Persona Design for Educational Chatbots

Abstract

Designing chatbots for use in an educational setting is challenging, as users' behaviours and expectations differ greatly. Little research has been conducted on designing and building context and persona-based chatbots for university students. New approaches to the design of educational chatbots are required. This study applied a design science research (DSR) methodology and was executed in three iterations, starting with persona elicitation using machine learning techniques. A data-driven persona development (Persona3D) method and modelling approach was undertaken that drew on the earlier machine learning and structured equation modelling concerning technology acceptance. Finally, instantiations of the Persona3D approach were realised through designing (with journey mapping and Persona3D models) and building a range of chatbot prototypes. The results from the machine learning analysis show that there are eight distinct student groups. The results of the second iteration show that performance expectancy, effort expectancy and habits are the three main predictors of behavioural intention towards the use of chatbots. The third iteration evaluated the approach by designing and building eight chatbots, including knowledge acquisition, knowledge assessment, assignment guider and lab support chatbots. This study makes several contributions, primarily the Persona3D method and model. Moreover, the design, build, and evaluation cycle extended the Extended Unified Theory of Acceptance and Use of Technology model (UTAUT2 model) into the domain of university students' acceptance and use of chatbots, as well as identified the chatbot features that support these constructs. Furthermore, this study is the first to examine the effects of persona moderators on students' acceptance and use of chatbot technology. Future work could evaluate the effectiveness of the resulting chatbots on students, utilising an adapted version of the System Usability Scale (SUS) or other approaches.