Personification: Crossover between... 1 RUNNING HEAD: Personification in Computer Mediated Communication Personification: Crossover between Metaphor and Fictional Character in Computer Mediated Communication

This contribution defines the psychological processing of the artistic, cultural, and communication aspects of agents in information interfaces and characters in media entertainment (i.e. video games) as a new research topic in communication studies and computer science. It offers a theory of what personifications such as (intelligent) adaptive navigation assistants for Web environments can achieve, cognitively and emotionally: Personifications can be user-friendly, be motivating, and suggest ideas but only if their behavior is task-relevant. The personification should not only be more-or-less pertinent to the interface goals or story line but also be adaptive to changes in tasks, personal and business goals, and work processes as these change over time. This paper offers a multiple-meanings theory of metaphoric communication (i.e. personifications) and a general communication model for human-agent interactions. Although a theoretical exploration, it offers methods for empirical verification, blending technical tools from EEG-measurement labs, reaction-time recordings, and structured questionnaires. Personification: Crossover between... 3 Personification: Crossover between Metaphor and Fictional Character in Computer Mediated Communication Communication between humans and computers sometimes seems a battle zone of tragic heroes (us) against invincible mechwarriors (them). Whenever emotion and intelligence meet technical rigidity, a communication vacuum emerges in which the end-user becomes the ‘end-looser,’ suffocating in the absence of human-centered feedback. Designers of computer systems and communication interfaces develop embedded agents as the utmost user-friendly software application. For example, agents supposedly help performing tasks, help achieving (business) goals, and facilitate work processes. One of the most important functions of agents is to emulate face-to-face multimodal communication between the system and the user. Personified multimodal communication would be richer and easier to comprehend than unimodal communication such as text. Personified interfaces can provide backchannel feedback to users in a more natural and less obtrusive manner than interface widgets such as dialog boxes (cf. Mayer, 1997; 2001; Mayer, Heiser, & Lonn, 2001; Mayer, & Moreno, 1998). Although it is widely admitted that integrating human interpretation is paramount in interface design, yet agent theory mainly “... is concerned with the question of what an agent is, and the use of mathematical formalisms for representing and reasoning about the properties of agents.” (Wooldridge & Jennings, 1995). Although that may be a bit of an overstatement (see Paiva, Machado, & Prada, 2001; Machado, Paiva, & Prada, 2001), there is a need for sophisticated theory that values the “artistic nature of the problem.” (Reilly, 1996). Reilly states that “one of the key steps in creating quality interactive drama is the ability to create quality interactive characters (or believable agents). Two important aspects of such characters will be that they appear emotional and that they can engage in social interactions.” As much as designers look into the technical requirements of agents (software architecture, rendering), they are less focused on the user requirements, i.e. the prerequisites of human communication. Inspection of the World Wide Web, for example, shows us Eliza-style chatterbots that rely on pattern-matching techniques to simulate intelligent conversational behavior. However, it is much more interesting to understand the agent’s impact on the user’s cognition. If this is well understood, then it remains an engineering challenge how to program them accordingly. The question is whether agents as compared to dialog are the ultimate solution to all possible applications for which one might use computers. Probably not (cf. Craig, Gholson, & Driscoll, 2002). Yet, agents are extremely interesting communication devices. In occupying the crossroads of art, culture, psychology, communication, and technique, they have the potential to become many-sided, and therefore, mature communication partners. It is worthwhile investigating for which applications they are suited. To get there, we need a revision of the computer system-centered approach. We need to treat intelligent assistants as cultural products (i.e. as fictional characters that sometimes crossover with metaphors), subjected to psychological processes of perception and (emotional) experience. Certain agents serve as metaphors for abstract ideas of, for instance, help and guidance, and therefore, they are personifications. If we can establish a general theoretical framework for communication through agents, empirical tests may follow that can decide on the specific requirements for particular application domains. The present paper takes up the first part by developing a new theory of human-agent communication. It wants to follow up on the work of Laurel (1991, p. 356), who states that: An interface agent can be defined as a character, enacted by the computer, who acts on behalf of the user in a virtual (computer-based) environment. Interface agents draw their strength from the naturalness of the living-organism metaphor in terms of both cognitive accessibility and communication style (p. 356). To explain and improve the supposed strength of agent communication, integration should take place between theory of processing fictional characters and theory of processing metaphors. This kind of theory is urgent and still wanting. It is becoming very relevant to software developers. Agents, avatars, bots, virtual and electronic creatures and icons inhabit the new generation of software in video games, active worlds and search engines on the World Wide Web. Our screens – and our lives (vide the rage for “tamagotchis”) – are becoming peopled by these electronic pets and guides. It is important to consider how they are perceived and experienced because that is the way developers can make them both attractive and user-friendly. Content providers such as publishers, media broadcasters, the entertainment industry, and cultural institutions are now preparing for distributing content online. A wealth of choices awaits the user. Any means that facilitates selection, classification, inspection, and manipulation of the huge supply and the retrieval and actual use of items will be welcome. With the increasing exposure to television, film, and digital media, emotional education and competence is acquired more and more via fictional characters (Buck, 1999). In the future, this topic will become increasingly important with the growth of, for example, multimedia performances, education through virtual reality, and ‘meeting’ people on the Internet (De Kerckhove 1991; 1995; 1997). Museums, cultural archives, publishing houses, and broadcasting companies employ multimedia devices (CD-ROM, Internet, and the World Wide Web) to bring culture to a larger audience. Therefore, it is necessary to investigate how graphic design and visualization techniques can improve the transfer of cultural information by multimedia. The issue is also pertinent to child psychologists and to media critics. It might even Personification: Crossover between... 4 throw much needed new blood in the world of literary criticism as its theory takes some of its cues from the ancient tradition of biblical exegesis: Indeed personification is the more accessible term for classical antiquity’s “allegory.” To what end and under what conditions can agents be put to use in communication interfaces (cf. Rickenberg & Reeves, 2000)? Are personifications appreciated differently from literal equivalents, other metaphors, or other fictional characters? Regarding cognition and emotion, is a personification differently processed from other types of metaphors and differently from other types of fictional characters? Is communication through personification more effective than via literal instruction, other metaphors, or other fictional characters? The theoretical aim of this article is to arrive at an interdisciplinary integration of literary theories on metaphor and fictional characters to help research into human-agent interaction with a communication model in which the cognitive psychology of users is described while processing personifications. Theory development should be facilitated by tests on the integrated model with various types of stimuli and real users. The results may support the practical aim, which is to promote better understanding of the relation between system usability and acceptance and to develop guidelines, models, and methods by which designers may be able to provide better human-oriented communication of computer systems. Agents as Fictional Characters Consider a merely voice-based information service that is accessed via a telephone line. There, the user may well believe that s/he is talking to a real or virtual person. This situation resembles the ambiguity of perceiving mediated people such as reporters and journalists (see Konijn and Hoorn, 2003, this conference). If the user is in the illusion of talking to a real person, the psychology of interpersonal communication applies. This situation falls out of the scope of the present paper. However, if the user/receiver is somehow aware that this conversation is with a machine, then the agent/sender is much the same like a character in a realistic theater piece or motion picture. The computer system offers a representation of a real conversation like the mediated journalist who is merely the representation of the real person. Therefore, the voice agent can be considered a character that is represented realistically but is yet fictional. This example stands for various types of agents such as “embodied conversational agents,” “life-like characters,” and “virtual humans.” They share the common approach in computer science to capitalize on (hyper)realistic rendering with neglect of other factors that evidently contribute to the receiver/user’s involvement and appreciation (Konijn and Hoorn, 2003; also Hoorn, Konijn, & Van der Veer, 2002). Figure 1. Herman the Bug, the animated pedagogical agent, teaches on plant growth. Animation agents like Herman the Bug (Elliot, Rickel, & Lester, 1997; Rickel & Johnson, 1998; Figure 1) are typical for fictional characters that are rendered unrealistically. This is not necessarily negative because fantasy and magical features may stimulate the imagination and increase fun, especially with children. This type of character is easier recognized as fictional but the theory to be presented in this paper pretends to account for all. Not only will certain agents be recognized as fictional, in many cases they will be relied on as teachers, guides, darlings, and servants. The roles agents play reflect the abstract function they have for performing the user’s tasks. As soon as this role or abstract function is recognized, the agent not only is a fictional character but can become a metaphor (e.g., a harbor pilot representing a software navigation facility). Agents that are fictional characters as well as metaphors for an abstraction correspond to the (artistic) device of personification (see below). Thus, a theory of communicating with agents should allow for realistic and unrealistic representation and integrate it with the metaphoric aspects of (certain) fictional characters. In conclusion, the theory of fictional characters should Personification: Crossover between... 5 be generic for all types of user-interface agents, animated agents, embodied characters, helper agents, tutor agents (e.g., Anderson, Corbett, Koedinger, & Pelletier, 1995), virtual product presenters (Figure 7), and automated news readers (Apple, 1990) as long as the receiver/user does not qualify them as real people (cf. Reeves & Nass, 1996). The theory of personification is valid for those agents that moreover have a metaphoric twist. Interaction Context An approach most suited for describing the interaction context of a user (group) in communication with a computer system is Groupware Task Analysis or GTA (Van der Veer & Van Welie, 2000). Figure 2 depicts an ontology of the task world of a user. In GTA, agents can be users as well as software components (also other than the puppets on our screens). In our case, however, the second meaning of agent does coincide with the puppets on our screens. Users and agents perform tasks like searching, navigating, and exchanging information. They use objects such as databases and hardware devices to perform those tasks. The roles agent and user play often is of sender and receiver, of supply and demand, of helper and helped, respectively. However, these roles may switch during the process when for instance the agent needs more information or the user interferes with an agent’s task. This is called interaction. All this action and interaction serves to an end, which is the goal of the user, for instance, to find information, to be entertained, to reach a business goal (cf. Marketing Information Systems), or to streamline a work process (e.g., tips and suggestions by Microsoft’s Clippit). Figure 2. The task world of the user (Van der Veer & Van Welie, 2000). External and internal events may induce the need for new tasks, for setting new goals, and inventing new work processes. By pressure from the market, for instance, a digital library can develop into an e-commerce company. The change in business model from an educational to a commercial company has an impact on the business goals that are set and the business processes that should help achieving those goals. Offering high quality content and keeping an exhaustive stock may be of interest to an educational institution but may be cost-intensive in the eyes of a commercial service. This changes the user’s requirements of a computer system or software application, in our case, the software agents. A library would want an agent to immediately point out new works that might be of personal interest to the user or improve the user’s comprehension of a scientific problem. Commercial industries, however, would hope that an animated agent attracts more traffic on their pages because it is fun to interact with a somehow life-like entity. They hope for ease of communication because the agents to a certain extent emulate human-human communication. Yet, Dehn and Van Mulken (2000) provide a survey on empirical work in the area and found that an animated agent does not necessary improve a user’s comprehension or recall of information. The added value has more to do with motivational aspects, e.g., learners maybe willing to spent more time with the learning application when an agent is present, or may feel less anxious in a comprehension test conducted by a synthetic tutor. Whatever the effect of an agent may be, the central issue is that the goals of the user are satisfied by the agent and that the work Software