One of the joys of researching AI coaching is the surprises I encounter along the way. In our most recent study that was accepted for publication recently in the journal “Interacting with Computers” (to be published soon) I collaborated with Dr Glen Wallis and Prof Martin Kidd to investigate the role of text versus voice as communication modalities in coaching chatbots.
We created two equivalent non-directive coaching chatbots that help people prioritise and set a goal for the day. The TextBot used text only input and output while the VoiceBot accepted user’s voice input and responded verbally. We asked participants to rate the TextBot (189 users) and VoiceBot (204 users) in terms of usability, performance expectancy and risk perception. We also looked at how introverts and extroverts rated these two bots.
If I had to ask you which coaching chatbot introverts would prefer (TextBot or VoiceBot) I bet that most would choose the TextBot. We stated this as our hypothesis and it is also the response I usually get when I ask my audiences. Turns out not to be the case! Introverts preferred the VoiceBot. In the article we offer possible explanations for this such as that introverts may feel more psychologically safe communicating with a chatbot and therefore chose to talk rather than text, and that introverts typically avoid frustrating situations and found the VoiceBot easier to use than the TextBot. Either way, a bit of a surprise.
The point I am trying to make here is that AI chatbot coaches are really a new frontier and there are many nuances we need to understand. We therefore need to proceed carefully and be guided by rigorous research to understand how to create and use these entities.