David Maulsby

Instructible agents: software that just keeps getting better

Agent software is a topic of growing interest to users and developers in the computer industry. Already, agents and wizards help users automate tasks such as editing and searching for information. But just as we expect human assistants to learn as we work with them, we will also come to expect our computer agents to learn from us. This paper explores the idea of an instructible agent that can learn both from examples and from advice. To understand design issues and languages for human-agent communication, we first describe an experiment that simulates the behavior of such an agent. Then we describe some implemented and ongoing instructible agent projects in text and graphic editing, World Wide Web browsing, and virtual reality. Finally, we analyze the trade-offs involved in agent software and argue that instructible agents represent a sweet spot in the trade-off between convenience and control.

U-TEL: a tool for eliciting user task models from domain experts

Eliciting user-task models is a thorny problem in modelbased user interface design, and communicating domainspecific knowledge from an expert to a knowledge engineer is a continuing problem in knowledge acquisition. We devised a task elicitation method that capitalizes on a domain expert’s ability to describe a task in plain English, and on a knowledge engineer’s skills to formalize it. The method bridges the gap between the two by helping the expert refine the description and by giving the engineer clues to its structure. We implemented and evaluated an interactive tool called the User-Task Elicitation Tool (U-TEL) to elicit user-task models from domain experts based on our methodology. Via direct manipulation, U-TEL provides capabilities for word processing, keyword classification, and outline refinement, By using U-TEL, domain experts can refine a textual specification of a user task into a basic user-task model suitable for use in model-based interface development environments. Our evaluation shows that U-TEL can be used effectively by domain experts with or without a background in programming or interface modeling, and that the tool can be a key element in promoting user-centered interface design in model-based systems.

Management of interface design knowledge with MOBI-D

Effective guidelines for interface construction require developers to apply a user-centered approach in their designs. Yet, developers lack integrated tools that would allow them to work with high-level concepts, such as user tasks, and to relate them to lower level elements, such as widgets, in their interface designs. The Model-Based Interface Designer (MOBI-D) is a suite of tools for the management, visualization, editing, and interactive refinement of interface-design knowledge at multiple levels of abstraction. MOBI-D represents knowledge via declarative interface models that assign specific knowledge roles to each model component. Developers work in an integrated environment with abstract concepts such as user tasks, domain objects, presentation styles, dialogs, and user types while being able to relate those concepts to concrete interface elements such as push buttons. MOBI-D is the first development environment to integrate the disparate elements of interface design into structured conceptual units—interface models—and to define an interface design as an explicit declarative element of such units.

Inductive task modeling for user interface customization

This paper describes ActionStreams, a system for inducing task models from observations of user activity. The model can represent several task structures: hierarchy, variable sequencing, mandatory vs. optional actions, and interleaved sequences. The task models can be used for just-in-time automation and for guidance in user interface design.

MOBI-D: a model-based development environment for user-centered design

MOBI-D (Model-Based Interface Designer) is a software environment the design and development of user interfaces from declarative interface models. End-users informally describe tasks and data, from end-users, from which developers construct formal models of user tasks and domain objects. The system supports development of presentation and dialog specifications from such models, and allows visualization of interface designs as units encompassing all relations and dependencies among the elements of task, data and user-interface specifications, MOBI-D is the first development environment to define an interface model as a comprehensive conceptual object, to identify an interface design as a declarative component of an interface model, and to establish a development cycle based on such a model. The sharable nature of the interface modeling language of MOBI-D, along with the open architecture of its system opens the door for many research areas in HCI to explore the benefits and potential of using interface models.