Workshop on
How much human-like should an Intelligent Interface be ?
in conjunction with the International Conference on
Advanced Visual Interfaces ( AVI'04 )
Gallipoli (Lecce), Italy, May 25, 2004
In an increasing number of application domains, new forms of intelligence are being added to interfaces, which implement the Stanford's 'Computers as Social Actors' metaphor, by giving to users the illusion to interact with an intelligent assistant or companion. Such interfaces aim to emulate human-style multimodal interaction in order to make human-machine communication more intuitive and expressive and also less challenging to use and therefore more attractive for less experienced user and even “computer illiterates”. The intelligence of such an interface manifests itself in a number of human-like qualities including eterogeneous abilities, such as recognizing and interpreting the affective state of the user and adapting their behavior accordingly, expressing a 'personality', showing some degree of autonomy and, more in general, establishing a social relationship with the user.
There is an ongoing debate regarding the degree of 'naturalness' or 'resemblance to human behavior' that should be taken as the objective to achieve. This raises the question of the range of properties (both peripheral as well as internal) which are required for an intelligent interface, and the question of how these properties should be implemented. For instance, humans are able to cooperate, influence, help, manipulate, be sincere or insincere, be polite or impolite, behave rationally or irrationally, show or hide their emotions and so on: even behaviors which may appear as 'irrational' may have their reason to exist and may be justified. Therefore, do we expect the same range of behaviors from an intelligent interface or do we prefer to establish rules which renounce to mimic the multifaceted characteristics of human behavior to insure us the 'effectiveness' of results? And does a decision on this depend on the application domain? For instance, we might decide to prefer a 'rational' decision to an 'emotional' one in medical applications. But we probably prefer some more advanced form of social and affective interaction in other domains, such as access to information services. In case of using animated interface agents, do we expect our agents to be realistic in their body appearance or can we establish a social relationship also with some sketched or cartoon-based character?
The previous questions assume that we are able to emulate human-like behaviors and appearance in the intelligent interfaces we build. But is that true at the present stage of research? Which are the main problems still open? Which kind of interdisciplinary cooperation do we need to advance in this domain?
The purpose of this workshop is to bring together researchers and developers of user interfaces that incorporate human-like qualities in one or the other form in order to exchange ideas and experiences but also stimulate controversial discussions regarding issues such as:
- Which human-like properties do we need? E.g.,
- which personality traits are particularly relevant in an interface?
- should traits of interfaces aim to resemble human-like personalities?
- assuming that emotions are an essential constituent of reasoning in humans, should an interface simulate emotion feeling? And if, which ones in which situations?
- should we enable users to explicitly turn on and off affective behavior ?
- How to reflect human-like properties at the surface level? E.g.,
- what is an appropriate audio-visual appearance of an interface that possess a certain set of human-like qualities so that the supported qualities become transparent to the user without rising over expectations?
- can we provide interface developers with practical design guidelines based on experience what works and what doesn’t ?
- How to evaluate this kind of interfaces? E.g.,
- Can we apply the same usability criteria and evaluation methods that have been developed for 'traditional' interfaces?
- Do we need to introduce new methods, e.g., which allow us to determine the, say “emotional intelligence quotient” of an intelligent affective interface?
- What are the most promising application areas for this kind of intelligent interfaces?
- What are the particular challenges in domains such as eLearning, eCommerce, public information kiosks, computer games etc?
2. Participation & Submission Format for Extended AbstractsAuthors are encouraged to prepare a 2000 words extended abstract and to specify whether they wish to show a demo. All submissions (in PDF or Postscript) should be sent not later than 29 Feb. 2004 to:
3. Important Dates
- February 29. Deadline for electronic submission of Extended Abstracts
- March 28: Notification of acceptance to authors
- April 25: Deadline for camera-ready papers
- May 25-28: Workshop at AVI 2004
- June 15: Update of workshop web page with wrap-up and conclusions
4. Organisation Committee & Address for further information
Please direct any questions related to the workshop (such as submission of extended abstracts or participation without a submission) to the organisers.
5. Programm Committee
- Elisabeth Andre, Univ. Augsburg, DE
- Ruth Aylett, Univ. Salford, UK
- Daniel Ballin, BTexact, UK
- Cristiano Castelfranchi, CNR, IT
- Jonathan Gratch, ICT, USA
- Kristina Hook, Univ. Stockholm, SE
- Brigitte Krenn, ÖFAI, AT
- James Lester, North Carolina State Univ., USA
- Christine Lisetti, Univ. of Central Florida, USA
- Ana Paiva, INESC, PT
- Catherine Pelachaud, Uni. Paris 8, FR
- Isabella Poggi, Univ. Rome, IT
- Helmut Prendinger, Univ. Tokyo, JP
- Daniela M. Romano, Univ. of Sheffield, UK
6. Local Arrangements
For information concerning travel and accomodation see the AVI'04 pages.
- Valeria Carofiglio (Università di Bari)
- Dora Cavalluzzi (Università di Bari)
Page updated on 15. January 2004 by Thomas Rist