Designing Interfaces to Maximize the Quality of Collaborative Work Judee K. Burgoon Bjorn Bengtsson University of Arizona Umea University Judee@u.arizona.edu bjorn@cs.umu.se Joseph A. Bonito Artemio Ramirez, Jr. University of Arizona University of Arizona Jbonito@u.arizona.edu Ramireza@u.arizona.edu Norah E. Dunbar University of Arizona ndunbar@u.arizona.edu Abstract Technological advances provide designers with tools to develop interfaces with anthropomorphic qualities. How- ever, it is not known how human participants accommo- date such design features in their interactions with com- puters, nor do we know if these features facilitate or hin- der information exchange and task performance. Study 1 examined the properties of mediation, contingency, and modality richness, whereas Study 2 examined the prop- erty of mediation. Results show that the some design features are better than others given the goal of the en- counter (e.g., passive involvement vs. relation building). Discussion focuses on the relation between user percep- tions, design features, and task outcomes. 1. Introduction Rapidly developing technology now affords organiza- tions, business, individuals and institutions of learning a cornucopia of options for engaging in communication and information exchange, especially for conducting dis- tributed tasks and collaborative work. Computer- mediated communication (CMC) is becoming ubiqui- tous, not only in such familiar text-based forms as email, NetMeetings, and computerized group support systems but also multimedia forms such as videoconferencing. Increasingly augmenting these tools are the use of com- puters as intelligent agents to conduct part or all of a transaction with individuals. Such human-computer in- teraction (HCI) utilizes computer interfaces that present and process information according to a set of predefined algorithms and that come in a variety of guises with various levels of functionality. For example, callers may be given a menu presented by a recorded or synthesized voice, and options are selected by pressing the appropri- ate pad on the caller's telephone. Technological ad- vancements in voice recognition software, synthesized voices, computer animation, and the addition of other nonverbal cues that simulate human facial expressions and gestures mean that interaction with virtual agents, as well as mediated human-human communication, is no longer limited to text or keyed responses. Other advances in artificial intelligence are also enabling more adaptive responses to human input by relying on for example, parsing of natural language text. When many options exist for designing a given communication and information system, the temptation is to include them all merely for the sake of including them. The real goals of building a computer interface to support collaborative group work, however, ought to be to facilitate clear and accurate information exchange, effi- cient transactions, high-quality collaborative work, and the like, i.e., to optimize the collaboration purposes for which the interface is being utilized. Therefore, the prob- lem facing designers is to understand how features of computer interfaces relate to features of the interaction between human users or between human user and com- puter, and the nature of the outcomes associated with the transaction. The investigations to be reported here offer a multi- dimensional look at this issue by examining user percep- tions when conducting the same task via several forms of CMC, HCI, and face-to-face (FtF) interaction. Results from two experiments are reported in which we examine user experiences along dimensions related to evaluations of the communication format or mode, the communica- tion process, and task partners. Study 1 was undertaken to explore the impact of increasing anthropomorphism in an HCI interface. The effects on decision quality, influ- ence, accurate understanding, and judged credibility 1 1 of task partners are reported in fuller detail in a separate HICCS report (see [1]). Of interest here is how users’ assessments relate to these outcomes, with an eye to identifying which features of interfaces yield the most beneficial results. The second investigation replicated the first but with human-human interaction under CMC and FtF conditions. Once again, emphasis is on how self- reported user experiences relate to the outcomes of deci- sion quality, influence, accuracy, and credibility. Because results of Study 2 come from an experiment still in pro- gress, only a subset of experimental conditions is pre- sented here. Even at this preliminary stage, however, many provocative relationships are evident that have im- plications for designing computerized group support sys- tems and computer interfaces for CMC and HCI. 1 Credibility is a multidimensional construct itself, comprised of interrelated judgments of character, truthfulness, competence, socia- bility, dynamism, and dominance. Proceedings of the 32nd Hawaii International Conference on System Sciences - 1999 0-7695-0001-3/99 $10.00 (c) 1999 IEEE Proceedings of the 32nd Hawaii International Conference on System Sciences - 1999 1