John Zimmerman, associate professor, recently received the final of four patents, co-invented with Anthony Tomasic. Zimmerman, who also holds an appointment in the Human-Computer Interaction Institute, invents new ways for people to work with intelligent computing systems. The series of patents cover work on intelligent agents that allow office workers to automate their work. This invention is based on an observation that people often retrieve information from the Internet in response to an email request. The patent describes a machine learning agent that learns to recognize incoming messages and to automatically retrieve supplemental information a user may need to complete the request.
The goal is to turn computer agents into power tools that office workers can use to automate mundane and repetitive tasks. A user can "train" the software, by showing it how a task is done. If the task is to retrieve information on a person mentioned in an email, the user can create a table next to the email, demonstrating what information they want and where it comes from. The agent studies the message and observes the data being copied and pasted. After only three or four examples, the agent will begin to automatically append the desired content into the incoming message. Users get to choose the tasks they want to automate. The agent takes over the repetitive work. As Zimmerman says, the aim is for the system to get the user as far along in the task as possible before human judgment is needed.
Others are working on large scale, such as tasks that everyone does often, such as detecting a meeting request in a message. What Zimmerman and Tomasic have developed is something that focuses on more personal or idiosyncratic tasks; ones to small to hire a program to automate. The system is customizable so that any office worker can use it for the tasks specific to their individual job.
Zimmerman began working on automation tools for office workers while on a DARPA project to develop intelligent cognitive assistants. Today his research focuses on how people can more effectively apply the power of computing to their personal situations and how interactions with computers can be more situated into people’s environments. His current work includes a crowdsourcing system that allows transit riders to create their own real-time arrival information system and work on how people can better access and engage their ever increasing collections of virtual possessions.