Document Centered Discussion

Pattern number within this pattern set: 
Todd Davies
Stanford University
Benjamin Newman
Stanford University
Brendan O'Connor
Stanford University

Supporting group interaction around a shared document is challenging for designers of two-dimensional interfaces and asynchronous, text-based groupware. The need to deliberate (collaborate, make decisions, or make comments) around documents appears to be one of the main reasons that groups that could otherwise interact virtually and asynchronously using the Internet choose to meet synchronously, either in person or online, often in a richer environment than text only (e.g., including audio, video, or a three-dimensional environment). When some or all stakeholders are unable to participate in synchronous meetings, distributed asynchronous interaction offers many advantages to groups deliberating about documents: more time for reflection, revision, and information seeking (cf. Holland and Stornetta 1992); the ability to accommodate people's conflicting schedules; flexible interaction modes through conversion of text to and from speech (e.g., for disabled or less literate users); the easier access, storage, and search afforded by digital archives; and the empowerment of those who are at a disadvantage when participation involves speaking in a live group (Price and Cappela 2002). But mapping in-person meetings onto an asynchronous interaction through distributed two-dimensional text displays entails several types of lost richness, including nonverbal grounding cues (Clark and Brennan 1991), spatial depth, the natural use of separate perceptual modalities for document (visual) and discussion (auditory), and the use of a shared temporal progression to guide attention.


In our usage, a document can be in any format, including images, audio, and video, but our primary focus here is on digital documents in which most of the information is in text. Discussion that takes place around a document consists of comments that pertain to either the document as a whole or some part of it. The document may be fixed or evolving as the discussion around it proceeds, but the document assumes an elevated status over the comments made about it because, for example, it has been chosen for careful discussion, its final version will have governing consequences, or it will represent the outcome of a collaboration. In the latter cases, the group must somehow reach a decision relating to the document (e.g., whether to adopt it). This pattern focuses on interfaces for visually abled users. Adapting the analysis presented here for visually impaired users might be possible, but our feeling is that that will require quite a different approach, one we hope to investigate in the future.


The projection of a three-dimensional, multimodality, co-located, synchronous deliberation experience (i.e., communicating face-to-face in a non-virtual place) onto a two-dimensional, primarily visual, distributed asynchronous interaction requires essential aspects of face-to-face deliberation to be remapped onto a screen interface. The needed mappings can be judged according to two broad goals:

  • Visible relationships. Relationships between comments and the texts they reference, between different comments, and between group members and the document and discussion should be as visible as possible.
  • Distinguishable boundaries. Separations between contextually related and unrelated text and comments and between individual authors of documents and comments should be as distinguishable as possible.

We first consider visible relationships. Exhibiting relationships between the components of document-centered deliberation (document, comments, and participants) implies a number of refinements of this goal. First, the document text that is the target of deliberation should be covisible (displayed simultaneously) with comments around it, and the identities of comment authors and document text, when relevant, should also be covisible with their output. Second, the referencing relationship between a comment and its target text should be visible, that is, the interface should incorporate ostensive pointing (meaning that a pointing relationship is displayed on the screen rather than being enacted through a peripheral device) and in-text placement of comments. Third, response relationships between comments should be visible through threading. And fourth, the reactions of deliberation participants should be visible through polling and decision features.

The other goal is distinguishable boundaries. Visible relationships can be inadequate, as anyone would know who has used a map with ambiguous place labelings. The interface should also mark boundaries between text that is and is not the reference target of a comment, for example, though text highlighting. Text authored by different people at different times should be distinguishable through textual boundaries. The topic of a text should be able to be viewed separately from its main body through headers. And obsolete comments (those made on a previous version) should be recognizable through pertinence markers which indicate which versions a comment pertains to, as well as those to which it does not apply, for example, because it has been addressed in the revision process.

The Deme environment for online deliberation is a tool for document-centered discussion, polling, and decision making that incorporates all of the elements derived above in a dynamically updating (no-page-reload) interface. The introductory image above shows the most recent design of the meeting area viewer in Deme. The shaded-in header of a comment in the discussion view pane on the right points to a shaded-in comment reference in the text of a document shown in the item view pane on the left. Deme provides covisibility between document and comments through an optional split-screen view. In-text comment references are transiently pointed to (the dotted-line arrow goes away as soon as the user scrolls) when clicked on, and comments are displayed in the context of hierarchical threads. Members can vote on documents under a variety of decision rules. Boundaries are provided through highlighting, text boundaries, headers, and a versioning system that remembers when comments become obsolete and marks them as such. The design takes advantage of no-page-reload web server calls to provide dynamic relationship visibility and boundary distinguishability.


Applying the principles of relationship visibility and boundary distinguishability in an integrated way puts online deliberators at less of a disadvantage relative to their face-to-face counterparts.

Verbiage for pattern card: 

Groups need to deliberate around documents. Distributed asynchronous interaction allows more time for reflection, revision, and information seeking; the ability to accommodate conflicting schedules; and easier access, storage, and search of digital archives. Groups can discus the document as a whole or parts of it. The document may be fixed or evolving, but the document should remain central.

Pattern status: 
Aaron Tam
Stanford University