My work focuses on machines that learn new tasks and goals from ordinary people in everyday human environments. This research works from the assumption that machines meant to learn from people can better take advantage of the ways in which people naturally approach teaching. My research works to understand and computationally model specific mechanisms of human social learning in order to build machines that participate in social learning interactions. This work has interconnected goals from Artificial Intelligence and Human-Computer/Robot Interaction: improving the performance of a machine's learning behavior through attention to human interaction and improving the experience of the human teacher by designing interactive learning algorithms based on how people teach. This work spans the fields of Artificial Intelligence, Machine Learning, Robotics, Human-Computer/Robot Interaction, and Cognitive Science.
Honors and Awards
Dissertation Research (Sept 2003-June 2006)
Socially Guided Machine Learning
While the topic of human input to machine learning algorithms has been explored to some extent, prior works have not gone far enough to understand what people will try to communicate when teaching a machine and how algorithms and learning systems can be modified to better accommodate a human partner. Interface techniques have been based on intuition and assumptions rather than grounded in human behavior, and often techniques are not evaluated with everyday people.
Using a computer game, an experiment with human subjects provides several insights about how people approach the task of teaching a machine. In particular, people want to direct and guide an agent's learning process, they quickly use the behavior of the agent to infer a mental model of the learning process, and they utilize positive and negative feedback in asymmetric ways.
Using a robotic platform, Leonardo, and 200 people in follow-up studies of modified versions of the computer game, four Socially Guided Machine Learning research themes are developed. 1) The use of human guidance in a machine learning exploration can be successfully incorporated to improve learning performance. 2) Novel learning approaches demonstrate aspects of goal-oriented learning. 3) The transparency of the machine learner can have significant effects on the nature of the instruction received from the human teacher, which in turn positively impacts the learning process. 4) Using asymmetric interpretations of positive and negative feedback from a human partner can result in a more efficient and robust learning experience.
MIT Postdoctoral Associate (June. 2006 - present)
The goals of my one-year post-doc with Prof. Cynthia Breazeal have been to publish and present my thesis research at four conferences; to prepare two journal publications of my thesis research; to conduct follow-up research with both the Sophie and Leonardo platforms; and to gain classroom and teaching experience. Additionally, this post-doc involves presenting work to sponsors, assisting in grant proposal preparations, and consulting with lab sponsors including senior management of Fortune 500 companies.
MIT Research Assistant to Professor Cynthia Breazeal (Sept. 2003 - June 2006)
In the Robotic Life Group, we develop robotic systems that can naturally work and learn with human partners. Additionally, this RA has involved presenting work, assisting in grant proposal preparations, and consulting with lab sponsors.
Microsoft Corporation, Redmond, WA, USA, Internship at Microsoft Research (June 2003 - Sept. 2003)
Interned with the Communities Technologies Group (now Social Computing), led by Marc Smith. Expanded the group's Usenet technologies for mining social statistics to the realm of email. Developed data mining tools, incorporated statistics into the Outlook email client, and ran a usability study to evaluate the accuracy of the mined statistics.
MIT Research Assistant to Professor Ted Selker (Sept. 2000 - June 2003)
Assisted in the ongoing development of the Context Aware Computing Group laboratory, incorporating Artificial Intelligence and interactive computer systems.
IBM, Austin, TX, USA, Scientist/Engineer RS/6000 Division (June 1999 - July 2000)
Worked with a group of 12 engineers charged with simulating (in C++) the memory subsystems of a RISC symmetric multiprocessor, for the purpose of circuit design verification. I wrote a cache preloader to save millions of cycles needed to achieve specific cache states in processor simulation. I interned with this same group for several semesters during my undergraduate studies.
Please download the .pdf for reference contact information.