Personalities in humanoid robots for cognitive training of older adults

In the last decade, the ageing of the population is occurring worldwide, and ageing increases the degeneration in the cognitive and physical domains of older adults. For this reason, technologies to support older adults in trying to slow down the progression of cognitive impairment are becoming more and more important. In particular, humanoid robots with social skills are increasingly common in the real world. Although life expectancy is increasing, the quality of life is not necessarily doing so. Thus, we may find ourselves and our loved ones dependent and in need of another person to perform the most basic tasks, which has a strong negative psychological impact. As a result, social robots may be the definitive tool to improve the quality of life by empowering people dependent on others and extending their independent living. In this context, humanoid robots can be an effective tool for the cognitive training of older adults, and to achieve this, their interaction with humans must be engaging. In this Thesis, we seek to understand if proposing robots with extraverted or introverted personalities could improve the user experience during a serious game scenario. Specifically, we design, implement, refine, and test a set of verbal and nonverbal parameters for such personality traits, which are general and potentially have different fields of application. The two personalities are implemented in an application that proposes typical cognitive training exercises using a Pepper robot. Additionally, we identify the requirements for designing and implementing a serious game to be a useful tool to be included in cognitive training. After evaluating the robot personalities with different tests and interviews with 52 users, including experts, healthy older adults and users with mild cognitive impairment, we address the problem of how to improve engagement and adaptation for older adults during repetitive cognitive training. The monotonous nature of repetitive cognitive training may cause older adults to lose interest and drop out. Social robots are used to reduce boredom and cognitive load when playing serious games as part of cognitive training, indeed. In this Thesis, a behaviour-adaptation technique is proposed to select the best actions, which consist of a combination of verbal and non-verbal interaction aspects, for the robot to maintain the attention level of older adult users during a serious game. The behaviour-adaptation technique proposed allows the robot to autonomously select the most appropriate actions to maintain the level of engagement of older adults during the full interaction session. After a session with 28 users, where both the adaptive and non-adaptive robot is used, a test to evaluate the adaptive behaviour of the robot is performed. The findings demonstrate users' ability to differentiate between the behaviours exhibited by the adaptive and non-adaptive robot. The users perceived the adaptive robot as displaying greater adaptability and engagement than the non-adaptive robot. This adaptability contributed to a more engaging and motivating user interaction with the robot. In summary, we provide a system and the guidelines to design a robotic behaviour of the future. The robot is able to autonomously adapt its personality to increase user engagement and experience, stimulating the users to continue the cognitive training.