Figure 1: People using the OnTop interface to share and discuss their personal files.
Anthony Collins
University of Sydney
School of Information Technologies
anthony@it.usyd.edu.au
Tabletop interfaces provide a new medium for collocated collaboration. Consequently, tabletops need to support access to file systems, just as a core facility of conventional computer systems is to provide an interface to a file system. However, the constraints of tabletop interfaces call for rethinking standard approaches to file system interaction. This paper presents the design of OnTop, a novel associative-search approach to file system interaction: users navigate multiple file systems by selecting focus files, retrieving similar ones. We report a small-scale qualitative evaluation of OnTop against a more conventional file browser approach: OnTop was consistently preferred and found to be more efficient, especially for larger file collections.
Tabletop interface, file system interface, single display groupware
I.3.6 Computer Graphics: Methodology and Techniques—interaction techniques; D.2.2 Software Engineering: Design Tools and Techniques—user interfaces; H.5.3 Information Interfaces and Presentation: Group and Organization Interfaces
Tabletop interfaces provide a novel means for collocated collaboration. Indeed, the collaborative possibilities afforded by tabletop interfaces make them appealing for sharing digital files and personal information between group members, resulting in the need to support file exchanges between multiple file systems. Just as conventional computer systems provide an interface to a file system, tabletop interfaces will need to provide similar facilities to allow flexible, transparent and seamless sharing of files from multiple personal file systems. The ability to navigate and view files from a remote personal file system directly on a tabletop (without having to explicitly copy them to a file system on the tabletop itself) is an important facility for collaborative tabletop interfaces.
Hierarchical organisation and navigation of file systems has become the standard in conventional personal computer systems. However, the constraints of tabletop interaction—where there is typically no mouse or keyboard present, and the interface is highly visual—call for rethinking standard approaches to file system interaction.
First, the overwhelmingly large nature of file systems is a major problem for tabletop file system interaction because of the constrained display area and interaction primitives. Navigation and visualisation of a large hierarchy on a tabletop display would require the presentation of a large amount of information, and this technique becomes even more difficult when multiple file systems need to be accessed simultaneously (which, arguably, is fundamental to sharing personal files at a tabletop). Second, the collaborative nature of the tabletop interface encourages users to share their files with each other, meaning that the files presented on a tabletop often need to come from multiple collections. Given that hierarchical file systems are inherently single-user oriented, the application of a hierarchical file system to a multi-user tabletop interface is awkward. Finally, hierarchical file systems require users to think about their information in terms of location in the file system, rather than content: this is often inefficient when managing a large file system. Related information may be stored in multiple folders or file systems, and the retrieval of this requires knowledge of exactly where the information is stored in each file system. The automatic merging of this related information, regardless of where it is stored, is important for the natural collaboration supported by tabletop interfaces.
Tabletop and pen-based interaction research has typically been restricted to small collections of information, such as digital photographs [3] or a virtual desktop of files [1]. There has been limited work towards importing files onto tabletops but these require explicit transfer of files between devices [4]. Other systems, such as UbiTable [6] and an approach using tangible drawers [5], require users to copy their files from a laptop computer or personal storage device into a shared region of the tabletop display (and so, the user is conceptually copying files onto the tabletop). Interaction with large hierarchies through association has been explored in the Personal Digital Historian (PDH) project [7]. However, hierarchical presentation tends to quickly fill the tabletop display, which makes navigation of multiple file system hierarchies problematic. Strict hierarchical interaction on tabletops also makes it difficult to present multiple collections that contain related content, but are organised uniquely.
The core aim of our approach is to allow seamless access to multiple, remote file systems with a tabletop interface that exploits the limited display area and restricted human input. A number of drivers have influenced this aim. First, the limited display area (and resolution), and restricted input of a tabletop interface call for approaches to file system interaction that show only relevant information, and only when it is needed.
Second, tabletops are often used as a collaborative tool for sharing information that has been created on another device. Thus the access of personal files in a ubiquitous computing setting—where little configuration is required—would be of great value to improving the integration of tabletop interfaces with everyday computing tasks.
Finally, the automatic merging of related information from multiple sources, such as the file system of each group member collaborating around a tabletop, would improve ease of use and efficiency using a tabletop for collocated collaboration.
OnTop embodies an exploration of a novel associative-search approach to file system interaction on tabletop displays. OnTop, shown in Figure 1, makes efficient use of limited display area by presenting files according to their content (rather than the hierarchical structure imposed upon them), and uses simple gestures to enable navigation of multiple file collections.
OnTop builds upon the Cruiser tabletop platform [2], a multi-user, gestural, collaborative tabletop application written in C++ and OpenGL. It uses the Mimio Capture whiteboard pen system for input.
The OnTop interface, shown in Figure 2, is designed around the notion of a focus file. Once a user selects a focus file, all other related files are displayed. A user may navigate their files by re-selecting focus items. To select a focus item, a user simply dwells on an image representation of the document. A brief 'click' sound gives feedback that it has been selected. A query is sent to each connected machine to find documents related to the newly selected focus item.
Figure 2: The OnTop interface showing files from multiple file systems, presented according to their relevance to the focus item (a text document about the Trojan War). The History Browser (top middle) shows a thumbnail of past focus files, and the icon on the far left of it allows users to return to the start view of the file systems. The Black Hole (bottom right), allows users to hide unwanted files.
To address the limited display area of the tabletop, files are presented in a non-uniform manner, where the initial size of each file's image representation is determined by its relevance to the focus item. Thus, most similar files appear large and prominently, while less relevant files appear small and unobtrusive. Users may then manipulate the displayed images (rotating or resizing them), overriding the initial size based on similarity.
When the interface is first launched (with no focus item selected), a broad starting view shows the first file in each exported directory of each remote file system. Enhancing this starting view by presenting a list of favourite keywords or tags is an area for future work.
A History Browser widget supports 'back' and 'forward' operations for file system navigation. This shows a thumbnail representation of each past focus item (typically the last five focus items, depending on the aspect ratio of the thumbnails). A user dwells on a past focus item to go 'back' to the display of the documents associated with it.
The access of files is performed transparently, and relevant content from each computer is merged to appear as a single information-space. This means that users do not need to know which computer files are stored on, or their place in the file system hierarchy. To make files accessible by the OnTop interface, users run the OnTop File Exporter on their personal computer (typically connected on the same Local Area Network). This lightweight application (written with the Apple Cocoa and Spotlight frameworks) allows full control over the information accessible by tabletops, and which tabletops may connect to the computer.
Given the personal nature of file system organization and content, it would be unreasonable to implement a static relevance calculation mechanism that cannot be customised. Users may customise how their files are determined relevant to a given focus, by selecting, on a scale of 1 to 10 how important certain meta-data attributes are to the organization of their information. For example, if the user deems documents by the same author as closely related, the user increases the weighting on the document author attribute. Currently, 11 meta-data attributes can be adjusted, including the full text content of a document. The attributes given a high weighting by default are keywords, path name (for items in the same directory), display name, and text content. Other attributes such as access time, and where the document was obtained from are also supported.
Our evaluation compared OnTop with a hierarchical alternative. Eight users, of varying backgrounds, were asked to complete five collaborative tasks in pairs. The tasks involved working together to create a set of information about ancient Greek history, and each task involved retrieving files. To simulate that each participant was sharing their own personal files, each participant was assigned their own collection of files that only they were familiar with. One participant had text documents related to a group project, while their partner had photographs and e-mails.
The primary goal was to compare the associative-search approach of OnTop with a hierarchical approach. The sub-goals were based on: success in task completion; number of errors made; error recovery; and affective aspects of user preferences.
A file system interface plug-in for Cruiser, known as the Browser plug-in, was used as the hierarchical alternative. The Browser interface (shown in Figure 3) represents each folder in a local file system hierarchy as a special Browser object on the table, which contains thumbnail representations of the folder's contents. A user drags thumbnails off a Browser object to copy the original file onto the table, at its original size. If a thumbnail represents a sub-directory, the first file in the sub-directory is used as an icon to represent it, and the sub-directory can be opened by dragging it off the parent Browser object (with another Browser object being created).
Figure 3: The Browser interface, which shows thumbnails of the contents of a directory. A sub-directory is represented by yellow triangular corners (top), and files may be dragged off a Browser object to see it at full resolution (bottom and right).
As both the OnTop and Browser interfaces are based on Cruiser, they share common gestures and idiosyncrasies. Similarity between the OnTop and Browser conditions is important to ensure that the only variance between them is the type of interaction involved when navigating a file system.
To gain insight into the impact of file system size for each of the interfaces, two sets of test data were used: one having 59 files, the second having 114 files with more, and deeper, sub-directories. The number of files used in each case was enough to potentially clutter the table and they were a plausible number of files for the scenario given to participants.
The experiment was conducted in a double-crossover fashion, where the ordering of both the conditions used, and file system size varied between experiments to ensure that the evaluations were comparable. Participants were asked to complete the same list of tasks for each interface and file system size.
Questionnaires indicate that participants found OnTop consistently easier to use, and seven of the eight participants considered OnTop to be more efficient. No participants found OnTop harder or slower when the file system size was doubled. In contrast, seven participants had problems with the Browser interface.
Participants embraced the idea of associative-search, employing many different ways to navigate the file systems. For example, some focussed on a text document to find documents matching text of the document. Some selected past focus items from the History Browser, using the document type, such as email, for the search. Some focussed on a document in their partner's file system to access related information in their own.
A key factor to task success and error rates was the degree to which each participant 'pruned' the tabletop of unneeded items (such as irrelevant files, or folders no longer needed). This was a problem for the Browser interface. Participants liked OnTop's automatic pruning, with files irrelevant to a focus item removed from the tabletop.
The Browser interface was found to have a familiar mental model, which made it initially easier to understand for participants. However, after a brief familiarisation period, participants found OnTop more intuitive to use as they could access their files based on content, rather than location. Two participants also stated that they would like to see the inclusion of some features common to hierarchical file browsers, such as being able to quickly navigate to specific files.
We have presented OnTop, a novel associative-search approach to file system interaction, where users navigate multiple file systems by selecting focus files, to retrieve similar files. We reported a qualitative evaluation of OnTop compared against a more conventional file browser approach. The evaluation indicates the promise of the associative-search approach, showing its advantages over hierarchical file system navigation: it makes efficient use of limited display area, and seamlessly merges related content from multiple remote file systems.
Thank you to Associate Professor Judy Kay for her supervision during this work, and Trent Apted for his suggestions and technical input.
[1] Agarawala, A., Balakrishnan, R. Keepin' it real: Pushing the desktop metaphor with physics, piles and the pen. In Proc. CHI 2006, ACM Press (2006), 1283–1292.
[2] Apted, T., Kay, J., Assad, M. Sharing digital media on collaborative tables and displays. In The Spaces In-between: Seamful vs. Seamless Interactions (in conjunction with UbiComp 2005), (2005).
[3] Apted, T., Kay, J., Quigley, A. Tabletop sharing of digital photographs for the elderly. In Proc. CHI 2006, ACM Press (2006), 781–790.
[4] Everitt, K., Shen, C., Ryall, K., Forlines, C. MultiSpace: Enabling electronic document micro-mobility in table-centric, multi-device environments. In Proc. Tabletop 2006, IEEE Computer Society (2006), 27–34.
[5] Hartmann, B., Morris, M. R., Cassanego, A. Reducing clutter on tabletop groupware systems with tangible drawers. In Adjunct Proc. UbiComp 2006, (2006).
[6] Shen, C., Everitt, K., Ryall, K. UbiTable: Impromptu face-to-face collaboration on horizontal interactive surfaces. In Proc. UbiComp 2003, Springer-Verlag (2003), 281–288.
[7] Shen, C., Lesh, N. B, Vernier, F., Forlines, C., Frost, J. Sharing and building digital group histories. In Proc. CSCW 2002, ACM Press (2002), 324–333.
Copyright is held by the author/owner(s).
CHI 2007, April 28–May 3, 2007, San Jose, California, USA
ACM 978-1-59593-642-4/07/0004.