Mixed-Initiative Interaction in Interactive Applications: A Programming Model Approach

Title: Mixed-Initiative Interaction in Interactive Applications: A Programming Model Approach

Abstract: This research explores the use of partial evaluation as a programming model for realizing mixed-initiative functionality in interactive applications. Mixed-initiative interaction is a type of communication where both parties can take turns to change and control the flow of the interaction. The study focuses on the aspect of mixed-initiative known as unsolicited reporting, where users can provide information out of turn. The researchers demonstrate how this can be modeled by "jumping ahead" to nested dialogs, using partial evaluation. Their approach allows for the view of dialog management systems in terms of their native support for staging and simplifying interactions. The study characterizes three different voice-based interaction technologies using this viewpoint. Notably, they show that the built-in form interpretation algorithm (FIA) in the Voice XML dialog management architecture is actually a (well-disguised) combination of an interpreter and a partial evaluator.

Main Research Question: Can partial evaluation be used as a programming model for realizing mixed-initiative functionality in interactive applications?

Methodology: The researchers use a mixed method approach, combining literature review, conceptual modeling, and empirical validation. They analyze existing voice-based interaction technologies and develop a conceptual model to demonstrate how partial evaluation can be used to model mixed-initiative interaction.

Results: The study shows that partial evaluation can be a useful programming model for realizing mixed-initiative functionality in interactive applications. It allows for the modeling of out-of-turn interactions by users, enabling them to provide information when they choose, not just when prompted by the system. The researchers also demonstrate that the built-in FIA in the Voice XML dialog management architecture is a combination of an interpreter and a partial evaluator.

Implications: The research has implications for the design and development of interactive applications, particularly those using voice-based technologies. It suggests that partial evaluation can be a powerful tool for creating more natural and personalized interactions. The study also contributes to the ongoing debate about the definition and classification of mixed-initiative interaction by providing a programming model perspective.

Link to Article: https://arxiv.org/abs/0110022v1 Authors: arXiv ID: 0110022v1