This paper proposes a learning analytics-based methodology for assessing human-AI collaborative writing, framed by the evidence-centered design (ECD) framework. The authors use elements of knowledge-telling, knowledge transformation, and cognitive presence to identify assessment claims, collect data from the CoAuthor writing tool, and apply epistemic network analysis (ENA) to make inferences about these claims. The study reveals significant differences in the writing processes of different groups of CoAuthor users, suggesting that the proposed method is a plausible approach to assessing human-AI collaborative writing. The research addresses the challenges of assessing writing crafted partly by a human and partly by AI, emphasizing the importance of understanding the cognitive behaviors involved in co-writing with generative AI. The findings support specific hypotheses regarding the impact of ownership over the final product, genre (creative vs. argumentative), and AI temperature settings on writing processes. The study also discusses limitations and future directions, including the need for more scalable and accessible assessment methods.This paper proposes a learning analytics-based methodology for assessing human-AI collaborative writing, framed by the evidence-centered design (ECD) framework. The authors use elements of knowledge-telling, knowledge transformation, and cognitive presence to identify assessment claims, collect data from the CoAuthor writing tool, and apply epistemic network analysis (ENA) to make inferences about these claims. The study reveals significant differences in the writing processes of different groups of CoAuthor users, suggesting that the proposed method is a plausible approach to assessing human-AI collaborative writing. The research addresses the challenges of assessing writing crafted partly by a human and partly by AI, emphasizing the importance of understanding the cognitive behaviors involved in co-writing with generative AI. The findings support specific hypotheses regarding the impact of ownership over the final product, genre (creative vs. argumentative), and AI temperature settings on writing processes. The study also discusses limitations and future directions, including the need for more scalable and accessible assessment methods.