ConversationWorkshop : DesignSpace

HomePage :: Categories :: PageIndex :: RecentChanges :: RecentlyCommented :: Login/Register
Most recent edit on 2005-04-02 11:26:03 by dhcp29311.wireless.unc.edu

Additions:
Contributors: PaulResnick, JohnRiedl <enter your WikiName here>

Computer Participation

One of the interesting possibilities in online conversation is that one or more computers may participate. For instance, (Ludford et al.) examined what happens to conversation if participants are emailed evidence of the 'uniqueness' of their potential conversation -- and discovered that conversation can increase significantly with such manipulations.
Another example of such computer participation is "Conversation Bots". My favorite is that one named after some lawyer who visits conversations and ever time someone says "that's not fair" responds "Do you mean 'that's not efficient'?"
This is a class of interactions that is interesting and important because:
- it is unique (nearly) to online conversation
- it combines expertise with conversation design with expertise in social science
- it has the potential to radically change the dynamic of online conversations

Deletions:
Contributors: PaulResnick, <enter your WikiName here>


Edited on 2005-04-02 04:39:59 by DerekHansen

Additions:

This is a draft version of a document defining the design space of online conversation, major design dimensions, and important contexts. It is intended to provide a general framework that relates many of the papers within this workshop together. It should be thought of as a collaboratively created document and all workshop participants are encouraged to make changes, recommendations, etc. Most of this work will occur after the workshop, but feel free to make changes at any time.

Online Conversation Design Space

"Conversation is food for the soul"

Introduction

1. Online Conversation Defined

Online
Conversation

2. Online Conversation Design Choices

Conversation Genre

Navigation

Grouping Innovations
Sorting Innovations
Linking Innovations
Some systems such as Everything2, utilize "softlinks" that automatically generate links between encyclopedia articles based upon reading behavior (DeMaagd2005). Other systems such as ConversationLens and CoLabspace allow links to be created between any two entities including messages, artifacts (e.g., files, hyperlinks), and people (VonMosch2005; Bobrow2005).
Searching Innovations

Access Control

Summarization / Distillation

There are a number of different criteria used to summarize conversations. Some are socially generated (e.g., quality, insightfulness), while others are automatically generated (e.g., number of messages in a given thread). Smith provides a list of over a dozen criteria available in threaded conversations alone. Some of the less obvious criteria include "Number of authors actively contributing to this thread","Names of recently added recipients", and "Thread days active" (Smith2005).

3. Online Conversation Contexts

Example Contexts

Social Support
Political Deliberation
Other Example Contexts

4. Relating Design Choices to Contexts

Conclusions

Footnotes

Deletions:
"This is a draft version of a document defining the design space of online conversation, major design dimensions, and important contexts. It is intended to provide a general framework that relates many of the papers within this workshop together. It should be thought of as a collaboratively created document and all workshop participants are encouraged to make changes, recommendations, etc. Most of this work will occur after the workshop, but feel free to make changes at any time."

Online Conversation Design Space

Conversation is food for the soul

Introduction

1. Online Conversation Defined

Online
Conversation

2. Online Conversation Design Choices

Conversation Genre
Navigation
Grouping Innovations
Sorting Innovations
Linking Innovations
Some systems such as Everything2, utilize softlinks that automatically generate links between encyclopedia articles based upon reading behavior (DeMaagd2005). Other systems such as ConversationLens and CoLabspace allow links to be created between any two entities including messages, artifacts (e.g., files, hyperlinks), and people (VonMosch2005; Bobrow2005).
Searching Innovations
Access Control
Summarization / Distillation
There are a number of different criteria used to summarize conversations. Some are socially generated (e.g., quality, insightfulness), while others are automatically generated (e.g., number of messages in a given thread). Smith provides a list of over a dozen criteria available in threaded conversations alone. Some of the less obvious criteria include Number of authors actively contributing to this thread,Names of recently added recipients, and Thread days active (Smith2005).

3. Online Conversation Contexts

Example Contexts
Social Support
Political Deliberation
Other Example Contexts

4. Relating Design Choices to Contexts

Conclusions

Footnotes


Oldest known version of this page was edited on 2005-04-02 04:22:25 by DerekHansen []
Page view:
"This is a draft version of a document defining the design space of online conversation, major design dimensions, and important contexts. It is intended to provide a general framework that relates many of the papers within this workshop together. It should be thought of as a collaboratively created document and all workshop participants are encouraged to make changes, recommendations, etc. Most of this work will occur after the workshop, but feel free to make changes at any time."

Editor: DerekHansen
Contributors: PaulResnick, <enter your WikiName here>

Initial Version: April 1, 2005
Latest Version: <enter date after modification>


Online Conversation Design Space


Conversation is food for the soul
Mexican Proverb

Introduction


Conversation is among the most frequent activities of human life. The many innovations that have supported communication throughout the centuries are an endorsement of its importance. More recently, the widespread use of networked computing devices has enabled new forms of conversation to take shape, often in dispersed groups of individuals. At the outset of this era of opportunity it is important to systematically analyze, design, and evaluate systems that support conversation.

This paper provides a framework for characterizing online conversation designs and understanding the choices available to designers of online conversation systems. It also discusses what factors should be considered when making such choices. In other words, it characterizes the design space of online conversation. Ideally, such a characterization will help clarify four things:
1) the scope of online conversation
2) design choices most important to online conversation
3) conversation contexts
4) a mapping of design choices to the contexts

Throughout this paper we will draw upon examples of popular systems used to facilitate online conversation, as well as prototype systems that incorporate novel design features. Papers describing many of these systems were discussed by workshop participants of the CHI 2005 workshop titled “Beyond Threaded Conversation” and are available at www.socialcomputingresearch.net/workshop/

1. Online Conversation Defined


Before discussing the design choices related to online conversation it is important to distinguish between online conversation and other forms of social interaction. To do so, let’s take a closer look at the terms online and conversation.

Online

In this context, the term online indicates that the conversation is facilitated by a network, is computer-mediated, and occurs between people who are not all co-located. The canonical example of an online discussion would be a group of dispersed individuals holding a conversation via a threaded discussion board, newsgroup, or email application. However, it is clear that many other situations are included in our somewhat broad definition of online.

We will not restrict the term online to mean on-the-Internet. In fact, many conversation systems of interest take place on private networks such as corporate Intranets or mobile phone networks. In addition, as communication devices such as mobile phones have become more sophisticated, they have taken on much of the functionality originally performed by traditional computers. Thus, the restriction that the conversation must be computer-mediated should not be interpreted as desktop-mediated. Rather, it should imply that the communication device provides a meaningful representation, storage capability, or computation related to the conversation. Finally, we are not interested in technology that facilitates conversation among entirely co-located groups (e.g., meeting support systems). There may be some individuals of the conversation that are co-located, but the systems we are interested in do not require that all members be physically present.

Conversation

We will define conversation broadly to include any language-based communication between two or more individuals where there is some form of responsiveness. A canonical conversation occurs when a small group of individuals exchange ideas and opinions either verbally or textually, as in a classroom discussion or newsgroup thread. But as we saw with the term online, we are interested in much more than the canonical examples.

Our definition of conversation does not require any specific form of communication. While the vast majority of systems support verbal and textual conversations, other forms of language-based communication such as signing are intentionally not ruled out. Although it has been said that writing is an example of having a conversation with oneself, we have limited our focus to conversations between two or more individuals. In addition, we have required that there is some responsiveness between the participants in the conversation. In other words, later discussion must be influenced in some meaningful way by earlier discussion. For example, a collection of announcements (e.g., an email list that includes job postings) would not be considered a conversation unless there was some discussion about the announcements.

In summary, online conversation is a subset of technology-mediated social interaction that occurs over a network and includes a responsive, language-based dialogue between individuals at least some of whom are not co-located.

2. Online Conversation Design Choices


The purpose of this section is to characterize the key design dimensions of online conversation systems and most, if not all, of the design options available for each of those dimensions. Specifically, four key dimensions are identified including: Conversation Genre, Navigation, Access Control, and Summarization. It is our hope that this characterization will help designers systematically consider design choices and act as an impetus for innovation. The four design dimensions and the design options related to them are described below, primarily from the point of view of a system user.

Conversation Genre

A logical first step in characterizing a conversation system is to describe the shared features and conventions of the basic unit of contribution. We call this collection of characteristics the conversation genre. Some systems such as instant messaging encourage brief, personal, informal contributions to the conversation. On the other extreme, systems that support the exchange of scientific knowledge typically have lengthy, objective, and formal contributions.

One way to identify the conversation genre is to consider the type of contributions that are acceptable both socially and technically. Some important characteristics of the conversation genre include the:

Conversation genres differ significantly based upon the purpose of the conversation, the technology used to support the conversation, and the social norms that develop over time. Some examples of online conversation genres that have become popular in recent years include blog entries (e.g., LiveJournal), online encyclopedia entries (e.g., Wikipedia, Everything2), and product reviews (e.g., Epinions). Current design innovations create new conversation genres in order to support the goals of the system. For example, the REALM forum makes the basic unit of contribution a single argument in a debate (Dave2005). The RadioActive system develops a novel navigational system that facilitates the creation of a new genre where single contributions are similar in spirit to text-based contributions to large-scale public discussions but in audio format (Zinman2005).

Some important questions related to conversation genre include:

Although some of the basic characteristics of conversation genres have been discussed here, a fuller treatment of this topic would be beneficial to designers and social scientists. Such a treatment may provide additional genre characteristics and a more complete description of each of these characteristics. For example, it may provide a taxonomy of different "tones" used in conversation.

Navigation

Online conversation systems differ significantly in how the user is able to navigate through the conversation. Many online conversation systems allow some form of text-based search, browsing through a threaded hierarchy, and/or viewing messages in chronological order. Although these are common, numerous other design options exist including collaborative filtering (see VonMosch2005), visual representations of the conversation (e.g., Donath2005, Smith2005), and automatically generated links between related documents (e.g., DeMaagd2005). This section describes the possible design options that relate to the navigation of a conversation.

Several techniques are used to help a user know how to navigate through a conversation. Three main navigational techniques include:

Each of these techniques facilitates certain types of navigational activities. For example, grouping facilitates browsing of the material that is grouped together. In other words, it is easy to move between items within a single group, but generally more difficult to move across groups. Sorting facilitates selection based upon comparison to other items. Moving between items based upon their similarity or difference become easy. Linking facilitates movement from one message to another even when the messages don't necessarily have an obvious connection. Searching facilitates the retrieval of specific information.

These navigational techniques are often combined together. For example, search results are generally sorted based upon some relevance criteria, while related links may be grouped together. However, it is often useful to consider them separately in order to understand the full set of possibilities.

The navigational techniques just described are based upon a number of factors including:

A traditional threaded conversation uses grouping based upon content (i.e., messages are grouped together into threads), and sorting based upon time (i.e., messages within threads and threads themselves are often sorted in reverse-chronological order). It may also provide a search interface that allows searching based upon content (i.e., does the content include a particular phrase?). Finally, the ubiquitous use of hyperlinks allows messages to use linking based upon content or other criteria. However, many more innovative navigational designs have been explored or are under development. A few of the innovations that emerged from our workshop include:

Grouping Innovations

Several systems group conversation about an artifact to the artifact itself. For example, conversation has been grouped and attached to a particular webpage (Golder2005) and work-related objects such as files or tasks (Bobrow2005; Davies2005). This has the advantages of providing quick access to relevant conversations and the context needed to understand the conversation. Anyone who has held a conversation about how to change a printer cartridge without a printer in the room will recognize the potential benefits of this approach.

Other systems such as LiveJournal, allow for the creation of pages that group xml and rss feeds based upon authors and their social networks (i.e., friends of friends) (Cherny2005).

Sorting Innovations

Two interesting sorting innovations related to visual representations of the conversation. Donath describes a system that visualizes the conversation as a landscape where items are visually sorted based upon various characteristics (e.g., content, ratings, recency) by varying the gradation of certain visual cues such as brightness, color, and size (Donath2005). Another approach that can be thought of as visual sorting was used in weblog, where conversation and artifacts gradually drift and fade to the outer edges of the screen over time (Windram2005).

Linking Innovations

Some systems such as Everything2, utilize softlinks that automatically generate links between encyclopedia articles based upon reading behavior (DeMaagd2005). Other systems such as ConversationLens and CoLabspace allow links to be created between any two entities including messages, artifacts (e.g., files, hyperlinks), and people (VonMosch2005; Bobrow2005).

Searching Innovations

[Need examples]

Access Control

Another important design dimension for online conversation is access control, or who has permission to perform certain actions. The design choices related to conversation access control involve decisions about (a) what the possible conversation-related actions are and (b) who can perform those actions.

Conversation related actions include:

Although there are innumerable ways of determining who can perform these actions, the decision will likely be based upon:

Conversation systems vary considerably in what actions are available and how they assign rights to those actions. Some systems such as Wikipedia and Slashdot have, based upon principle, allow anonymous users significant permissions (e.g., editing, deleting, posting). This, of course, opens the door for deviant behavior that may disrupt the conversation. Sophisticated social and technical methods, such as Slashdot's moderation system, have been developed to address this issue. A complete characterization of the design options related to access control would include descriptions of various options related to addressing such issues. For our purposes, we will consider it enough to point out that there are significant costs and benefits associated with access control decisions that must be considered by designers. Additionally, designers must consider how easy it is to perform each action since some individuals (e.g., newcomers) will be excluded by default if the action is too difficult to perform.

Several workshop papers discussed innovations in this area. For example, the Kontiki system modifies a Wiki platform by allowing users to change the structure of the page as well as the content (e.g., user-defined fields can be created and assigned to individuals) (Wattenberg2005). Another paper discusses the use of "friends lists" as a tool for access control (e.g., posts on LiveJournal can be locked by the author so that only friends may read them) (Cherny2005). Rodriguez2005 introduces the idea of a growing flexible mailing list that keeps track of all of the contributors to a particular conversation within the system and makes it easy to send messages to all (or a subset) of them.

Summarization / Distillation

Some online conversation systems provide social or technical means to facilitate summarization or distillation of the conversation. Summarization refers to the presentation of the substance of the conversation in some condensed form (either textual or visual). Distillation refers to the separation or extraction of the essential (or most pure) elements from the conversation. Among other things, summarizations and distillations of the conversation can provide context to a particular message, help newcomers quickly learn the major themes of the conversation, facilite navigation and discovery within the conversation, and select the very best material from the conversation. A popular example of summarization in action is Wikipedia, where each encyclopedia entry is the summary of a conversation that either occurs on the page itself or via a discussion page linked to the summary article.

The main design decisions in this dimension include:

The process of summarization or distillation may include multiple steps. For example, the Health FACTS system distills a traditional email list conversation by having community members nominate a high-quality message, which is sent back to the author for final editing and classification, and then posted to the FACTS archive (Hansen2005). An alternative process is used in I-DIAG where moderators oversee a brief, but large threaded conversation and then after-the-fact an individual (or small group of individuals) summarizes and distills the conversation with technological aids (Ackerman2005). Yet another approach was used in REALM where the distilled arguments of a conversation are first presented, followed by a process whereby new arguments can be added (Dave2005). It is important to consider that if the distillation or summarization occurs during the conversation it may influence the conversation itself.

Some systems use automated methods to summarize or distill the conversation either through textual analysis of the messages themselves (Zhou2005) or through the use of metadata that describe messages (Smith2005; Erickson2005; Donath2005; Medynskiy2005). Other systems require human involvement, although technical tools are generally provided to aid in the process (Ackerman2005; Hansen2005; Wikipedia). In systems that do require human involvement a designer must determine if the conversation will be summarized or distilled by an individual editor (or small group of editors) (Ackerman2005) or collectively as a group (Hansen2005).

There are a number of different criteria used to summarize conversations. Some are socially generated (e.g., quality, insightfulness), while others are automatically generated (e.g., number of messages in a given thread). Smith provides a list of over a dozen criteria available in threaded conversations alone. Some of the less obvious criteria include Number of authors actively contributing to this thread,Names of recently added recipients, and Thread days active (Smith2005).

Finally, designers must decide how the results of the summarization or distillation are presented. Many novel visualization techniques were presented at our workshop that summarized or distilled the conversation by selectively highlighting certain characteristics of messages and allowing patterns to be recognized that would not have been as apparent otherwise (Donath2005; Erickson2005; Smith2005; Medynskiy2005; Yee2005).

3. Online Conversation Contexts


To fully characterize the design space for online conversation, we must consider the contexts in which online conversations occur. Each context can be thought of as a collection of contextual factors, some of which may be shared with other contexts. The following questions help uncover the underlying factors that make up a particular context:

Example Contexts

Social Support

Many conversations ultimately exist to enable individuals to provide social support to one another. For example, a number of online health communities exist where individuals with a similar medical diagnosis come together to help one another cope with new physical and emotional concerns. The goal of participants is generally to make sense of a new and often frightening situation, learn coping skills, and help others who are in a similar situation.

Many of these social support communities include hundreds of individuals, only some of whom actively participate in the conversation at any given time. These groups may include a medical professional as a facilitator, but this is not always the case. Some members play the conversational role of support provider, giving encouragement to those who play the role of support seeker. However, members may find themselves switching between these roles at different times. Other conversational roles that may be important in such communities include medical expert, lurker, empathizer, or advice provider.

Social support conversations may take place in a publicly accessible forum, a closed network from terminals in hospital rooms, a corporate website, or may even include occasional face-to-face meetings at a local church. The conversations may be linked to a medical advocacy group, an insurance company, or a group of friends and family. Each of these specific contexts is likely unique in some significant way.

The tone of social support conversations is generally friendly, personal, empathetic, and encouraging. Conversation typically follows a pattern of “vent” and respond. Venting includes describing in a very personal manner why you are in need of social support (e.g., “I can’t believe my doctor didn’t tell me sooner…”). Responses are typically empathetic (e.g., That’s terrible. I know how it feels to…”), encouraging (e.g., “Hang in there. I’m sure it will work out…”), show interest (e.g., “Why did he…”), or may include advice. The “venter” generally follows up at a later date with thanks and an update on the situation.

Political Deliberation

[Someone with more experience than me in this area should probably write this one. I think it would be a good one to write out in order to contrast with the last one.]

Other Example Contexts

Although creating a complete taxonomy of online conversation contexts is beyond the scope of this article, it is worth mentioning a few other popular contexts. These include:

4. Relating Design Choices to Contexts


Once design choices and contexts related to online conversation are sufficiently understood, it is possible to analyze what choices make sense in what contexts. Findings from prior research or experience are much more informative when the key design features and the context of use are both described sufficiently. Without such a characterization, it is difficult to know if findings were the result of a specific design feature or the result of a unique contextual factor. Work that is able to test similar designs in multiple contexts or test multiple designs in a single context will be especially helpful to future designers.

Several active research projects discussed at our workshop focus on matching appropriate conversation features with specific contexts. For example, design choices that support various types of workgroups were discussed in Bobrow2005, Davies2005, Hupfer2005, and Wattenberg2005. Several papers discussed online deliberation and debate (Ackerman2005; Dave2005; Yee2005) or scholarly publishing (Pontes2005; David2005). Additional papers looked at contexts related to mobile devices (Zinman2005), supporting ongoing relationships over distance (Windram2005), social support (Hansen2005), and e-learning (Erickson2005).

Conclusions


We are at an exciting and important time in the development of online conversation systems. Many novel ways of conversing online have come about in recent years and countless other opportunities to support various forms of conversation have yet to be discovered. This article outlines the design space of online conversation systems. The design space consists of a description of what is included in it (i.e., a definition of online conversation), what design choices are available within various design dimensions, what contextual factors are important, and ends with a call to systematically relate design choices to contextual factors. We hope that this will be the beginning of a long conversation on improving such systems.


Footnotes
Page History :: 2005-04-02 11:26:03 XML :: Owner: DerekHansen :: Search:
Valid XHTML 1.0 Transitional∞ :: Valid CSS∞ :: Powered by Wikka Wakka Wiki 1.1.5.3∞
Page was generated in 0.1866 seconds