5. Análisis
5.3 Primeras claridades
5.3.7 Otros hallazgos sobre los componentes de las adaptaciones
As mentioned above, argumentation is an important skill. But since many people are poor arguers (Tannen, 1998), learning to argue is to be regarded as a favorable goal. However, due to teachers’ time and availability constraints, teachings on arguing often fall short. Since the last 20 to 25 years, software tools have been built to fill this gap, for a number of different domains including law (see e.g. Aleven and Ashley, 1997), science (see e.g. Linn et al., 1998), conversational argumentation (see e.g. McAlister et al., 2004) or, as would seem natural, the field of computer supported collaborative learning (CSCL) (see e.g. Andriessen, 2006; Baker, 2003; Scheuer et al., 2010; Stegmann et al., 2007).
Scheuer et al. (2010) published a review on computer support for argumentation. They categorize software tools into tools designed for single users, for small groups and for com- munities. I will exemplarily describe one tool for each of these categories. However, one needs to be aware that tools might be used differently than they are intended to. Tools for single users might be used by several people sharing a single computer and individuals can use tools designed for collaboration.
LARGO(Pinkwart et al., 2006, 2007) is a tool designed for single users. It is described as an
intelligent tutoring system that aims at helping law students to develop legal argumentation skills. Students have to transcribe given arguments from the US Supreme Court into graph- ical representations (see Figure 3.3, left). The argumentation model this system is built on is a domain-specific model aligned to the requirements of legal argumentation. Feedback
3 Application Areas
Figure 3.3: Related work on support for remote argumentation. From left to right: LARGO (Pinkwart et al., 2006), BELVEDERE(Suthers et al., 2001 and COLLABORATORIUM(Klein and Iandoli, 2008).
about certain weaknesses (structural weakness, context weakness and content weakness) in arguments is provided in form of self-explanation prompts. This type of feedback is espe- cially helpful in ill-structured domains (such as law) and is aimed at encouraging students to explain their solution.
BELVEDERE (Suthers et al., 2001) is one of the most well-known argumentation support
tools and is designed for small groups. Several evolving versions of this tool exist, mov- ing from a focus on scientific reasoning to more evidential argumentation (Scheuer et al., 2010). BELVEDEREis a graph-based system. Nodes represent the statements (classified into
“data”, “hypothesis” or “unspecified”) and links can be added to display relations (“for” or “against”). An example is displayed in Figure 3.3, middle. BELVEDEREoffers on-demand
feedback for students. Similar to LARGO, it shows feedback in form of suggestions and
questions, and highlights problematic parts in the graphical representation.
An example for a system designed for larger groups is the COLLABORATORIUM(Klein and
Iandoli, 2008). The authors’ main goal was to build a system that combines the advantages of open source/peer production tools with argumentation systems. The resulting system COLLABORATORIUMis web-based and allows people to create argument maps with the aim
to support “collaborative deliberation” (Walton and Krabbe, 1995) (see Figure 3.3, right). In this system, moderators evaluate the entries in terms of correctness and validity.
I will now review some of the specific characteristics of these systems that are important in the scope of this theses. Doing that, I will follow the review of Scheuer et al. (2010), who categorized existing argumentation systems. They identified five different graphical representations of arguments in the literature: (1) linear, (2) threaded, (3) graph-based, (4)containerand (5)matrix. An example forlinearis a chat, which is a kind of computer- mediated communication (CMC) in textual form. However, the problem of sequential inco- herence exists, meaning that contributions such as questions and answers cannot be matched to the statement they refer to. That is the reason why dedicated argumentation support tools rarely or never use this form. Threaded argumentation in contrast solves this problem by enabling message-reply sequences, an option that, for instance, has been made use of in the system HERMES(Karacapilidis and Papadias, 2001). The most common representation form
aregraph-based representations which use nodes and links to display argument components and their relations (e.g., BELVEDERE). Other types are thecontainerrepresentation in which
elements that are interrelated (e.g., claim and evidence) are contained inside of a frame, and the matrixrepresentation that indicates relations inside of the cells. To our knowledge this representation is only used in BELVEDERE, additionally to the graph.
Moreover, Scheuer et al. (2010) discuss theamount of autonomythat the systems provides to users when creating arguments. This is especially relevant in the context of learning be- cause a different amount of autonomy might relate to different learning goals. Scheuer et al. (2010) identify five different levels: (1)free-form arguments, meaning that users can freely choose the content of their argument components (to a predefined topic), (2)argumentation based on background materials, implying that users are provided with background materials, (3) arguments rephrased from a transcript, where users are provided with a transcript that they should convey into a more structured form (such as in LARGO), (4)arguments extracted from a transcript, meaning that users are able to reuse the wording from the transcript and copy and paste it into a new structure and (5) system-provided knowledge units, where the components already exist but the user has the task to relate them to each other.
Finally, it is important to discuss the topic ofontologiesin the context of argumentation sys- tems, as all systems are based on a specific ontology that may differ from the ones of other systems. “An ontology is an explicit specification of a conceptualization” (Gruber, 1993). Traditionally used more in the context of artificial intelligence and knowledge representa- tion, ontologies also take effect in argumentation. In this context, ontologies“describe the components of arguments, together with relations between components and modifiers of the components and relations” (Scheuer et al., 2010). In the context of argumentation system, ontologies are used to make users aware of the available conceptual components (Suthers, 2003). Schwarz and Glassner (2007) differentiate betweeninformal ontologiesandeducated ontologies. Whileinformal ontologiesare based on argumentation found in natural conver- sation, educated ontologies contain definitions and rules and result from a more reflected process. Another differentiation in the scope of ontologies is that between rather general
(e.g., “claim” and “evidence”) anddomain-specific ontologies(e.g., “hypotheticals”, a cate- gory used in law).
Schwarz and Glassner (2007) state that such ontologies that result from an evolutionary and reflected process can be classified as educated ontologies. They are learned in schools and universities in the form of definitions and rules. This contrasts with informal ontologies, which are based on reasoning that typically occurs in natural conversations. While educated ontologies seem especially appropriate for argument modeling, their informal counterpart may be more suited to support structured - and typically less formal - communication. One variation are sentence opener interfaces, which do not explicitly expose categories but which scaffold new contributions through predefined sentence-starting phrases. Typically, these in- terfaces are based on an underlying model of desired communication acts and processes, for instance, dialogue games (McAlister et al. 2004). One general problem that communica- tion ontologies and sentence openers strive to address is to help students to stay on topic by limiting user options.
3 Application Areas
Several of the systems that Scheuer et al. (2010) review are of collaborative nature. All of them are designed for the use on single computers, implying that each collaborator is working from a single computer. Yet in reality, argumentation often occurs in face-to-face situations. I am especially interested in exploring the potential of technological support of such situations. Hereafter, I will discuss first approaches of supporting face-to-face argu- mentation through technology.