Supporting and assessing human-AI collaborative digital multimodal composing: A process-oriented framework

Human-AI collaborative digital multimodal composing (HAI-CDMC) has emerged as a new form of DMC in which AI can negotiate meaning and co-construct DMC with learners. This poses challenges for assessment and highlights the importance of human-AI collaboration process, while existing DMC assessment frameworks remain product-oriented and grounded in non-AI contexts. Addressing this gap, this study aims to develop a process-oriented assessment framework to conceptualise and assess the quality of human-AI collaboration in HAI-CDMC. A preliminary assessment framework was developed, informed by research on assessing collaborative DMC and GenAI-assisted DMC. Adopting a collaborative action research approach, we collected self – and peer-assessment sheets and reflection journals from 77 students and conducted interviews with six of them. The analysis identified recurring HAI practices that participants themselves perceived as more conducive to collaboration with AI. Informed by learners’ assessment practices and perceptions, our preliminary framework was refined into two dimensions: (1) Human Control over the Interaction, and (2) Human Engagement with AI Outputs, each operationalised through two descriptors: (1) Interactional Grounding and Interactional Regulation, and (2) Evaluative Judgement and Transformative Integration. Learners perceived the framework as supporting human-AI collaboration both retrospectively and prospectively. The study contributes an exploratory, process-oriented framework that makes human-AI collaboration processes more visible and assessable in DMC contexts.