Conference Agenda

Proposal Information (461)

Recent developments in generative AI writing tools are radically changing how second language learners prepare, write and revise their texts. More specifically, in EFL, they are able to quickly refine grammar, structure and clear expression of meaning by producing instant personalised feedback and automatic suggestions (Dong, 2023). But this help introduces a new assessment challenge: in other words, if AI helps shape most of the final text, does that text still embody the learner's own language proficiency? This conflict is especially salient in high stakes international arenas like IB English B HL where the aim of assessment is to reflect independent communicative competence and real voice. Studies with EFL students suggest that, though AI-assisted assessment can counter linguistic uncertainty and enhance overall quality, over-reliance may result in over-harmonisation of texts and a loss of culturally situated individual expression (Werdiningsih, Marzuki & Rusdin, 2024). Even students have stressed the importance of mixing AI guidance with their own judgment to keep the originality and authenticity (Werdiningsih et al., 2024).

From a language assessment standpoint, generative AI undermines fundamental notions of validity, fairness and reliability. In a recent systematic review, it has been reported the risks of bias, inequality of access, and unequal impact along large-scale, local and classroom levels of assessment if AI is introduced without scrutiny (Chuang & Yan, 2025). Simultaneously academic integrity frameworks currently do not effectively classify and regulate the multitude of AI-based writing tools currently accessible in education, leading to a potential grey area as opposed to evident instances of violation (Roe, Renandya & Jacobs, 2023).

Instead of treating AI as a threat to humans alone, a range of researchers advocate for a participatory, human-AI model, in which teachers are at the heart as interpreters, guides and ethical gatekeepers of evaluation (Amin, 2023). This raises the question of what constitutes “authentic” writing within digitally mediated settings. This study therefore asks: In what way does the deployment of AI software tools challenge the authenticity of language writing assessment within IB English B HL, and how assessment could change in order to maintain relevance and legitimacy on international stage? The study is grounded on sociocultural theory of mediated learning, which sees AI as a medium that enhances but also redistributes cognitive efforts. For this reason, authenticity is defined as a state of learner choice in how and when AI is offered input rather than one that reflects a lack of tools. For that, in IB classes — a place in which students have different linguistic resources — preserving voice becomes even more crucial (Werdiningsih et al., 2024).

Through examining AI mediation in novel ways beyond the framework of AI vs. no-AI writing, the project endeavours to provide an applicable evidence for assessment policy and practice in generative AI era (Chuang & Yan, 2025).

This study addresses the following main research question: to what extent do assessment tools improved by artificial intelligence demonstrate validity and reliability in assessing the intercultural communicative competence of foreign language teachers (ICC) prior to employment, compared with human expert evaluators, and how can human-artificial intelligence collaboration optimize the assessment process in multilingual contexts?

Additional research questions are the following:

1. What is the degree of mutual agreement between the assessments of ICC obtained using artificial intelligence and the expert assessments of teachers before starting training using criteria-based headings?

2. Which model of human-artificial intelligence interaction ensures maximum assessment reliability while maintaining the professional freedom of action of the teacher in a multilingual educational context?
Theoretical Framework

This research integrates four complementary theoretical frameworks to conceptualize and validate AI-enhanced assessment of ICC in pre-service teacher education.
The model of intercultural communicative competence developed by Byram (1997) provides a pedagogical framework for the definition and practical application of ICC. The five aspects of the model: savoir (knowledge of oneself and others), savoir etre (openness and curiosity), savoir comprendre (skills of interpretation and establishing relationships), savoir apprendre/faire (skills of discovery and interaction) and savoir s'engager (critical cultural awareness) form the basis of the evaluation criteria. The Byram system is particularly suitable for teacher education because it clearly defines competencies that can be taught and evaluated, making it practical using criteria-based headings. In this study, the Byram model is adapted to develop performance descriptors that can be evaluated using both human experience and artificial intelligence-assisted pattern recognition.
The procedural model of intercultural competence developed by Deardorff (2006) complements Byram's concept, focusing on internal and external results, providing a development perspective when acquiring ICC. The Deardorff pyramid model distinguishes between necessary attitudes (respect, openness, curiosity), acquired knowledge and skills (cultural awareness, deep understanding of culture, listening and observing skills), internal results (adaptability, flexibility, ethnorelevant perspective, empathy) and external results (effective and adequate communication and behavior). This process orientation defines the structure of the assessment, recognizing that the development of the ICC is gradual, which requires the use of assessment tools that take into account gradual growth, rather than binary assessments of competence.
The concept of Human-Artificial Intelligence Cooperation (HAIC) (2024) provides for the integration of artificial intelligence into the assessment process while maintaining human control. This framework conceptualizes collaboration in three modes: artificial intelligence-oriented (automation under human control), human-oriented (AI to support decision-making), and symbiotic (collaborative judgment formation). The three main factors for evaluating the system - goals (matching AI results to pedagogical tasks), the quality of interaction (transparency, explainability, trust) and the distribution of tasks (optimal allocation of assessment responsibilities) determine the design of the human-AI assessment system. For the ICC assessment, the symbiotic mode is considered optimal because it uses artificial intelligence capabilities to process large amounts of multimodal data (text, audio, video), while preserving human experience in interpreting cultural nuances, contextual relevance, and critical cultural awareness, aspects that defy algorithmic reduction.
The Technology Acceptance Model (TAM) (Davis, 1989) makes it possible to assess how much teachers accept AI assessment tools before starting training. TAM argues that technology adoption is determined primarily by perceived usefulness (the degree to which users believe technology improves productivity) and perceived ease of use (the degree to which users believe that the technology is effortless). TAM's expanded concepts, including self-efficacy, digital literacy, and subjective norms, contextualize acceptance within teacher education. This structure is important because the successful implementation of an assessment using artificial intelligence depends not only on technical validity, but also on the approval of the teacher, which determines the actual use and pedagogical impact.

The digital transformation of education, accelerated by the integration of generative AI (GenAI), has fundamentally reshaped language assessment practices. Secondary EFL teachers now face the dual challenge of navigating digitally mediated evaluation while addressing discipline-specific pedagogical, ethical, and technological demands (Cui et al., 2025; UNESCO, 2024). Existing competence frameworks often adopt a fragmented view, focusing either on generic digital literacy or discrete assessment skills, thereby overlooking the integrated and contextual nature of teacher competence in AI-enhanced environments (Estaji et al., 2024). Grounded in the “competence as a continuum” perspective (Blömeke et al., 2015), this study conceptualizes teacher competence as a multidimensional construct spanning dispositions, situation-specific skills, and observable performance. In the GenAI era, this continuum must also encompass emerging competencies—such as prompt design, critical evaluation of AI-generated outputs, and human–AI collaborative feedback—which are not yet systematically represented in language assessment literacy models (Kremmel & Harding, 2020; TALiGAI framework). This study therefore addresses the following research question: How can a continuum-based digital assessment competence framework for secondary EFL teachers be developed and validated to incorporate both established dimensions and AI-responsive capabilities?

This study aims to develop and validate an integrated digital assessment competence framework for secondary EFL teachers that aligns with a continuum-based understanding of teacher competence and responds to the evolving demands of GenAI. The framework synthesizes established dimensions of language assessment literacy—such as those identified in empirical models like the Language Assessment Literacy Survey (Kremmel & Harding, 2020)—with forward-looking, AI-integrated competencies highlighted in recent scoping reviews (e.g., knowledge of AI ethics, critical appraisal of AI outputs, adaptive pedagogical practices). It intends to provide a structured, discipline-sensitive tool to guide teacher professional development, curriculum design, and reflective assessment practice in digitally enhanced language learning environments.