Abstract: Students, researchers and professional analysts lack effective tools to make personal and collective sense of problems while working in distributed teams. Central to this work is the process of sharing—and contesting—interpretations via different forms of argument. How does the “Web 2.0” paradigm challenge us to deliver useful, usable tools for online argumentation? This paper reviews the current state of the art in Web Argumentation, describes key features of the Web 2.0 orientation, and identifies some of the tensions that must be negotiated in bringing these worlds together. It then describes how these design principles are interpreted in Cohere, a web tool for social bookmarking, idea-linking, and argument visualization. p. 2

The Need for Distributed, Collective Sensemaking Tools p. 2

this paper focuses specifically on the challenge of designing engaging, powerful tools for distributed, primarily asynchronous work which, in particular, exploits the strengths of the “Web 2.0 paradigm”. p. 2

2.2. Emergent, Not Predefined, Structure and Semantics p. 3

Argumentation focuses on a particular kind of semantic structure for organising elements. Of central interest, therefore, is the Web 2.0 emphasis away from predefined information organizing schemes, towards self-organised, community indexing (‘tagging’) of elements, resulting in so-called “folksonomies” that can be rendered as tag clouds and other visualizations. Persuading ‘normal people’ (in contrast to skilled information scientists or ontology engineers) to create structured, sometimes high quality, metadata was previously thought impossible, and the success and limits of this approach is now the subject of a new research field that studies collaborative tagging patterns, e.g. [5]. p. 3

Blogs demonstrate one way to negotiate the formality gulf successfully, providing expressive freedom (essentially, traditional prose and graphics), with just enough structure to reap some benefits of hypertext (entries are addressable as URLs, timestamped, tagged, and syndicated as web feeds – see below). The limitation of blogging at present is that like the Web at large, there are no semantics on the links between postings, thus failing to p. 3

provide any support for an analyst who wants to gain an overview of the moves in a debate, or indeed, any kind of inter-post relationship. p. 4

Lit review field p. 4

A significant strand in COMMA research focuses on the design, implementation and evaluation of practical software tools for creating and analysing arguments. Following the entity-relationship modelling paradigm that lends itself so well to software, as well as the work of pioneering argument and evidence mapping theorists such as Wigmore and Toulmin, these tools provide a way to construct arguments as structures comprising semantically linked elements taken from one or more argumentation schemes. The argument structures may be left implicit behind text-centric user interfaces, or rendered explicitly as trees or networks to help the author and reader visualize and edit the p. 4

argument [7]. The intended users of such tools include members of the public engaged in a public consultations and societal debate [8], students or educators in a learning context [9], lawyers [10], and analysts in many other walks of professional life such as public policy [11] and scholarly publishing [12]. Research in this field examines issues including the translation of argumentation theory into computable representations [13], the nature of expert fluency with such tools [14, 15], and empirical studies of the tools’ usage in all of the above domains. p. 5

Overview of existing Web 2.0 tools for argumentation p. 5

Finally and most relevant, there are a number of Web-native applications, designed from the start to support large scale, multi-user construction. Some websites now provide a very simple structure for structuring the two sides of a debate, while others provide a more articulated argumentation language. Beginning with the least structured, we see the emergence of sites such as Debatepedia, which is modelled on Wikipedia, providing a debating resource showing unstructured prose arguments for and against a particular proposal, demarcated in two columns [17] p. 5

CoPe_it! [18] is designed for community deliberation, and provides a way to synchronise views between IBIS graphs (it also integrates with Compendium in this respect), an IBIS outline tree, and a conventional threaded discussion forum. CoPe_it! also provides a mechanism to evaluate the strength of a position, and so represents another interesting development. Its interaction design is at present rather rudimentary compared to Web 2.0 interfaces. It does not have an end-user customisable semantics, interoperability with existing Web data sources, or mechanisms to syndicate content outside the application. p. 5

Parmenides is designed to support web-based policy consultation with the public, and incorporates a formal model of argumentation [19]. It provides a forms-based, questionnaire interface to elicit views from the user, populating an argumentation structure, which it then reasons over to elicit further views. Parmenides enforces a particular argument ontology (it was not designed as a social web application) and does not appear to support any other Web 2.0 characteristics. p. 5

TruthMapping goes much further than this, aiming specifically at tackling some of the limitations of threaded discussion forums, with a clear distinction between unsupported premises, which when supported become claims, and a way to post rebuttals and responses to each of these [22]. p. 5

DebateMapper uses a combined graphical and outline structure to map debates using the IBIS scheme, with contributions tagged as issues, positions and arguments [23]. DebateMapper perhaps illustrates most clearly some the Web 2.0 interaction design principles, but provides no open semantics, or an p. 5

open architecture to enable services on the data. p. 6

The ArgDF system [24] is the first argumentation tool to adopt a Semantic Web architecture based around the W3C standard Resource Description Framework (RDF) for distributed data modelling and interchange. Moreover, ArgDF is probably the first interactive tool to ground its argument representation in the recently proposed Argument Interchange Format (AIF) [25]. This combination of AIF+RDF is a notable advance. However, while proving the conceptual and technical feasibility of a semantic web orientation for argumentation, it does not yet have a user community, and it cannot be regarded as a Web 2.0 application as defined above. p. 6

. Emergent Semantics: Negotiating the Formalization Gulf p. 6

Our approach is to start with relaxed constraints in order to foster engagement with the idea of structuring ideas in general, but provide tools to incrementally add structure as the user recognises the value that it adds in a given context. p. 6

Cohere is, therefore, styled to invite playful testing by people who may not first and foremost be interested in argumentation. Instead, the website invites them to make connections between ideas. This broader framing aims to meet the need of many sensemaking communities to express how ideas or resources are related (whether or not this is argumentative) in a way that goes beyond plain text blog postings, wikis or discussion forums. A typical pair of connected Ideas in Cohere is illustrated in Figure 1. p. 6

In Cohere, users are free to enter any text as an Idea and its detailed description. The examples seeding the database convey implicitly that Idea labels are generally short and succinct. Users are encouraged by the user interface to reuse existing Ideas, with an autocomplete menu dropping down as they type to show matching Ideas already published: as far as possible, we want them to describe the same Idea using the same label. p. 6

Users must, however, constrain their contributions by: creating labelled connections between Ideas (e.g. is an example of) reusing, or creating, a connection from a list of either positive, neutral or negative connections p. 7

Users can optionally: assign roles to Ideas (e.g. Scenario; Problem) add descriptive details (displayed when the Info icon is clicked) assign websites to Ideas (listed when the Idea is selected) p. 7

The provision of mechanisms to enable flexible linking of web resources around what we are calling Ideas is a goal shared by the Topic Maps community [26], whose data model is very close to that of Cohere. Intruigingly, the two were developed entirely independently, yet arrived at the same core data model, which we take to be a form of empirical validation. In the more formally defined, and more wide-ranging Topic Map Data Model, topics (=Cohere Ideas) point to one or more resources (=websites); topics can be linked by associations (=connections), and topics may play one or more roles within a given association (=roles). p. 7

The provision of mechanisms to enable flexible linking of web resources around what we are calling Ideas is a goal shared by the Topic Maps community [26], whose data model is very close to that of Cohere. Intruigingly, the two were developed entirely independently, yet arrived at the same core data model, which we take to be a form of empirical validation. In the more formally defined, and more wide-ranging Topic Map Data Model, topics (=Cohere Ideas) point to one or more resources (=websites); topics can be linked by associations (=connections), and topics may play one or more roles within a given association (=roles). p. 7

A Web 2.0 application called Fuzzzy [27] is based on the Topic Map standard and shares some similarities with Cohere, as does the HyperTopic system [28]; neither, however, provide support for argumentation. p. 7

While not mandating that the user engage in argumentation, the language of deliberation and argument is nonetheless at the heart of Cohere: (i) the roles that Ideas can play in a connection include the IBIS scheme’s Question, Answer, Pro, Con and the user can define new ones (e.g. Datum, Claim, Warrant for Toulmin); (ii) the connection types offered to users are clustered by positive, neutral or negative polarity, with defaults including discourse moves such as proves, is consistent with, challenges, refutes. These default connection types are also leveraged in the predefined visualization filters offered by the Connection Net tab, described later p. 7

the fact that all connections are classed as broadly positive, neutral or negative provides a way to express not only disagreement in the world of discourse, but could signify inhibitory influence (e.g. in biological or ecological systems modelling), or antagonistic p. 7

relationships (e.g. in social networks). p. 8

Cohere now provides a platform for collaborative deliberation and debate mapping over the internet, with primarily asynchronous use in mind to start with. (Real time mapping requires a tool like Compendium which has a user interface optimised for rapid mapping. However, it is our intention to optimise for real time mapping in the longer term, perhaps by adapting Compendium as an applet for Cohere). p. 8

All Ideas other than one’s own have their owner clearly indicated iconically. Clicking this displays the user profile, making it possible to learn more about the person behind the ideas. We are beginning to add metrics to make users aware when they arrive at the site how many active users there are, and what the most recently posted, reused and linked Ideas are. Web feeds in the future will enable users to request notification whenever one of their Ideas is embedded in someone else’s connection, or in someone else’s website (see below). p. 8

Central to O’Reilly’s notion of Web 2.0 is the notion of web data as the platform on which many applications can compute. Cohere exposes and consumes data in a variety of ways: Publishing and importing XML Web feeds Importing XML data from the Compendium offline dialogue and argument mapping tool Embedding pointers to its data in other applications as URLs and HTML ‘snippets’ Exposing data in a variety of standards to engage different communities p. 8

Could be combined with a P2PU course? p. 9

Cohere aims to leverage this by enabling users to import/refresh the Web feed (RSS or Atom) for any bookmarking or blogging site. Entries are converted into Ideas and annotated with the relevant URL, ready for linking. We are in the process of implementing an RSS feed so that users can track new Ideas as they are published. We plan to develop this capability, so that individual Ideas can be subscribed to, with alerts everytime someone connects to or from them. p. 9

Data formats for structured argument data. p. 10

Multiple Import/Export data formats: By the time of the conference, we will have implemented further data formats for importing and exporting Cohere structures. A priority is to provide Argument Interchange Format compatibility, with other candidates being Topic Maps, Conceptual Graphs, and OWL. p. 10

We have started to design views that help render the structures that will result from many users working on common Ideas. This is a long term challenge, but Figure 5 shows the first tool called Connection Net, which uses a self-organising graph layout algorithm that can render all of one’s personal data, or filtered views of the world’s data. In particular, Ideas with a border are used by more than one person, and as shown, right-clicking on it enables the user to view all the owners of that Idea. In this way, just as folksonomies enable disparate users to discover related resources and people, Cohere aims to reveal new connections and users working on the same Idea, or perhaps more interestingly, making similar or contrasting connections. p. 10

Filter buttons in the Connection Net view make use of the connection types, as shown in Figure 6. A number of saved filters are shown, for example, Contrast searches the database from a focal Idea on a specific subset of connection types of a contrasting nature, e.g. challenges, has counterexample, is inconsistent with, refutes. Users can define their own custom searches, and in the future will be able to save them as shown in the example buttons. p. 11

Cohere is implemented on Linux, Apache HTTP server, MySQL database, and PHP. p. 11

We are now in disussion with other COMMA research groups to explore how their argument modelling approaches could be integrated with Cohere. We are moving to a web services architecture [cf. 35] and plan to enable data exchange via the W3C’s RDF and OWL, and the proposed Argument Interchange Format [25]. These developments are designed to evolve Cohere from a closed web application, towards a collaboration platform for structured, social argumentation and deliberation for wider experimentation by both end-users, developers and argumentation researchers. Future technical work will support different argument layouts, more flexible visualizations, group permissions, and the management of template ‘pattern libraries’ (currently managed via the offline Compendium tool). Our pilot usability evaluations are leading to interface changes that are being added at the time of writing, and will be followed by more in depth studies in the lab (cf. [12]), and with end-user communities. p. 12

Finally, our goal with Cohere is to provide an environment for the emergence of social and argument structures. While not currently exposed in the user interface, Cohere has built into its data model the Cognitive Coherence Relations modelling scheme described in the COMMA’08 debate modelling paper by Benn et al. [36], which seeks an integrated approach to modelling social networks and argument networks. p. 12

The Open Sensemaking Communities Project [www.kmi.open.ac.uk/projects/osc] is funded by the William and Flora Hewlett Foundation, as part of the Open University’s OpenLearn Initiative [www.open.ac.uk/openlearn]. p. 12

6. References p. 12