@inproceedings{toushik-wasi-2026-position,
    title = "Position: Biomedical {NLP} Demands Specialization, Not Generalization",
    author = "Toushik Wasi, Azmine",
    editor = {Danilova, Vera  and
      Kurfal{\i}, Murathan  and
      S{\"o}derfeldt, Ylva  and
      Reed, Julia  and
      Burchell, Andrew},
    booktitle = "Proceedings of the 1st Workshop on Linguistic Analysis for Health ({H}ea{L}ing 2026)",
    month = mar,
    year = "2026",
    address = "Rabat, Morocco",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-eacl/2026.healing-1.7/",
    pages = "77--93",
    ISBN = "979-8-89176-367-8",
    abstract = "Multimodal Artificial Intelligence (AI) promises to transform biomedicine by integrating imaging, genomics, and clinical data for superior decision-making. Yet, we contend that the current pursuit of large-scale generalist models is fundamentally misaligned with the high-risk nature of biomedical applications. This position paper argues that biomedical NLP demands specialization, not generalization, challenging the assumption that greater model scale and generality inherently ensure robustness in healthcare. We propose a theoretical framework built on three biomedical axioms: error cost asymmetry, multimodal data fragility, and interpretability{--}utility coupling, alongside a formal proof of criticality in biomedical NLP, showing that generalist models are intrinsically unsuited for medical tasks. As a secondary contribution, we advance a task-first design paradigm centered on modular, specialized, and ethically grounded AI architectures for biomedical use. Through analysis and illustrative cases, we contrast this approach with scale-centric strategies, exposing risks such as bias amplification, reduced interpretability, and exclusion of rare or underrepresented populations. We call for a realignment of research, funding, and regulation toward specialization as the sustainable path for meaningful and equitable biomedical AI, aiming to spark critical discourse on what constitutes genuine progress in machine learning for health."
}