Knowledge Elicitation of Human Activities Using a Graphical Modeling Language

Abstract: Recent, and ongoing, research has been addressing the general problem of representing complex human activities for the purpose to provide intelligent software agents with a way to identify, reason, and evaluate human activities. Reliable evaluation of complex human activities requires the intelligent agent to obtain a representation that comprises distinguishing features of an activity. This is a challenging task, since a person´s activity is driven by goals, motives, and norms that may be conflicting in a situation. In order to provide knowledge about human activities to an intelligent software agent, we require tools that can allow us modeling human activities enabling knowledge elicitation. This study has evaluated a software prototype of a graphical modeling language with the overall research question to find the minimum language elements required to elicit the knowledge of activities from a domain expert. Eight participants tested the prototype through think aloud usability sessions were their understanding of the structure of the language was tested. Qualitative data analysis was conducted using a Grounded theory approach, which validity has been discussed. The findings indicated that characteristics of the hierarchical structure of Activity Theory are a supportive theoretical framework for the graphical language that resembles the way occupational therapists reason when analyzing human activities. The study proposes a set of minimum elements for the graphical language. A focus a for future study is to target the intelligent software system to further expand and tune the language by the systems requirements.