Conversation Analysis and the User Experience Allison Woodruff and Paul M. Aoki Palo Alto Research Center 3333 Coyote Hill Road Palo Alto, CA 94304-1314 USA [email protected], [email protected] aspects of human-human interaction. In a previous paper ABSTRACT We provide two case studies in the application of ideas [15], we attempted to illustrate these uses and provide “how drawn from conversation analysis to the design of to” instructions for incorporating a trained conversation technologies that enhance the experience of human analyst into the iterative design process. We continue that conversation. We first present a case study of the design of discussion here, showing how CA has led us into new an electronic guidebook, focusing on how conversation research areas. analytic principles played a role in the design process. We Our story consists of two interrelated narratives. After a then discuss how the guidebook project has inspired our brief description of CA, we discuss our project on the continuing work in social, mobile audio spaces. In design of an electronic guidebook for historic houses that particular, we describe some as yet unrealized concepts for facilitates social interaction between visitors. This is adaptive audio spaces. mature research, largely conducted during 1999-2002; the discussion demonstrates the use of CA to identify INTRODUCTION conversational structures that are important in facilitating The sociological discipline of conversation analysis [9] users’ social goals. We then turn to a description of an (hereafter CA) has long held a significant position in user ongoing project on the design of mobile audio experience design. The idea that ethnomethodology, and communication systems. The idea is to provide CA in particular, can have direct application to design is communication technologies that actively facilitate social widely credited to Suchman [12], who not only proposed interaction by monitoring spoken conversation in a their methodological use in the study of human-machine mediated communication channel, recognizing the presence interaction but also observed that an awareness of human of specific conversational structures, and then changing social practices of (e.g.) conversational repair can itself specific properties of the communication channel to support serve as a resource for design. Since then – even leaving the social goals implicit in the use of these structures. This aside studies of technology use that apply conversation is research that has resulted in some early prototypes but is analytic methods – a number of attempts have been made to still very much in-progress. apply CA to HCI in a very direct way; we briefly discuss some of these in [15]. CONVERSATION ANALYSIS Given this history, which is nearly as long as that of the Conversation analysis, the most visible and influential form SIGCHI conference itself, it may seem odd to speak of CA of ethnomethodological research, is concerned with as anything but a traditional methodological source for describing the methods by which the members of a culture experience design. “Tradition” in the sense of history, engage in social interaction [9]. A key goal of CA is to however, does not imply “traditional” in the more common examine social interaction to reveal organized practices or sense of customary or characteristic use. Most popular patterns of actions, under the fundamental assumption that design methods, such as contextual design, are broadly interaction is structurally organized. Social actions include applicable and can be learned from courses or textbooks. talk, gesture, and use of objects. While ethnomethodology By contrast, CA focuses on human-human interaction, and and CA share this concern for how actions are organized, a professional level of proficiency in CA methods is best the goal of CA is to describe both how sequences of action attained through long practical apprenticeship. As a result, are organized and situated in a particular instance of researchers often find it difficult to apply CA to design in activity, as well as to abstract features that generalize across ways that are both productive and consistent with its a collection of similar instances. sociological outlook. A conversation analytic research program involves Nevertheless, CA can be very helpful in system design, analyzing a collection of interactive encounters. The most clearly for systems that involve spoken language. We analysis is twofold. First, the analyst makes a moment-by- base this claim on our own experience – for several years, moment, turn-by-turn transcript of the actions in each we have been drawing on CA to inform the design of encounter. Second, the analyst examines these encounters computing technologies that are intended to facilitate 1 individually and then comparatively to reveal a practice’s descriptions delivered through speakers played into open generalizable orderliness. air, and audio descriptions played through headphones. Analysis of the use of this prototype revealed that visitors To make the discussion more concrete, consider the had a strong preference for speaker audio since it afforded a procedure we used in our own studies of electronic shared listening experience for companions. Further, CA guidebook use. As we discuss further below, we use CA to revealed that when visitors had a shared listening describe visitors’ systematic practices as they use an experience, they oriented to the guidebook as though it was electronic guidebook to tour a historic house with a a human participant [14]. This was achieved through companion. To identify these systematic practices, we careful audio design (e.g., using audio clips that did not examine in detail the data collected during selected visits. exceed expected human turn lengths; scripting audio Specifically, for each visit, we create a video that includes content in ways that a human speaker would “design” a the audio and video recordings of the visitors, as well as conversational turn; ensuring that listeners could hear audio audio of the descriptions and video of the screens of each clips simultaneously). Visitors structured their visitor’s electronic guidebook. The resulting data are conversations around the guidebook’s audio, creating a transcribed and analyzed. Both of these steps require place for it in their social interaction, e.g., visitors made a careful, repeated viewing of each video. place for the guidebook to take turns in the conversation. More specifically, CA demonstrated that visitors oriented to COMPLETED WORK: ELECTRONIC GUIDEBOOKS the guidebook descriptions as though they were stories, In this section, we review the salient aspects of some work following discourse patterns that have previously been we have done on handheld audio guidebooks for historic observed with human storytellers [8]. This was desirable houses. This information has been reported elsewhere in since it meant that the flow of the visit could take the form more detail [2,14,15] but it forms an important backdrop for of an ongoing conversation between visitors into which the the section that follows. guidebook content could be fitted, as opposed to a series of Visitors often go to cultural heritage locations, such as long lectures from the guidebook. museums, with companions. Many seek what has To minimize “audio clutter” when a large number of sometimes been called a “learning-oriented” experience [6]. visitors wish to play descriptions simultaneously, we To facilitate learning, institutions typically present designed a second prototype with a technological information through guidebooks and prerecorded audio mechanism that allows visitors to have a shared listening guides as well as through labeled exhibits and docent-led experience with headphones [2]. Specifically, devices are tours. However, sharing the experience with companions is paired and communicate via a wireless network. Each often a higher priority than learning, particularly for visitor in a pair always hears the content they select infrequent visitors [7]. Unfortunately, typical presentation themselves, and additionally, each visitor has a volume methods interfere with the interaction among visitors. For control for determining how loudly they hear content from example, visitors frequently complain that audio tours with their companion’s guidebook. headphones isolate them from their companions, and visitors have few opportunities to interact effectively with CA studies of the use of this second prototype indicated that each other while docents “lecture” to them. the shared listening experience was preserved and that the visitors continued to orient to the guidebook as a storyteller. Our project had the goal of designing an electronic Further, the analysis revealed interesting patterns of guidebook that would facilitate rather than hinder social engagement. In general, when people are gathered together interaction. To achieve this goal, our project followed an and involved in an activity, conversational interaction may iterative design process. We designed and implemented occur, then lapse, then occur again. After a lapse, people two major prototypes (as well as a number of more re-engage turn-by-turn talk; alternatively, when people incremental prototypes). We observed visitors using the suspend turn-taking and dis-engage turn-by-turn talk, a prototypes during self-guided tours of a historic house and lapse occurs. To accomplish states of re-engagement and we conducted semi-structured interviews about their dis-engagement, people