Since its first appearance in the field of language testing, cognitive diagnostic assessment (CDA) has attracted attention for its ability to extract the intricacies of students' cognitive abilities. However limited research has discussed the issues in the implementation of CDA. Therefore, this article offers an overview of CDA's implementation in language proficiency evaluation. The article also engages in a comprehensive discussion on the conundrum and considerations within CDA, particularly the ongoing debate between distinct classifications of cognitive diagnostic models. It elaborates on the distinctions between the models and their implications for assessment depth and diagnostic insights. Additionally, this article delves into the clash between retrofitting existing items and developing new diagnostic items, highlighting the strategic considerations in each approach. Apart from that, the contentious issue of validating Q-matrices, crucial in CDA, is thoroughly examined, presenting the battle between expert-based and empirical validation methods. The persistent challenges in CDA have profound implications for both theoretical frameworks and practical applications. The theoretical debate not only influences our understanding of cognitive processes but also shapes the conceptualization of diagnostic information extraction. In practical terms, decisions regarding item development, retrofitting strategies, and Q-matrix validation methods directly impact the effectiveness of CDA in providing targeted interventions and personalized learning strategies in real-world educational contexts. Future research directions are also presented, emphasizing the need for more development of entirely new diagnostic items, hybrid CDMs, and adaptive cognitive diagnostic assessments. Practical recommendations are provided for practitioners, encouraging a strategic approach based on specific assessment goals.
