Inference of Human Intentions in Context Aware Systems

Inference of Human Intentions in Context Aware Systems

Katsunori Oyama (Iowa State University, USA), Carl K. Chang (Iowa State University, USA) and Simanta Mitra (Iowa State University, USA)
DOI: 10.4018/978-1-61692-857-5.ch019
OnDemand PDF Download:
No Current Special Offers


Most of context models have limited capability in involving human intention for system evolvability and self-adaptability. Human intention in context aware systems can evolve at any time, however, context aware systems based on these context models can provide only standard services that are often insufficient for specific user needs. Consequently, evolving human intentions result in changes in system requirements. Moreover, an intention must be analyzed from 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 context aware system design. The authors investigated the possibility for inferring human intentions from contexts and situations, and deploying appropriate services that users require during system run-time. This chapter first focus on describing an inference ontology to represent stepwise inference tasks to detect an intention change and then discuss how context aware systems can accommodate requirements for the intention change.
Chapter Preview

1 Background

1.1 Contexts and Situations

The definitions and interrelationships of contexts, situations and human intentions have been intensively discussed in pervasive computing (Bihler et al., 2005) and service computing (Rolland et al., 07). These notions were originally studied in different areas and then later were put together with the goal of improving human-centric systems including robots, software agents in smart home, and mobile systems.

One definition of context is “any information that can be used to characterize the situation of an entity” (Dey, 2001) for providing relevant information or services to the user. Typically, context awareness research centers on designing context models or context ontology (Strang & Linnhoff-Popien, 2004; Krummenacher & Strang, 2007), and developing various context reasoning and monitoring methods. Many of the existing approaches synthesize information about locations, surrounding objects, physiological readings of people etc., for use in real world applications. For example, location-based services (Bellavista et al., 2008) provide personalized functions based on the location gathered from context information. As one of the drivers for a pervasive computing paradigm, context awareness creates a new means of human-computer interaction because as it is deployed all over the environment, it changes the way users interact with the environment or the software system. CA systems rely heavily on the information of various sensor data to adapt to changing situations of the user who is using a CA service. Normally the CA system contains three major steps: collecting environmental data through sensor network or control frameworks, understanding contexts from the environmental data based on a context model for the system, and triggering actions to provide services. These three steps constitute a spiral of executions of a CA system; any change in the existing environment could cause changes in the context model and therefore trigger new actions, which in turn again affect the environment and so on.

The definition of situation has evolved over a long period. In the 1980's, situation was originally studied to investigate a mathematically based theory of natural language semantics (Barwise & Perry, 1980). In a later work (Barwise & Perry, 1983), situation was defined in terms of information conveyed in a discourse that can be clearly recognized in common sense and human language. Based on this definition, events and episodes are situations in time. Today, researchers have different perspectives on how to define situation, often depending on research objectives. For instance, some define situation as “histories” (i.e., finite sequences of primitive actions (Levesque et al., 1998)). Situation can also be reasoned by aggregating a set of predicates that explicitly represent information about sensory data (Mastrogiovanni et al., 2008). Sequencing such predicates can form the history of a finite number of instances in a first-in first-out order. For our purposes, generally a situation is a set of contexts in the application over a period of time that affects future system behavior (Yau et al., 2008).

Key Terms in this Chapter

Evolution of Services: The development process that creates new generation of a context aware service. There is an emerging challenge in requirements engineering to analyze an individual user’s behavior based on context information.

Human-Dimensional Inference Ontology: A task ontology designed to infer human intentions based on DIKW hierarchy. This ontology models execution of step-by-step inference tasks in a CA system that proceeds in a hierarchical fashion from low level processes where device information is captured to higher level human-machine interaction processes

Intention Change: An uncertainty factor that drives system requirements to be obsolete or contrary to user needs. Human intention change, once monitored, facilitates capture of new requirements; these new requirements trigger the next cycle of system evolution for upgrading the system to a new version.

Human Intention: Temporal sequence of user actions in order that the system user can achieve a goal. It belongs to the decision-making process which is within the scope of human knowledge and wisdom.

Inference Ontology: Task ontology in which a set of inference tasks are modeled. In this chapter, the inference tasks are referred to as the tasks for inferring human intentions from context information.

BDI Model: An agent planning model/architecture that mimics Belief-Desire-Intention of human mental state, originally described by Bratman in 1987.

DIKW: Abbreviated name of Data-Information-Knowledge-Wisdom that classifies a human thinking for knowledge management.

Complete Chapter List

Search this Book: