Ph. inishe D.

I successfully defended my dissertation on May 13, 2008 and  now work as a research computer scientist with Starkey Labs.

 

---

 

I am interested in building Adaptive Human Computer Interfaces. My research harnesses the advances in the fields of Artificial Intelligence, Machine Learning, Robotics and Human Computer Interaction to build adaptive human-computer interfaces.

My adaptive interfaces framework (sycophant) relies on different learning paradigms like genetics based machine learning techniques (GBML), context-sensitive learning and data mining to build applications that are more responsive to a user's needs. I work under Dr.Sushil Louis' supervision in the Evolutionary Computing Systems Lab

In more detail, my research interests lie at the intersection of Human-Computer Interaction and Machine Learning . Currently, my research focuses on better personalization of adaptive user interfaces related to interactive desktop applications.

Modern software is highly interactive and ubiquitous. Despite the considerable maturity of human-computer interaction in the last few decades, user interfaces still tend to be fixed and behave in the same way irrespective of the individual users. user interfaces meagerly adapt to user's with different experience levels, expertise and work habits. An interface's lack of adaptiveness coupled with the variation in user characteristics clearly necessitates the need for user interfaces to be more adaptable towards individual users. Adaptive user interfaces/Intelligent User Interfaces are systems that learn a user preference-model from interaction data gathered from that user in different contexts.

Learning user preferences is a challenging task. Consider the case of an interactive media player. If you are listening to music on your desktop, you might pause the music while the phone rings in your office, while I generally prefer to decrease the volume. With the same media player, if you are talking with someone in your office, you might prefer to decrease the volume while I might prefer to pause the music. That is, UI preferences not only vary from user to user, they also vary for a single user depending on the context of use. In my doctoral work, I address the problem of learning user preferences for adaptive user interfaces based on contextual information gathered from a user's environment. Sycophant, my context-based user modeling framework, supports multiple desktop interfaces. Sycophant enables such interactive interfaces to adapt their behavior towards individual users and effectively learn their preferences for different interface actions. Results from a pilot and short-term studies have shown that for an application such as Google Calendar, Sycophant can successfully predict with an 88 percent accuracy the type of action to take for individual users.

---

---

---

Class Diagram of Sycophant's Application-Level API

---