On the Design of Intelligent Memory Functions for Virtual Meeting Places: Examining Potential Benefits and Requirements Version 11

I. INTRODUCTION

On the other hand, virtual meeting places represent a major step beyond the classical computer-user relationship. Rather than having a single user interact with the machine, the computer-mediated user-user interaction is becoming a new paradigm. Communicating through computers enables new services which are to facilitate taking part in a cyberspace conversation, and which will assist the participants in keeping trace of the communicative actions performed during conversations. Besides, virtual meeting places can take advantage of multimedia technology allowing people to communicate through multiple modalities. In this paper, we use the term virtual meeting places in a broad sense and consider that they may simultaneously integrate traditional human-computer interaction processes (database or Web access, for example) and also be related to what could be called a whole group story made of synchronous and asynchronous interactions. In this contribution, we focus on the issue of shared memory functions. Our hypothesis is that a variety of added-value communication services can emerge from a system’s capability to memorise information units that have been obtained from observing how people communicate and interact which each other in a virtual meeting place. Our working background is a European project called Magic Lounge¹ in which participatory design studies achieved by Bernsen & Dybkjaer (1998), Masoodian & Cleal (1999), and sociological studies have been carried out. After an introduction to related work, we briefly sketch the project’s goals and provide some illustrative scenarios in order to demonstrate the potential of memory functions for virtual meetings and for human-human mediated communication in general. Though we will stay away from providing technical details on implementation work in this paper, we discuss some of the requirements and technical challenges that have to be tackled when implementing "intelligent" memory functions.

In the context of telecommunication, answering machines for the telephone constitute a first example of an add-on service that – at least to a small extent - relates to the concept of remembering. Though useful for many purposes, the mere recording of audio streams is only a very low-level memory function with the drawback of allowing tedious sequential access only. In our work we are especially interested in memory functions which go beyond the mere recording and storage of data and which are to support the participants of virtual meetings. The concept of virtual meetings changes in two ways the manner we think about a meeting. Firstly a meeting can be distributed in space, and secondly, a meeting can be distributed in time, therefore integrating synchronous and asynchronous interaction. In that respect, our work is related to what has been done in the area of Computer Supported Co-operative Work (CSCW) and computer-based conferencing systems.

Since Engelbart’s pioneering work on NLS (oN Line System), cf. Engelbart & English (1968), a large variety of so-called groupware systems and tools have been developed of which many are already available on the market (e.g., see Young & Buchanan (1996) for a listing of tools), while other systems are research prototypes exploring specific directions such as awareness (that is, how to mutually inform the collaboration partners about their states and activities) or explicit versus implicit structuring of conversation. As proposed by Johansen (1988) time and space are commonly used as the basic dimensions to categorise actual and possible CSCW tools and systems (see Table 1). Usually, the attributes “synchronous” and “asynchronous” are used to indicate whether a collaboration/conversation takes place at the same time or not. The attributes “co-located” and “remote” are used to indicate whether the collaboration partners are all present at the same place or whether they are at geographically dispersed locations. Virtual meeting places in that restricted context usually aim at providing collaboration support for a group of people who are not at the same place but who want to interact with each other at the same time. To avoid confusion, the notion of virtual meeting must not be mixed up with that of a virtual team room. In the later case, the metaphor of a shared space is often used just for a location on which to place electronic documents, notes or messages. Thus, the issue of supporting real-time interactions among the group members is less apparent in such systems.

Table 1: Using the “classical” time and space dimensions to characterise CSCW systems.

Until now most of the work done in the area of CSCW aims at studying how people work together and how collective software (i.e. groupware) affects the group behaviour. In such systems memory functions or often not available or limited to the mere recording of exchanged data. For instance, most of the systems do not provide any means for structuring information according to criteria other than the temporal axis (i.e. the sequence of exchanged messages). Furthermore, existing relationships between messages remain unrecognised and thus cannot be exploited to provide the user with coherent descriptions at a higher level of abstraction.

Recording and transcription services are also offered by some commercial teleconferencing operators (e.g., Genesys Inc. offers a transcription service for multi-party telephone conferences). To meet the required quality standards for business applications, however, such services are quite expensive as they are done manually by human operators. The Coral system prototype by Minneman et al. (1995) is a remarkable exception in the field of conferencing systems. In addition to the communication tools (e.g. for audio and video conferencing) the system also comprises a number of tools for recording, indexing and retrieval of media and collaboration activities. These tools are to assist a minutes taker in producing written documentation after a meeting.

Other types of groupware systems focus on asynchronous collaborations through documents. For example, system like Lotus Notes have been used by Bouillant (1997), Courbon & Tajan (1997), and Khoshafian & Buckiewicz (1998) to create, store and update databases for text documents. Therefore, Lotus Notes provides powerful functions for version management, synchronisation management, replication management, and search of information. In an enterprise, such a created document database may actually serve as a sort of collective memory to be consulted by a large number of different users. However, in this case the concept of memory is very close to the idea of a central container in which to store and retrieve information in a rather fixed way.

In order to identify possible functions that memory can play in groupware and teleconferencing systems, one can adopt the classification of co-operation types by Schael (1997) who distinguishes between: co-ordination (people have to co-ordinate their actions), collaboration (people have to work on a same task to achieve their goal), and co-decision (people have to aim at making a common decision). For each of these co-operation types, Schael defines requirements for tools, which aim at supporting communication and co-operation. Some of these requirements are directly related to memory capabilities. For example, a co-ordination tool should comprise a “conversation history” that allows the group members to trace back who has contributed to whom and what in a conversation. A collaborative tool should provide means to trace back how the results of collaborations (e.g. text documents with multiple authors) have been created. Co-decision tools should provide appropriate means to ensure that all participants share all the information that is relevant for decision making including decisions which have been made in previous meetings.

Much work in CSCW has aimed at supporting professionals at work. Thus, tools have been developed for more or less well defined tasks and workflows to support corporate users, e.g., see Mark, Haake, & Streitz (1995). With the opening of the Internet to the general public, text-based chat systems and derived advancements of them have become very popular among non-corporate users as well. Among such systems are especially MUDs (Multi-User Dungeons) and so-called Virtual Community Servers. MUDs enable remote but synchronous text-based communication run over the Internet, e.g., see Dourish (1998). While early systems were text-based only, more recent systems include other media as well. For example, a number of systems promote the concept of an avatar to co-represent a human user in an electronic space. Thereby, an avatar may appear as a 2D cartoon graphics like in Comic Chat by Microsoft Corp. (1998). Likewise, the electronic space may come as an emulation of a 3D environment inhabited with 3D embodied avatars, e.g. see Damer (1998). Yet other approaches aim at augmenting text-based communication by audio/video conferencing tools or shared whiteboards, e.g., Curtis & Nichols (1993), Tolone, Kaplan, & Fitzpatrick (1995). Most of these MUD-type systems keep records of the conversational contributions made by the participants. For instance, the chat tool of Microsoft’s Netmeeting™ offers the possibility to create a log file that lists all chat contributions from the different conversation partners. The entries in this log file include the message text, a date, a time stamp, and the sender of the text message. Since the messages as such are not processed further, there is only very limited support for accessing the recorded data in a way other than browsing through a temporally ordered list of entries. An attempt beyond the mere recording of textual messages has been made in the AIDE system by Mase, Sumi, & Nishimoto (1998). AIDE relies on keyword spotting techniques to guess the semantic content of dialogue contributions. Using a metric for semantic distance, clusters of related dialogue contributions are then displayed as visual groupings on a screen for the chat partners.

Last but not least, an interesting source of knowledge is given by previous studies on “real meetings”, i.e. people meeting physically in the same place and at the same time. This work includes taxonomies of meeting types as proposed by Ellis, Gibbs, & Rein (1991), the study of meeting phases or the identification of roles that participants may fill. Unfortunately, less work have been done in this area with regard to questions such as how people reuse the result of a previous meeting or how a late-comer gets informed of the current state of an ongoing meeting.