Content as Conversation Space: Promoting Positive ... - CiteSeerX

0 downloads 0 Views 35KB Size Report
seen first (e.g. Slashdot [7] and Urban Dictionary [8]). In such systems, commentary from other users must be rated as positive or negative in order to rise or fall in ...
Content as Conversation Space: Promoting Positive Contributions Through Vetted Annotations Scott A. Golder MIT Media Laboratory 20 Ames St., E15-391 [email protected] INTRODUCTION

More and more, web users are beginning to create content as well as consume it, keeping and commenting in weblogs, writing product reviews and engaging in conversation with others [4]. On most web pages that allow them, user-generated comments are placed in a well-defined space allotted by site designers, set apart from the content on which the visitors are commenting. Blogs, product review sites, and conversation sites like Slashdot all work in this way. We envision a system that, instead of setting visitorcreated content apart, superimposes visitors’ annotations directly on top of the content on which the visitor is commenting. For example, a reader of the New York Times’ website might place a comment directly on top of a news article, and that comment may serve as a seed for an entire conversation, made up of superimposed annotations by subsequent visitors. The system we are building allows any document on the web to be annotated in this way. We consider the webpage as an interesting public space or location that is itself the context for further discussion. A threaded conversation space with no messages is barren and lifeless; by contrast, a webpage with no annotations is still full of content that could give rise to a new conversation. Also, by treating the web page as a space for visitors to inhabit, questions of ownership arise. For example, whether or not a visitor should be able to augment someone else’s page is a question open to serious social and legal discussion, as it is not clear whether doing so is more similar to picketing their storefront or defacing it. At its best, annotations by users are meant to inform, educate or entertain subsequent users. We seek to evoke in this work the metaphor of the medieval manuscript, the handsomely handwritten texts before the invention of printing, and the “manuscript culture” that emerged around it. Each copy of a manuscript passed through generations of scholars and other learned people; at each step, the current owner would add commentary about passages of text or about the commentaries left by previous owners. These comments would lie in the margins (“marginalia”) or between lines of text (“interlinear glosses”). In this way, each copy would be unique and valuable in its own right; the holder would be

Judith Donath MIT Media Laboratory 20 Ames St., E15-392 [email protected] guided not only by the wisdom of the text but by those through whose hands the manuscript had passed. Likewise, we envision a system with which visitors’ annotations increase the value of content for future visitors. This is a best-case scenario. At its worst, a system like this, in which anyone could annotate a document, could result in the proliferation of uninteresting content or basic vandalism. Unlike the medieval manuscript, which was rare and restricted to a privileged few, the web is accessible to many, without restriction. Large amounts of low-quality user commentary could obscure the main content of the page without providing any benefit itself. For this reason, we propose a system by which visitors vet those annotations by previous users which they find interesting, a kind of collaborative filtering [e.g. 6]. CONTENT MANAGEMENT

As more and more users contribute, there is an everincreasing amount of content for subsequent users to wade through. In a perfect world, all of this content would be interesting and valuable. However, in reality, much user-generated content is uninteresting or useless to the vast majority of readers. Even if all the commentary by all users were interesting, it would be difficult to wade through it all to find that which is most likely to be most interesting. Systems have chosen to mitigate this problem by forcing more highly-rated content to rise to the top, so that it is seen first (e.g. Slashdot [7] and Urban Dictionary [8]). In such systems, commentary from other users must be rated as positive or negative in order to rise or fall in the ranking. Two potential problems in such a scenario arise. First, users are given an administrative burden, though they may simply wish to read; secondly, a paucity of “votes” leaves much user content unjudged [3]. In this system, we attempt to address these problems with the following rules: 1.

All user comments are shown to the next ten unique visitors. After ten visitors, the comment is destroyed.

2.

If a user chooses to “vet” another’s comment, that comment “lives” for an additional n users beyond the initial ten (we have not decided on the value of n, but

it will likely be between 2 and 5). Vetting anothers’ annotation is done by clicking an icon on the annotation. These two rules have several effects. First, they provide a built-in mechanism for disposing of non-vetted commentary, while allowing higher quality commentary to persist. That is, the system works as intended even with a severe lack of vetting data. Secondly, since the system does not require that user feedback, it takes the administrative burden off of the users. PREVIOUS WORK

The idea of allowing users to annotate any document on the web is not new; some examples include wikis and “graffiti”-oriented sites. A wiki is a collaborative website that anyone can edit [9]. Wikis, however, are less about attaching commentary to a particular place, than adding commentary to a space of one’s own. Having a space of one’s own, even one that can be edited by any visitor at any time, is valuable, but this does not address the problem of allowing anyone to add commentary to any web site at any time, which raises questions of authorship, ownership and authority. Graffiti the Web [1] does allow visitors to add comments to any web page. However, it is demonstrative of what happens when absolutely no filtering takes place. Its “most vandalized site,” msn.com, is so covered in graffiti that the page’s original content is completely illegible, as are most of the graffiti. A now-defunct commercial product called Third Voice seems to be the closest in spirit to what I am attempting to do here. Because the company is out of business, it is difficult to obtain information about how their software worked in practice or to obtain screenshots. However an organization called Just Say No To TV1, as well as several magazine articles, including ones from FirstMonday, Salon and Wired, suggest that Third Voice did not promote thoughtful conversation and commentary as effectively as it may have wanted to.

Figure 1. A Firefox window containing the toolbar for our extension, which contains three buttons, and the homepage of google.com, along with two annotations.

after the tool is used by subjects, we can analyze usage patterns. The system retains information about all annotations, even those that have expired. We are also exploring ways to use this historical data to visualize past interaction within the space. Information about past activity, even if the activity itself is not visible, can guide subsequent users in their exploration of the text [2]. REFERENCES

1. “Graffiti The Web.” www.yeahbutisitart.com/graffiti/ 2. Hill, William C., James D. Hollan, Dave Wroblewski and Tim McCandless (1992) “Edit Wear and Read Wear.” Proceedings of the ACM Conference on Human Factors in Computing Systems (CHI). 3. Lampe, Cliff and Paul Resnick. (2004) “Slash(dot) and Burn: Distributed Moderation in a Large Online Conversation Space.” In Proceedings of Conference on Human Factors in Computing Systems (CHI). 4. Lenhart, Amanda, John Horrigan and Deborah Fallows. (2004) “Content Creation Online.” Pew Internet and American Life Project

IMPLEMENTATION

5. Mozilla Firefox. www.mozilla.org/products/firefox/

We are building an extension for the Firefox web browser [5] that allows users to click on any page to annotate it, as well as displays the annotations of past users. It is developed in the XML User Interface Language (XUL) [10] and Javascript, like all Firefox browser extensions.

6. Resnick, Paul, Neophytos Iacovou, Mitesh Suchak, Peter Bergstrom and John Reidl. (1994) “GroupLens: An Open Architecture for Collaborative Filtering of Netnews.” Proceedings of the ACM Conference on Computer Supported Cooperative Work (CSCW).

When users navigate to new pages, annotate pages, or vet others’ annotations, CGI scripts are called that extract existing annotations, record new ones and record vetting, respectively. Each user is tracked via a unique ID, so that

7. Slashdot. www.slashdot.org/ 8. Urban Dictionary. www.urbandictionary.com/ 9. “Wiki.” Wikipedia. en.wikipedia.org/wiki/Wiki 10.XML User Interface Language (XUL) Project. http://www.mozilla.org/projects/xul

1

http://worldzone.net/internet/pixelsnttv/index.html

2