Autonomous Characters as User Interface Agents


Sponsor John Stasko
stasko@cc.gatech.edu
253 CoC
Area Human-Computer Interaction

Problem
The purpose of this project is to familiarize yourself with research in software agents, particularly research on the use of personified characters as agents in user interfaces. Often, these characters are simulated human beings that look for input from the user, make suggestions, and take useful actions. Building such as interface character is quite challenging. Some of the difficulties involved include doing the I/O interactions well, making the character take helpful and useful actions, making the character not take extraneous actions, and doing the character's presentation well. In fact, some researchers question whether these kinds of interfaces really have any intrinsic value and whether we should pursue this type of interface paradigm. Your first task here is to do some background research and reading to familiarize yourself with prior work in the area.

Background
First, if you have not seen the Apple Knowledge Navigator video, view it. It is perhaps the prototypical vision of these types of interfaces. Next, read the articles listed below to learn more about the area:

The video and the papers can be acquired through Prof. Stasko. (Many more related papers are available too if you are interested.)

Next, examine the autonomous characters available from 3D Greetings, Microsoft, and Haptek. Download the freely available software and try each out.

You may also want to look at the toolkit from the Center for Spoken Language Understanding (CSLU) at the Oregon Grad Inst (cslu.cse.ogi.edu).

Finally, suppose you have software that is able to generate a personified talking head and allows you to control it remotely. You could be the "brains" behind the interface, do speech recognition and natural language understanding, then make appropriate replies to queries from a user. Thus, you'd be able to simulate an intelligent agent in the interface. Assuming this exists, design an experiment to evaluate this style of agent interface using the software. State the purpose of the experiment, what you'd hope to uncover, and the methodology you'd use to run it.

Deliverables
You should turn in a 5-10 page report describing what you learned about autonomous characters in user interfaces. One item to include in your report should be a critique of this approach (Do you think it will be an important UI paradigm or not?). The report should also include the description of your proposed experiment.

Evaluation
You will be evaluated on the quality of your report and the insights within it. We will judge the thoroughness and depth of your examination of the area and critique of these approaches.