Adaptive vs Plain-text AI Chatbot Outputs : Effects on Students’ Perceived Clarity, Cognitive Load, and Format Preference

Sammanfattning

Most current AI chatbots default to Plain-text outputs with only slight formatting, such as headers and bullet points. Recent industry updates have begun to introduce more adaptive elements such as widgets and cards. The Cognitive Load Theory states that the structure of presented information directly affects the mental effort required to process it. There is no direct research on how Adaptive AI chatbot output formatting affects cognitive load compared to Plain-text. Therefore, the purpose of this study is to compare adaptive and Plain-text AI chatbot output formats in terms of how university students perceive their clarity and cognitive load, and to examine how their preferences for each format vary across three task types. A within-subject user study was conducted, with a total of 20 university student participants. The Adaptive vs Plain-text formats were tested on three different task types - Grammar checking, decision making, and exam preparation. Mixed methods were used to collect quantitative ratings for perceived clarity and cognitive load, and qualitative feedback. Custom-built prototypes were used for both conditions to control content and response time. Results showed Adaptive UI rated higher in perceived clarity and lower in cognitive load across all three task types, and participants preferred the Adaptive UI for all task types. These findings suggest that Cognitive Load Theory and multimedia learning principles extend to AI chatbot interfaces, where Plain-text outputs impose avoidable extraneous load, while adaptive formats reduce it.