Can You See What I See? Developing Coherent Support for Distributed Collaborative Video Analysis with MiMeG

This paper describes the process of trialing our first version of the MiMeG system, an e-social science tool designed to support ‘distributed data sessions’ in which social scientists conduct real-time, collaborative video analysis across remote sites. MiMeG provides a shared workspace in which remote participants can playback video, communicate vocally through VoIP, and create and archive annotations over video in real-time, as well as interlink textual transcripts and other data. Here we detail how redesign work is being informed by our evaluation of the user experience, which took the form of video recording and closely analyzing ‘real world’ trials in which a research team were engaged in meaningful data sessions on real project issues. We outline the technical problems encountered with the MiMeG tool, some of the ways the problems were solved in practice, and how users ‘shaped’ the resources available to them. In particular, we highlight how local talk and action was used as a key resource in coordinating distributed work. We then specify how we are feeding the user experience into technical redevelopment work. Introduction: The Imprint of the Users Mixed Media Grid (MiMeG), a research node of the ESRC’s e-Social Science programme, is a collaborative project between the University of Bristol and King’s College London. The node is dedicated to better understanding collaborative video analysis and developing tools for its support across distributed research groups. Version one of our MiMeG software is freely available under a GPL open source license and enables multiple participants in geographically distributed sites to control and view video data synchronously. It allows researchers to create and archive real-time freeform annotations over the video using a penbased input system, to interlink textual transcripts or other relevant mixed data sources, and provides real-time audio communication through VoIP channels. The MiMeG software is designed to support the burgeoning groups of social scientists and practitioners who use video materials as a primary form of data. Group data analysis sessions with video, or ‘data sessions’, are a key work practice for the research communities engaged in video analysis. In these standard, co-present data sessions, multiple individuals gather to view, comment on, and collaboratively analyse raw qualitative data with colleagues, with a key strand of work being to have others ‘see’ (or see in a certain way) the phenomena which you can ‘see’ within the video data fragment under study. Given the increasingly distributed nature of research projects, bridging across institutions, disciplines and nations, there is a growing need for effective technological solutions and support for remote collaboration over video data. Indeed, there is a strong demand for systems and technologies to support virtually collocated research meetings in general, with a particular and persistent challenge in the development of CSCW systems being the design of effective support for synchronous collaboration over and around common documents, objects and datasets. Thus, MiMeG was designed to better support ‘distributed data sessions’ (Fraser et al., 2006), where groups of geographically dispersed researchers can simultaneously view video data and conduct meaningful analytic work with those data. Far from professing that MiMeG could replace face-to-face meetings, our software design addressed the demand for such systems to supplement existing face-to-face data sessions. Indeed, Olson & Olson (2000) suggest that remote collaborative tools tend to work best when participants also meet up regularly aside from their remote meetings. The technical development of the MiMeG software has been outlined elsewhere (see Fraser et al., 2006), but here we turn to discuss the (sometimes surprising) ways in which the software was used in practice to support the collaborative analysis of video data. The paper evaluates the user experience of the first version of MiMeG, outlining the technical and social problems facing participants during their remote collaboration over video data, including the synchronizing of playback and annotations across sites. Our study then attends to how participants confronted these problems in practice, through the use of talk, gesture and playback control, and takes seriously the ways that users ‘shaped’ the resources available to them. We then specify how we are feeding the user experience into the redevelopment of the software, towards MiMeG Version two. Studies: Remote Data Sessions We conducted a programme of in-depth research into the use of MiMeG Version one by various groups of social scientists, who collaboratively analyse video as part of their everyday work. We undertook video-based field studies of the trials of the technology and the analysis was informed by ethnomethodology and conversation analysis (e.g. Heath and Luff, 2000). In adopting this approach, rather than conducting a traditional, technical evaluation of the system and its properties or functionality, we have attempted to explicate the interactional practices that emerge in managing the technology in use. Our understanding of these practices is intended to stimulate further issues for research and to inform the design of future technologies. While MiMeG has been widely distributed to social scientists, we provided additional technical support and assistance to a small number of groups, who have agreed to be studied during their use of the system. The distributed data sessions were video recorded at each participating site, to capture data concerning the embodied conduct of participants including the use of media and objects, as well as talk, gesture, body movement, and the display screen in use. Whilst, on the surface, the tasks of remote data sessions are chiefly the same as copresent ones – those of spotting and getting others to see phenomena, making analytic claims, supporting claims with video evidence, etc. – our study points to the practices and coordination of these tasks being quite distinct in remote collaboration. This paper refers to distributed data sessions (through MiMeG) held by an existing research group comprised of four members, from three different departments, situated within two geographically remote UK universities. Consisting of close colleagues, collaborating together on a funded research project, they can be seen as a “gelled social group” (Aoki et al., 2003), and they engaged in a series of meaningful remote data sessions, tackling real project issues and tasks. We video-recorded two one-hour distributed data sessions by using two video cameras at each of the geographical sites – one camera to depict the data session participants clearly, and the other to display a close-up image of the on-screen activity. Their research project is concerned with car sharing and the group analyse video data of action and interaction in vehicles to explore such issues as journey routines, instruction-giving and wayfinding. For comparative purposes, we also recorded a series of co-present data sessions held by their group, and also by other research groups. Issues and Challenges in Deployment In the course of developing this software and putting it to work with existing research teams in the UK and elsewhere we have encountered a range of social and technical problems concerning the distribution, sharing and discussion of social scientific video data. These include, for example, a number of thorny issues associated with institutional firewalls. Firstly, while on the one hand universities are keen to support inter-institutional research, on the other they are keen to restrict access into their own networks. As a result, we have faced persistent (although variable, depending on the institutions involved) problems in securing solid connections between institutions to support MiMeG. Secondly the proliferation of video file types and compression standards also presents non-trivial challenges to those concerned with the remote sharing of video data. We realized how important the quality of the video is in order for the video analysts to get a good sense of the action and interaction depicted on screen. Thirdly, in trialing MiMeG, we encountered the persistence of paper. Whilst the technology enables textual transcripts of the video materials to be shown and shared onscreen, participants would continue to use paper to produce comments, notes and analytic memos. Fourthly, we have recognised how valuable it is for colleagues to be able to cluster around screens, so that they are within easy reach for referential work, and yet also able to sit back and produce personal, paper-based notes. Fifthly, while in co-present data sessions video analysts routinely enact or mimic on-screen behaviours to elaborate analytic points (see Tutt et al. 2007b), we realised that participants are very constrained in the flexibility available for embodied analytic work when using MiMeG. However, whilst a number of such issues arose, here we will consider one key issue in a little more depth. Remote Problem-Solving: Taking Notice of the Problematic and Incidental While analyzing how the technical and social problems were solved by the participants in practice, we particularly noticed the importance of an unexpected interactional resource, namely the interplay of ‘local’ (designed for co-present colleagues) and ‘general’ (designed for the remote site as well) talk and action (see also Tutt et al. 2007a). In this case, the work of fixing problems alerted us to the value of a misunderstood and potentially underused social resource in distributed work – local action. Consider Example 1, in which Henry faces the problem of getting others (at both the local and remote sites) to see something that he has noticed in the video data – in particular evidence of ‘no entry’ to a certain side street. The research team is studying a car journey through a cit