Article Preview
TopContent
A Context Aware (CA) room may know that it is dark outside and that the resident has just entered the home and decide to turn on the lights, whereas if the resident enters the home with guests, it may decide to not only turn on the lights but to also turn down the heat a notch to adjust for more people. To be able to respond in this context aware manner, software systems in the CA room usually incorporate an adaptable and distributed device framework that consists of sensor networks and mobile computing techniques. Here the framework’s role is to acquire stimulus data using sensors, match the perceived sensory stimulus to a context, and trigger actions based on the recognized context.
It is well known that changes in circumstances under which individuals use a system influence them to change goals and hence intention to use the system. For example, life style and assistive environments may coevolve to maintain wellness of residents in a smart home (Helal et al., 2012). Necessary support of body wellness varies from person to person. Timely services would be required for a sudden situation even if such situation were not precisely defined during the system development; for instance, critical situations involve different behavior pattern from usual that leads to a health problem (Kim & Helal, 2013). Consequently, evolving human intentions result in changes in system requirements. As researchers and software system developers, we must answer the question, “How can the system find out about changing human intentions so that the system can respond with appropriate services?” In the past, intention has been discussed as the BDI model (Bratman, 1987) and implemented in software systems. Researchers today have found various types of sensor information helpful in designing computational methods to mine human intentions.
Unfortunately, most context models in pervasive computing have a limited capability for involving human intention in system evolvability; human intention may demonstrate a need for a service to adapt, however, CA systems based on these context models can provide only standard services that are often insufficient for specific user needs. As intention changes may occur at any time, pre-defined system requirements can become obsolete or contrary to user needs (Oyama et al., 2008a). Consequently, such a CA system cannot continue to deliver services as expected. Extraction of an intention from sensor contexts is hard challenge due to their tangled relations with different types of contexts. In the past, this complexity has prevented researchers from using computational methods for analyzing or specifying human intention in CA system design.
This study introduces understanding human intention as an important factor for the development of effective CA services. Inference process of human intention from context information has been investigated to discover causal factors of a potential failure and then deploy appropriate services on the fly. This paper first presents an inference ontology to represent inference tasks and then discuss how CA systems can identify an intention change in a smart home environment, and then review the steps of inference tasks which proceed in a hierarchical fashion from low level processes where sensor contexts are aggregated for higher level human-machine interaction processes.