This paper presents a human-robot mutual learning system for language acquisition grounded in homeostatic needs. The system uses differential outcomes training (DOT) where the robot provides specific feedback based on its internal needs. The study found that DOT can enhance human learning efficiency, which then enables more efficient robot language acquisition. The robot’s language is similar to a human infant’s during the ‘babbling’ phase, associating vocabulary with internal needs like hunger, thirst, and curiosity. The study suggests this approach could facilitate cognitive interventions, such as for people with dementia, by engaging humans more actively in training tasks.
Publication date: 20 Oct 2023
Project Page: https://arxiv.org/abs/2310.13377
Paper: https://arxiv.org/pdf/2310.13377
