While teachers often monitor and adjust their learning design based on students' emotional states in physical classrooms, synchronous online environments often limit their ability to perceive the emotional climate of the class. Drawing from the concept of social translucence, it is suggested that making students' emotional states "visible" in online settings can help foster empathetic interactions. Recent advancements in emotion recognition technology are enabling the creation of learning analytics (LA) systems that can estimate students' cognitive-affective states in real time. Yet, the adoption of these systems raises ethical, cultural, and practical concerns when implemented in learning environments, including potential challenges related to accuracy, privacy, and data integrity. To address these concerns, we conducted an in-depth qualitative study exploring the perspectives of 12 undergraduate students on modelling and visualizing their cognitive-affective states in the context of a Mexican higher-education institution. The study provides insights within a particular cultural context, which can guide the design of more human-centred emotion recognition--based LA tools for online educational contexts or contribute to informed decisions about the necessity of modelling and visualizing cognitive states.