Learning to Eat Healthily With a Robot: Creating and Evaluating a Dialogue Design for Social Robots to Support Children in Learning About Healthy Nutrition by Stimulating Reflection

Abstract

With the current shortage of teachers, virtual agents and social robots can provide a solution by supporting pupils in learning independently. They can do so in several manners, where stimulating reflection is considered particularly relevant. With only half of Dutch children meeting nutrition guidelines, health literacy is considered an important learning domain. However, to date it remains unclear how the interaction between agent/robot and child should be designed, especially in those domains. A systematic literature review showed that agents and robots are employed in various domains with various behaviors, but that only a small number of papers focus on reflection and/or health. To design a learning session with a robot that incorporated both, a human-centered design approach was adopted. This involved the creation of personas, scenarios, and reflection strategies, which resulted in a dialogue design for a learning session about healthy nutrition. The reflection strategies were evaluated during a between-subjects experiment (N = 28), consisting of a condition in which a social robot stimulated reflection on learning goals and learning progress, and a control condition, in which these reflection strategies were not employed. The outcome measures consisted of 1) recall of the learning goals, 2) learning outcomes, and 3) motivation to eat healthily. The former two measures were assessed on two occasions: directly after the session (short-term) and one week later (long-term). No significant results were found. However, the control condition showed a stronger decrease in scores over time. Future work is recommended to employ multiple sessions with the robot to stimulate reflection more effectively and to assess long-term effects.