Personas from Static Profiles to Conversational User Interfaces Ilkka Kaate, Joni Salminen, Soon-Gyo Jung, Trang Thi Thu Xuan, Jinan Y. Azem, João M. Santos, Bernard J. Jansen ACM Designing Interactive Systems (DIS2025), Funchal, Madeira, Portugal. Corresponding author: [email protected]
for end users to interact with personas in natural language. • In a randomized controlled experiment, we compare two persona interaction modalities, document and dialogue personas, both generated using AI approaches from a real survey dataset (Pew Research, N = 10k). • Participants (N = 54) were researchers and engineers (44.4% female, avg. age 33 years) and the task was about locating specific information about the persona. • We measured participant perceptions of personas (Persona Perception Scale), their perceptions of the persona system’s usability (System Usability Scale), and task success rate.
use • Less enjoyable experience • More complete • More transparent • Were found less helpful for the task • Higher task success rate • Easier to use • More enjoyable experience • Less complete • Less transparent • Were found more helpful for the task • Lower task success rate
different types; for example: • Document personas caused users to struggle with information overload (“high information personas”) • Chat personas imposed a ‘fourth wall’ effect in that people experienced the interaction as if they were talking to a representative of the persona rather than the persona itself (prompting matters!) • Task success rate is not necessarily connected with better usability in the context of using personas (false confidence?).