Revisiting communicative competence in the age of AI: Implications for large-scale testing

Abstract

Abstract Changes in the characterization of communicative competence, especially in the context of large-scale testing, are typically driven by an evolving understanding of real-world communication and advancements in test construct theories. Recent advances in AI technology have fundamentally altered the way language users communicate and interact, prompting a reassessment of how communicative competence is defined and how language tests are constructed. In response to these significant changes, an AI-mediated interactionalist approach is proposed to expand communicative competence. This approach advocates for extending the traditional concept of communicative competence to encompass AI digital literacy skills and broadened cognitive and linguistic capabilities. These skills enable effective AI tool usage, as well as the interpretation and application of AI-generated outputs and feedback, to improve communication. Embedding these competencies into language assessments ensures alignment with contemporary communication dynamics, enhancing the relevance of language assessments, and preparing learners for navigating AI-augmented communication environments. While high-stakes testing faces considerable challenges in adopting this expanded construct, low-stakes formative assessments, where scores do not influence critical decisions about individuals and where opportunities exist to rectify errors in score-based actions, if any, provide a fertile ground for exploring the integration of AI tools into assessments. In these contexts, educators can explore giving learners access to various AI tools, such as editing and generative tools, to enhance assessment practices. These explorations can start to address some of the conceptual challenges involved in applying this expanded construct definition in high-stakes environments and contribute to resolving practical issues.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 18 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.