CHI 97 Electronic Publications: Tutorials
Creating Conversational Interfaces for Interactive Software Agents
Tandy Trower
Advanced UI Design Group
Microsoft Corporation
One Microsoft Way
Redmond, WA 98052 USA
+1 206 882-8080
tandyt@microsoft.com
ABSTRACT
While much research and design has been presented on designing interactive agents and on speech interfaces, little has been said about combining these areas. This tutorial presents recommended guidelines for creating conversational interfaces with agents presented as interactive characters.
Keywords
Software agents, interactive characters, conversational interfaces, social user interface, speech interfaces
© 1997 Copyright on this material is held by the authors.
INTRODUCTION
This tutorial will define guidelines for creating conversational interfaces for interactive software agents. As defined for this session, a conversational interface is one that supports a natural dialogue between the user and a software application.
REVIEW OF AGENT AND SPEECH RESEARCH AND IMPLEMENTATION
Much excellent research is ongoing on agent design, including ongoing work by Dr. Pattie Maes [5] at MIT and Dr. Joseph Bates [1] at CMU. In addition, significant research has been done in the area of speech interfaces [3,6,9,11]. However, development has not been limited to research. Agents have begun to emerge in the commercial market, from the Internet to the desktop. Similarly, auditory (phone-based) agents that employ speech interfaces have begun to find their way into the market with products like Wildfire [10]. This opens an opportunity for a new form of interface where both interactive, visual agents and speech are combined.
COMPUTERS ARE SOCIAL ACTORS [7]
The subject of agents often raises the question of whether anthropomorphism is appropriate or not. However, such discussion may be moot, as it may be impossible to avoid to design software with out creating a social context. The simple phrasing of text on the screen can be sufficient to stimulate a social response [8]. The only question is how to use this channel to enhance human-computer interaction.
SPEECH RECOGNITION IS NOT ENOUGH
A common perception is that a conversational interface relies on the quality and accuracy of the recognition engine. However, even human speech recognition can break down, yet we are still able to communicate. Therefore, creating an effective conversational interface must also leverage others aspects of the human dialogue. Further speech recognition is not always practical and must be supplemented by other forms of dialogue interaction.
GUIDELINES FOR DESIGNING EFFECTIVE CONVERSATIONAL INTERFACES
Drawing from basic principles of user interface design, animation, and speech interfaces, the following are guidelines creating effective conversational interfaces with interactive characters.
* Use good aesthetics. While it doesn't take much to elicit a social response, people are accustomed to high quality output. This doesn't imply photographic or human images, but it does require attention to detail, such as the basic principles of animation [4].
* Provide good feedback. Any well-designed interface requires on providing good, appropriate, and timely feedback. Similarly, a conversational interface, enhanced with recognizable facial features, arouse interaction and expectations.
* Be non-exclusive. Delegation and speech interface are not always the best way to complete a task. Provide users with options to interact as they prefer.
* Be polite. Software often breaks simple rules of etiquette. In any social context, humans expect reciprocity of politeness. Gricean [2] principles provide a good guide of what to consider.
* Use praise. All of us respond to praise, even when it isn't warranted.
* Be personal. Adapt to the user's personality type and work style.
* Create a team player. We work more effectively with others when we feel we are matched as a team. Treat the user as a teammate.
* Provide good error handling. Communications errors are unavoidable, so an effective interface depends on how well you support error recovery.
* Clarify and limit choices to context. Speech technology rapidly breaks down when an engine is expected to handle any utterance, yet natural dialogue typically does not follow random paths. Where limitations must be set, they must be clear to the user.
* Use natural dialogue techniques. Speech is a natural interface, yet basic models of dialogue such as the use of anaphoria, ellipsis, rephrasing, and directive prompting are often overlooked.
CONCLUSION
Both agent and speech interface technology is rapidly emerging. Combining them together creates an exciting new form of interface design. Unlike tradition GUI design, there are no established guidelines for designing agent-speech interface. By understanding the model of human communication and the limitations of technology useful recommendations can be defined.
REFERENCES
1. Bates, J. The Nature of Character in Interactive Worlds and the Oz Project. Technical Report CMU-CS-92-200. School of Computer Science. Carnegie Mellon University. Pittsburgh, PA. October 1992.
2. Grice, H. Logic and Conversation. In Syntax and Semantics 3: Speech Acts. P. Cole and J. Morgan, eds. New York: Academic Press, 1975.
3. Herman, J. Newstalk: A Speech Interface to a Personalized Information Agent. MIT Master's Thesis, Program in Media Arts and Sciences. June 1995.
4. Lassiter, J. Principles of traditional animation applied to 3D computer animation. In Proceedings of SIGGRAPH '87 (1987), ACM Press, 35-44.
5. Maes, P., Agents that Reduce Work Overload and Information Overload. In Communications of the ACM, (July 1994), 31-40.
6. Marx, M. Toward Effective Conversational Messaging. MIT Master's Thesis, Program in Media Arts and Sciences. June 1995.
7. Nass, C., Steurer, J., and Tauber, E. Computers are social actors. In Proceedings of the CHI '94 Conference (Boston, MA, April 1994). ACM Press. 72-77.
8. Reeves, B. and Nass, C., The Media Equation: How People Treat Computers, Televisions, and New Media as Real People and Places. Cambridge University Press, 1996.
9. Schmandt. C. Voice Communication with Computers: Conversational Systems. New York: Van Nostrand Reinhold, 1994.
10. Thomas, F. and Johnson, O. The Illusion of Life. New York, Abbeville Press, 1981.
11. Wildfire Communications, Inc., Lexington, MA, 1995.
12. Yankelovich, N., Levow, G., and Marx, M.
Designing SpeechActs: Issues in Speech User Interfaces. In Proceedings of CHI '95 Conference (Denver, CO, May 1995). ACM Press. 369-376.
CHI 97 Electronic Publications: Tutorials
