What is the Best Action for Children to "Wake Up" and "Put to Sleep" a Conversational Agent?

Abstract

Popular conversational assistants like Amazon Alexa, Google Assistant, and Apple Siri are activated by pronouncing their wake-up word ("Alexa", "OK Google", and "Hey Siri", respectively), and they stop listening when they recognize a pause. We explore whether this paradigm is appropriate for children and whether there are alternative, more effective ways for children to "wake-up" and "put to sleep" a conversational system. We consider seven possibilities: wake word, free-form speech, physical button, digital button, mouse, gaze, and buzzer. We propose a methodological framework of analysis based on the Multi-Attribute Value Theory (MAVT). As part of the decision-making process, we ran a study involving 42 children aged 9 to 10 who experienced the seven approaches. Our results suggest that the physical button is the most appropriate solution for this target group, which opens up new directions in the design of interaction affordances of conversational agents for children.