilingual networks remains underexplored. This study models the psychocognitive networks of Hong Kong trilingual learners using network analysis and machine learning. Results show that L3 learning anxiety plays a key role, with higher proficiency levels showing more complex connections. Key components include psychological states, L1/L2 proficiency, and cognitive abilities (e.g. auditory processing and working memory). These findings support CDST’s core tenet of dynamic interactions between abilities, offering new insights and directions for language learning research.