|
The field of human-robot and human-agent interaction has always acknowledged the relevance of nonverbal interaction, so that the method seems to come naturally to it. Two remarks, however, need to be made on the notion of "interaction": • From an ethnomethodological viewpoint, only the participants themselves establish whether they regard the exchange with an artifact as an interaction [24]. • Where human-human interaction is the issue, it is understood as more or less synonymous with communication, or conversation [13]. The field of human-machine interaction has always taken this concept more widely in both senses: here it includes both communication (social interaction) and operation (physical interaction), and action-reaction sequences as well as interaction proper. This is the sense in which we will use the concept here. If one wants to study the relationship between companions and their owners, one must not preclude, by misleading terminology, that this relationship can be both social and a-social – at different times or even at the same time: the companion can be seen both as something to talk to and as an object or machine. Video interaction analysis allows to include in the analysis other semiotic fields [25] than language use, and studies how participants mix and relate them. Transcripts of human-robot interactions usually leave these out. The reason simply is that robots are, to date, heavily deficient in sharing the semiotic fields of objects, spaces, gestures etc. One example of such a shared semiotic field in our field study is the house key, whose presence or absence is "known" to the Nabaztag through a micro-switch in the key hook, so that the meaning of the key as "going out" and "coming home" can be shared. On the other hand, the Nabaztag has no access to the semiotic field of any other household object. When one participant takes the key off with the trash bag in the other hand, the meaning of the action – taking the bag to the bin outside – is obvious to human observers from the same culture, but not accessible to the robot. Tests that limit the potentially relevant semiotic fields by clever design (e.g. in the lab) bear the risk that others than the targeted channel of interaction (e.g. language) are neglected. In studying human-robot interaction "in the wild", they make themselves felt and have to be taken into account, even if, at this stage, as not accessible to the machine. Observational data on real-life, long-term use of companions are extremely rare as yet, and in such cases, formulating hypotheses that can be tested in analysis can lead research far off the mark. Inductive explorative analysis is the approach of choice to prepare the field for later systematic hypothesis-based research, in that it can point to new and interesting phenomena and generate the concepts and questions with which to work in the field. The instruments used in this part of the study owe much Conversational Analysis (CA) [15] and Grounded Theory (GT) [14]. From CA, we take the microscopic look at the interaction, where the constant questioning of the data and the perspective of the particpants' orientation toward the interaction is used to attend to every potentially relevant detail. GT inspires the open approach to coding the material, as well as the research interest in uncovering not only the mechanisms of interaction, but also the attitudes, socio-cultural stance and practical theories that subjects develop and apply to their relationship with companions. The verbal interaction of the videos was transcribed roughly. Through repeated viewing, these transcripts were extended to full descriptions, including position, posture, gestures, facial expression and actions of the subjects. An effort was made to rigorously separate the description from any other remark, note, or thought that would inevitably present itself, following the procedure recommended by the documentary method [16]. Such interpretations were written down separately in notes and memos. Common keywords or titles to the notes started to present themselves, and these were noted down separately as the first tentative, low-level concepts. Other data invited to use the same or related concepts. In this process, the attention is drawn not only to those cases where the phenomenon giving rise to the concept is present, but also to the cases of its absence, as well as to its different dimensions. CLOSINGS IN HUMAN-PANION CONVERSATION Thus, closings of interactions in our data were first brought into focus by a few video sequences where participants concluded an interaction with a greeting. Subsequently, we analysed systematically for the different ways in which interactions were terminated. Only through taking both presence and absence of closings into account was it possible to develop the concept of "closure". Closure as a concept is here distinguished from the conversational segment "closing". Rather than to observed behaviour, it refers to the orientation of participants to the social norm of bringing interaction to an orderly end. It is part of everyday politeness not to terminate a conversation strictly unilaterally. If it is done nonetheless, reasons or excuses have to be given, or else such behaviour is sanctioned. While closings are the mechanisms by which this is achieved, closure as a concept points to an attitude toward the other which makes them necessary and desirable. In human-human interaction, this orientation is taken for granted to such a degree as to be invisible. This is different in human-machine interaction. In our data, we could distinguish, roughly, five different types of termination: • greeting: a verbal closing ending with a formal greeting, e.g. "bye", by the participant • verbal: the participant closes with an utterance which marks a termination or adjournment of the conversation, but no greeting. • nonverbal: in most cases, this is a nod by the participant before leaving or turning away. Turning away alone is not qualified as a closing, as it is already part of the action that follows the end of a conversation. • waiting: this behaviour can occur, in principle, in addition to and typically before any of the other ways of terminating an interaction, but in most cases, no closing follows. It cannot be said with certainty what exactly the participants are waiting for in these instances: it can be a continuation, a new topic, or a move towards closing by the Nabaztag. The identification of these pauses as "waiting for closing" resides in the observation of the contrasting cases where a closing of some kind occurs. • no closing: none of the above signs occur. The participant takes up whatever action occurs after the interaction: turning away, leaving the room, taking up some household task, etc. What can be seen is that the cases of no closing become more frequent as the field study goes on. There are, roughly, three stages: the first is characterized by pauses (waiting), the second is a stage of more frequent verbal closings, culminating in greeting, while afterwards the closing sequences decrease in frequency. Contrary to expectations, however, they do not disappear, and we even find again a sequence with greeting towards the end of the data collection period. With such a small number of relevant interactions even in our best sample, this kind of analysis does not allow any conclusions on the development over time. The map is shown here only to illustrate occurrence and distribution of closing-types. Schegloff and Sacks [17] talk of the closing of conversations as a "problem", not for the analyst, but for the participants in the conversation itself, in the sense of a task that participants have to execute at the end of each conversation with different partners, situations, contents, and modalities. The end of a conversation is (except for the intervention of external forces) not something that just happens, but has to be brought about by the participants in a collaborative way. Schegloff & Sacks have shown that closings and changes of topic are closely related, and that the closing problem can be seen as the opposite of the "turn-taking machinery" which is fundamental in generating turn after turn of a conversation in an orderly manner. To arrive at an orderly end of this stream, the participants have "to organize [their] simultaneous arrival ... at a point where one speaker's completion will not occasion another speaker's talk, and that will not be heard as some speaker's silence" [17]. 网站原创范文除特殊说明外一切图文作品权归所有；未经官方授权谢绝任何用途转载或刊发于媒体。如发生侵犯作品权现象，英语论文题目，保留一切法学追诉权。()，英语论文题目 |
