Principles of Mixed-Initiative User Interfaces

Vocabulary-style flashcards covering the principles of mixed-initiative user interfaces, decision theory in UI, and the Lookout system architecture.

Mixed-Initiative User Interface

An interface design where intelligent services and users collaborate efficiently to achieve the user's goals, coupling automated services with direct manipulation.

Direct Manipulation

A methodology focusing on new metaphors and conventions that enhance a user's ability to directly interact with interface objects to access information and invoke services.

Interface Agents

Automated services centered on building machinery for sensing a user's activity and taking automated actions.

Value-Added Automation

Automated services that provide genuine value over solutions attainable solely through direct manipulation.

Inference Under Uncertainty

The use of machinery for inferring and exploiting the uncertainty about a user's intentions and focus of attention.

Lookout

A testbed system for mixed-initiative UI that overlays automated scheduling services on Microsoft Outlook to assist with reviewing calendars and composing appointments.

Support Vector Machine (SVM)

The text classification methodology used by Lookout to assign probabilities of user intention based on message content.

Expected Utility

A decision-theoretic measure used to direct actions based on the costs, benefits, and uncertainties of a user's goals.

Threshold Probability ($p^*$)

The specific inferred probability of a user having a goal at which the expected value of action and inaction are equal.

Social-agent Modality

A mode of interaction projecting an explicit social presence using animated characters, audio channels, and anthropomorphic behaviors.

Dwell Time

The amount of time a user prefers to review the content of a message before accepting automated calendaring and scheduling operations.

Sigmoid Function

The mathematical function used to approximate the relationship between message size and the preferred time for deferring offers of service.

Life-long Learning

The ability of automated services to continue to improve by observing and recording user goals and needs over time.

$$u(A_{tr}, G)$$

The utility representing the outcome where the system takes a desired action when the user indeed has the goal.

$$u(A_{tr}, \neg G)$$

The utility representing the outcome where the system takes an automated action when the user does not have the goal.

Working Memory of Interactions

The maintenance of recent interaction history that allows users to make efficient and natural references to objects in shared short-term experiences.

Graceful Degradation of Precision

A strategy where an agent reduces the precision of a service to match current uncertainty, such as showing a week view of a calendar instead of a specific time.