ISSN 1068-5723 May 16, 1994 Volume 2 Issue 2

HUMAN-COMPUTER-HUMAN INTERACTION: HOW COMPUTERS AFFECT INTERPERSONAL COMMUNICATION by Rodney Fuller Claremont Graduate School

Abstract There are two contemporary paradigms of the human-computer interface (HCI) -- the conversation paradigm and the direct manipulation paradigm. Neither one of these paradigms provides good models for designers of electronic media. A paradigm based on using the computer as a medium for conversation (rather than as the target of conversation) might resolve this problem. With this new paradigm, users of electronic media may be less likely to mis-perceive someone else's personality. Two groups of subjects were tested on their abilities to make personality assessments of people they had communicated with. The first group of paired subjects were users of traditional electronic media. Each pair consisted of Person A, who took the test as if they were someone (Person B) they had communicated with but had never seen, and Person B, who took the same test as if they were themselves. A second group of paired subjects who had never communicated using electronic media were asked to do the same task. The electronic media group consistently perceived the person they communicated with to be more analytical (p<.003) and judgmental (p<.03) than that person perceived themselves as being. There were no significant differences in perceptions in the face-to-face (no electronic media) group. Introduction Two contemporary design paradigms exist under the human-computer interface (HCI) umbrella: the human-computer conversation paradigm, and the direct manipulation paradigm. "The HCI banner is used to cover a wide range of interests and topics, but it usually refers to an individual human [computer] user interacting with a computer workstation.... It is often thought of in terms of a conversation paradigm in which the user converses with the machine" (Moran and Anderson, 1990, p. 381). The second and more recent paradigm in HCI is the "direct manipulation" paradigm in which the user interacts directly with his environment, using the computer as a tool (Moran and Anderson, 1990; Shneiderman, 1992). Although these two design paradigms cover the design needs of a vast majority of contemporary computer applications, they may not be the most appropriate paradigms for all of the applications that exist. One of these applications is electronic media (electronic mail and bulletin boards). The design focus in electronic media is one of using the computer as a medium for conversation between users, rather than one of a user conversing with the computer. This change in design focus (from object of action to medium for communicating action and intent) creates a new level of complexity for the designer of electronic media. It is well known that the selection of a data representation places an upper limit on the complexity of what can be modeled, as well as making some elemental forms of interaction ambiguous (Shannon and Weaver, 1963; Cover and Thomas, 1991). This creates special concerns for the electronic media designer because although the complexity of the human-computer interaction may be unambiguous, the complexity of the communication may overwhelm the medium (Sproull and Kiesler, 1986). Frank Ludolph (A Xerox Alto/Star and (original) Apple Macintosh designer) has said: As to paradigm shifting from workbench to communicator, I'm not sure that the latter precludes the former. Perhaps the communicator is just a tool on the workbench? To fully shift to communicator, the workbench work would likely be performed by the computer as an intelligent agent that an individual communicates with. A full shift would effectively turn an individual into a paraplegic, incapable of dealing with the world directly. I think that the best path is co-existence of the two paradigms (Personal communication, January 9, 1991). If the two paradigms -- human-computer interaction (HCI) and human-computer-human interaction (HCHI) -- are to coexist, then guidelines regarding how to combine these two paradigms need to be created. But before guidelines can be developed, we need measures that will determine how effectively the interaction models the behavior of those participating. One possible measure--the use of social cognitions to measure the effectiveness of communication behavior -- is investigated here. Two common types of electronic media are electronic mail and bulletin board services. Electronic mail (e-mail) is probably the most common example of electronic media. It consists of sending text from one computer user to another as if it were a letter. Because electronic media are intended for others' use, they could be thought of as facilitating human-computer-human interaction (HCHI). An interesting characteristic of communication by electronic media is that they allow the differences between design structures and communication patterns to be studied. Users of electronic media sometimes communicate with colleagues whom they have never met. If these interactions continue over a period of time the participants usually develop a mental picture of what the other person is like. It is not uncommon to meet someone with whom you have been communicating for several years by e-mail and realize that they are entirely different from what you had thought they were like. This is an example of a "breakdown" in communication (Heidegger, 1962). Winograd and Flores (1986) describe breakdown as a the invention of a network of understanding that is incorrect and private; an understanding that cannot be shared, and one that inhibits communication. Because of the limitations imposed on communication by electronic media (only text documents can be sent, with little chance for an interactive conversation), the fact that communication breaks down cannot be surprising. If one person's model is highly discrepant from another person's model, a breakdown in communication will occur. The models of other people's expectations and prior knowledge that people bring into communication can influence not only the tone of the discussion, but also the expectations of one person regarding someone else's personality. They can influence the communication patterns and styles of the person sending the message. Recent work regarding the effect that social cognitions (judgments and attitudes) about a stereotype have on an individual's communication patterns has shown that people can mis-communicate in predictable ways, depending upon the model or stereotype they employ (Fiske and Taylor, 1991). While it may be possible to coexist electronically without ever facing this breakdown in expectations, any discrepancy in communication models becomes more readily apparent when talking face to face. By measuring the perceived personality traits of people with whom participants communicate frequently by electronic media, but whom they have never met, and then comparing these judgments to that person's own perceived personality traits, we can determine tendencies towards differences in perception that are common in electronic media use. If these tendencies follow a common pattern, and if users of electronic media develop certain types of mis-perceptions more often than others, then these tendencies might be taken by computer systems designers as pragmatic guidelines of probable user errors as they build electronic media systems. Method The Myers-Briggs Inventory is a short, well known test that has high reliability on several basic personality scales. It has been used before in HCI to test the styles of programming that computer programmers use (Shneiderman, 1980). A shorter test modeled after the Myers-Briggs was developed by Keirsey and Bates (1978). While reliability of this version is unknown, it is more feasible to administer via questionnaire. The test determines scores on four sub-scales; introversion-extroversion, intuition- sensation, thinking-feeling, and judging-perceiving. These four sub-scales are then combined to give a personality "type." Four hundred and fifty tests modeled after the Myers-Briggs instrument (Keirsey and Bates, 1978) were distributed to two different groups of participants: People who took the test as if they were a co-worker who was known only by communication through e-mail, and people who took the test as if they were a co-worker whom they had never communicated with through e-mail (the face-to- face group). For the email-only group, 200 tests were distributed at random to members of the Cognitive Science Society (a multi-disciplinary field). Two hundred more tests were distributed at random to the attendees of the 1990 conference on Computer Supported Cooperative Work. For both groups, only people who listed e-mail addresses in the United States were contacted. Respondents were asked to identify someone whom they had communicated with, but with whom they were personally unacquainted, and to fill out the test imagining that they were this other (targeted) person. In the face-to-face group, 28 graduate students and 22 employees at a retail outlet were asked to identify someone they had worked with but with whom they had never communicated electronically. The same questionnaire was then sent to those people who were identified (targeted) by the original participants. This second group of people filled out the questionnaire as if they were themselves. After these measures were obtained for each pair of participants (one taking a test as if they were someone else, and the someone else taking it for his or herself), the scores were compared to see if users of electronic media are more likely to develop certain mis-perceptions more often. Results Ninety-three questionnaires were returned from the original mailing of 450. In the email-only group, 60 tests were returned, 30 of which were returned blank with notes explaining that the respondents did not communicate with anyone whom they did not know personally. Fourteen of the remaining questionnaires were returned from a second follow-up mailing. Thirty-three questionnaires were returned from the original face-to-face participants, with 26 returned from the second mailing (see Table 1). Table 1 EXPERIMENTAL DESIGN AND RESPONSE RATES ----------------------------------------------------------------- Response Rate Usable Rate Respondent Type N % N % ----------------------------------------------------------------- A (original) email only 60 15 30 7.5 FTF 33 66 33 66 A' (targeted) email only 14 45 14 45 FTF 26 52 26 52 ----------------------------------------------------------------- Note: Half of the people who responded from the email only group said that they do not communicate with anyone that they don't know personally. The individual responses were allocated to the 4 sub-scales: introversion-extroversion, sensation-intuition, thinking-feeling, and perceiving-judging. Correlations on the sub-scales between participants in the e-mail group exhibited two significant differences. The original respondents were more likely to rate the targeted respondents high on the thinking sub-scale (and low on the feeling scale) than were the targeted respondents themselves (r=.83, p<.003). Original respondents were also more likely to rate the targeted respondents as more judgmental (and less perceiving) than were the targeted respondents themselves (r=.60, p<.03). No significant differences between the perceptions of the paired respondents were found for the face-to- face group. Discussion Users of electronic media develop consistent mis-perceptions across several personality dimensions when judging others that they communicate with, but whom they do not know. These perceptual differences contrast with the lack of perceptual differences found within the face-to-face group. The significant differences in the e-mail group versus the lack of differences in the face-to-face group indicates that either people are not using electronic media as it was conceived, or that the technology limits the quality of communication that can take place between participants. These findings imply that users of electronic media make consistent and erroneous judgments concerning the thinking and judging dimensions of others' personalities. These limitations could be influenced by either the limited bandwidth that electronic media imposes upon human communication or by the use of electronic media outside of it's intended design function. One important finding is the fact that half of the original email-only participants explicitly mentioned that they do not communicate with anyone that they do not know personally. Although this study does not address the reasons why half the users of electronic media failed to maintain relationships based solely on electronic communication, this surprising result might lead to the commercial development of more semantic human- computer communication models for electronic media (see Targowski and Bowman, 1988). (The lack of "e-mail only" relationships is especially surprising since the majority of the participants were computer or cognitive scientists, who might be presumed to be comfortable with computer-based media.) It remains to be seen whether future systems designed to prevent predicable breakdowns that occur during e-mail use will result in an increase in relationships maintained only through electronic media. In summary, this study shows that social cognitions can be used to measure the effective bandwidths of electronically mediated conversations. The use of Myers-Briggs tests and other similar measures will provide designers of future electronic media with guidelines for combining today's conversation and direct manipulation paradigms with tomorrow's communication paradigms. Notes 1. This research was partially supported by the Jenness Hannigan Research Fellowship. References Cover, T.M. and Thomas, J.A. (1991). Elements of Information Theory. New York: John Wiley and Sons. Fiske, S.T. and Taylor, S.E. (1991). Social Cognition (2d ed). New York: McGraw-Hill. Heidegger, M. (1962). Being and Time. New York: Harper and Row. Keirsey, D. and Bates, M. (1978). Please Understand Me. Del Mar, Calif.: Prometheus Nemesis Book Co. Moran, T.P. and Anderson, R.J. (1990). The workaday world as a paradigm for CSCW design. Proceedings, Conference on Computer-Supported Cooperative Work, October 7-10. Los Angeles. Shannon, C.E. and Weaver, W. (1963). The Mathematical Theory of Communication. Urbana, Ill.: University of Illinois Press. Shneiderman, B. (1980). Software Psychology: Human factors in Computer and Information Systems. Cambridge, Mass.: Winthrop Publishers. Shneiderman, B. (1992). Designing the User Interface: Strategies for Effective Human-Computer Interaction (2d ed). Reading, Mass.: Addison-Wesley. Sproull, L. and Keisler, S. (1986). Reducing social context cues: Electronic mail in organizational communications. Management Science 32, 1492-1512. Targowski, A.S. and Bowman, J.P. (1988). The layer-based, pragmatic model of the communication process. The Journal of Business Communication 25, 5-24. Winograd T. and Flores, C.F. (1986). Understanding Computers and Cognition: A New Foundation for Design. Reading, Mass.: Addison-Wesley.