Happy Chatbot, Happy User

We compare our own embodied conversational agent (ECA) scheme, BotCom, with seven other complex Internet-based ECAs according to recentlypublished information about them, and highlight some important attributes that have received little attention in the construction of realistic ECAs. BotCom incorporates the use of emotions, humor and complex information services. We cover issues that are likely to be of greatest interest for developers of ECAs that, like BotCom, are directed towards intensive commercial use. 1 Using ECAs on the Internet Many embodied conversational agents (ECAs) are targeting the Internet. However, systems that are bound to this global network not only benefit from several advantages of the huge amount of accessible information provided by this medium, but inherit its common problems as well. Among those are the difficulty of relevant search, complexity of available information, unstructuredness, bandwidth limitations etc. So, what are the main arguments in favor of deploying an ECA on the Internet? First of all, the preference for real-time events, real-time information flow, expresses an innate need of mankind. Internet ECAs have this advantage as opposed to any other on-line customer-company communication method, such as web pages, email, guest books, etc. In addition, secondary orality, the communication by dialogues as opposed to monologues, is also far more effective when dealing with humans [5]. Furthemore, even though ECAs and simpler chatterbots may give wrong answers to certain questions, they create some sort of representation of themselves in the customers mind [13]. An ordinary website can be considered not only less interactive than one with an ECA, but the way it operates is closer to monologues than to dialogues. We have developed “BotCom”, a fully working prototype system, as part of a research project. It is capable of chatting with users about different topics as well as displaying synchronized affective feedback based on a complex emotional state generator, GALA. Moreover, it has a feature of connecting to various information T. Rist et al. (Eds.): IVA 2003, LNAI 2792, pp. 5-12, 2003.  Springer-Verlag Berlin Heidelberg 2003 6 Gábor Tatai et al. sources and search engines thus enabling an easily scalable knowledge base. Its primary use will be interactive website navigation, entertainment, marketing and education. BotCom is currently being introduced into commercial use. There is no space to discuss all the features and interesting implementation experiences with our BotCom ECA in this paper. Therefore we focus on some highlights where, we think, our ECA is special or when a theoretical or practical observation has proved to be particularly useful, so that others might benefit from these as well. 2 Comparison of Popular Internet Chatterbots During design and implementation we have analyzed, evaluated and constantly monitored existing ECAs in order to reinforce and validate our development approach. We did not follow only one methodology; several of them ([2], [4], [12], [15]) served as a basis of our own compound method, as, in spite of the similarities, overlaps frequently occurred and all of them contained unique evaluation variables. We studied the following (either commercial or award-wining) chatbots (see Table 1. for the results): Ultra Hal Assistant 4.5 (Zabaware, Inc., http://www.zabaware.com/assistant) Ramona (KurzweiAI.net, http://www.kurzweilai.net/) Elbot (Kiwilogic, http://www.elbot.com/) Ella (Kevin L. Copple, http://www.ellaz.com/EllaASPLoebner/Loebner2002.aspx) Nicole (NativeMinds, http://an1-sj.nativeminds.com/demos_default.html) Lucy (Artificial Life, http://www.artificial-life.com/v5/website.php) Julia (Conversive, http://www.vperson.com) 2.1 Visual Appearance In most cases visualization is typically solved by 2D graphics focusing only on the face, or photo-realistic schemes of still pictures (photos). Some tend to limit animation to only certain parts of the body (e.g. eyes, lips, eye-brows, chin), the roles of which are considered to be important in communication [11]. 3D animations are also applied occasionally, for instance in Lucy’s case. Despite the more lifelike and realistic appearance of 3D real-time rendered graphics, there is no underpinning evidence of differences in expressiveness amongst cartoons, photos, movies etc., though various studies confirm that users assume high-quality animated ECAs to be more intelligent [15]. Aiko, a female instance of BotCom, runs on the users’ web interface. The representation of her reactions and emotions is implemented through a 3D pre-processed (pre-rendered), realistic animation. Since the face and gestures provide the significant secondary communication channels [8], only the head, the torso (shoulders, arms) and occasionally the hands were visualized. To be able to diversify and refine the reactions, the collection of animations is extendable, but the right balance should be kept