Activity qualifiers in an argumentation framework as instruments for agents when evaluating human activity

Abstract

Theoretical frameworks have been developed for enabling software agents to evaluate simple activities such as walking and sitting. However, such frameworks typically do not include methods for how practically dealing with uncertain sensor information. We developed an argument-based method for evaluating complex goal-based activities by adapting two qualifiers: Performance and Capacity defined in the health domain. The first one evaluates what a person does, and the second one how “well” or “bad” an activity is executed. Our aim is to deal with uncertainty and inconsistent information; generate consistent hypotheses about the activity execution; and resemble an expert therapist judgment, where an initial hypothesis assessment can be retracted under new evidence. We conducted a pilot test in order to evaluate our approach using a Physiotherapy assessment test as a goal-based activity. Results show that skeptic argumentation semantics are may be useful for discriminating individuals without physical issues by considering Performance and Capacity; conversely, credulous semantics may be suitable for obtaining information in the evaluation of activity, which an intelligent agent may use for providing personalized assistance in an ambient assisted living environment.