Social Sciences in China (Chinese Edition)
No. 3, 2026
Reflection on the Premises of Comparing Human-Machine Linguistic Competencies from an Equivalence Perspective
(Abstract)
Liang Dandan
Large language models have intensified debate over the comparative assessment of human-machine linguistic competencies, making it necessary to scrutinize the theoretical premises underlying such comparisons. Existing comparative practices often tacitly adopt the typical adult native speaker as a homogeneous benchmark, thereby obscuring the internal diversity of human linguistic competencies. By analyzing the structural differences between descriptive and evaluative comparisons at the levels of tasks, indicators, and extrapolation, it becomes apparent that category mismatches in current research constitute the primary cause of legitimacy disputes. In response, a regulatory framework centered on equivalence can be constructed that, along three key dimensions—referential equivalence, contextual equivalence, and risk equivalence—sets regulatory constraints on cross-subject capability assessment. Such a framework aims to enhance the scientific character of model evaluation while providing defensible premises for the reconstruction of linguistic theories and the ethical governance of artificial intelligence in the digital-intelligent age.
