The PAL project took a cyclic –human-centered, iterative and incremental– development process, called Socio-Cognitive Engineering (SCE). The end-users have been involved in all design and evaluation phases to address their needs and integrate their knowledge in the PAL system. Part of expert’s tacit knowledge was made explicit in a formal (logically correct) model, i.e. an ontology, which is interpretable by the relevant human stakeholders and the PAL system. It resulted in an extendable set of self-management objectives (focusing on learning) and related task content, with a coherent and concise structure. Further, social and cognitive theories have been integrated into the PAL ontology, as a transparent and verifiable foundation of PAL’s supportive behaviors (e.g. affect, memory, agreement and explanation). PAL acts as a partner for child’s disease management by the setting of joint objectives, entering into agreements, sharing of experiences, personalized action selection, provision of feedback and explanations, and showing appropriate learning styles. For each of these partner functions, the research provided scientific and technological results that have been or will be published. For example, specific experience-sharing, explanation and learning methods were developed for which, subsequently, specific user responses and preferences were identified in evaluations. As a second example, the Predictive User Model (for knowledge-level tracing and gaze estimation) and the aspect-based Sentiment Mining Module proved to out-perform state of the art technology. The final summative evaluation of the 3rd design-test cycle compared child’s self-management with the PAL-system versus “care as usual”, for a period of twice 3 months (with children aged 7-14y, in the Netherlands and Italy). PAL proved to partially support the three human basic needs that affect the development and habituation of human behaviors in a social environment, such as disease self-management (Self-Determination Theory). First, PAL enhanced the acquisition of competencies (knowledge and skills) for diabetes self-management. Second, children liked the PAL-robot and were motivated to continue the robot-mediated tasks (“relatedness”). Third, PAL enhanced child’s subjective self-care and diabetes related quality of life assessment, which may indicate progress in autonomy development. The re-usable PAL design rationale, ontological models and Co-design for Child-Computer Companionship (C4) suite are maintained and accessible in the Socio-Cognitive Engineering Tool.
