Computational Models for Bilingual Aphasia: A Systematic Review of Language Deficit Research with a Focus on Code-Switching and Translation

This systematic review synthesises computational models of language deficits in bilingual aphasia, focusing on code-switching and translation. Per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, this study identified 42 publications (2010–2025). Of the three deficit types, lexical retrieval and naming deficit simulations are most mature, replicating naming errors and predicting cross-language generalisation. Code-switching and translation deficit modelling is extremely limited: only a single computational lesioning study has addressed both through mechanistic simulation. Machine learning methods show preliminary promise for predicting treatment outcomes but remain largely data-driven. Three key challenges emerge: (1) methodological fragmentation—heterogeneous model types, lesion implementations, and evaluation metrics, with no standardised quantitative validation; (2) an underdeveloped theoretical foundation, as most studies are confined to behavioural fitting rather than testing hypotheses on impaired language control via computational lesions; (3) limited clinical translation, with few predictive studies integrating longitudinal patient data, with approximately 70% of studies concentrated in North America/Western Europe and biased toward English–Spanish bilinguals, limiting theoretical generalisability. Future work should shift from behavioural fitting to explainable mechanistic simulation, develop predictive frameworks for personalised rehabilitation, and establish standardised protocols bridging computational modelling with clinical application.