ArticlePDF Available

The metafora tool: Supporting learning to learn together


Abstract and Figures

Collaboration in complex learning scenarios does not succeed automatically without structuring the learning process. The Metafora project ( is designing a pedagogy and a platform of web-based software to support learning to learn together (L2L2) in the context of math and science. The platform serves both as a toolbox of various learning tools and as a communication architecture to support cross-tool interoperability. The central tool in the Metafora system is a web-based application offering a visual language for planning, enacting and reflecting on learning activities. In the demonstration we will present our pedagogical approach for supporting L2L2 activities and the platform developed on the basis of this understanding. In particular we will demonstrate how the platform can be integrated in successive activities.
Content may be subject to copyright.
The Metafora Tool: Supporting Learning to Learn Together
Reuma De-Groot, The Hebrew University of Jerusalem, 91905 Jerusalem, Israel,
Toby Dragon, Center for E-Learning Technology, Saarland University, C5.4, 66123 Saarbrücken, Germany,
Manolis Mavrikis, London Knowledge Lab, Institute of Education, University of London, London, UK,
Andreas Harrer and Kerstin Pfahler, Catholic University Eichstätt-Ingolstadt, 85072 Eichstätt, Germany,
{andreas.harrer, kerstin.pfahler}
Bruce M. McLaren, Human Computer Interaction Institute, Carnegie Mellon University, Pittsburgh, PA, USA,
Rupert Wegerif, Graduate School of Education, University of Exeter, Exeter EX1 2LU, UK,
Chronis Kynigos, School of Philosophy, Ed. Tech. Lab, Athens, Greece
Baruch Schwarz, The Hebrew University of Jerusalem, 91905 Jerusalem, Israel,
Abstract: Collaboration in complex learning scenarios does not succeed automatically
without structuring the learning process. The Metafora project (http://www.metfora- is designing a pedagogy and a platform of web-based software to support learning
to learn together (L2L2) in the context of math and science. The platform serves both as a
toolbox of various learning tools and as a communication architecture to support cross-tool
interoperability. The central tool in the Metafora system is a web-based application offering a
visual language for planning, enacting and reflecting on learning activities. In the
demonstration we will present our pedagogical approach for supporting L2L2 activities and
the platform developed on the basis of this understanding. In particular we will demonstrate
how the platform can be integrated in successive activities.
Learning to Learn Together and the Metafora tool
Most knowledge creation is conducted by teams and not by individuals. In addition, learning mediated by the
Internet is often focused on learning together with others. It is therefore important that we teach and support the
complex competence of learning to learn together (L2L2). While there has been some research on learning how
to learn (L2L, e.g., Claxton, 2004; Fredriksson & Hoskins, 2007; Higgins et al., 2006), there has been little
research on L2L2. Learning how to learn together implies that all the group members are able to coordinate,
regulate and plan the learning task by balancing issues of individual ability, motivation and expectations
through constant dialogue. The process of L2L2 can be described and studied by analyzing the groups’
collaborative learning activities and behaviors as a set of sub-skills: distributed leadership, mutual engagement
for fulfilling collaborative tasks, a dialogue where students can discuss their ideas and create new ones, and peer
group assessment, where members give and accept feedback from each other, routinely reflecting on their work.
The Metafora project (, funded by the EC, is designing a pedagogy and
a platform of web-based software to support L2L2 in the context of math and science. A key technical and
pedagogical innovation of the project is to support L2L2 within a group of learners. We present our platform
(see Fig. 1), which serves both as a toolbox of various learning tools and as communication architecture to
support cross-tool interoperability. The toolbox facet of the system provides a graphical container framework in
which the diverse learning tools can be launched and used. Basic functionalities that is globally available are
user management (login/logout and group membership for both local groups of students sitting at one computer
as well as remote, collaborative groups), a chat system to discuss and organize work between group members,
and a help request function that is present across the entire platform. Below we describe in some more detail
certain components and features of the Metafora system.
The planning/reflection tool
The planning/reflection tool offers a visual language that enables students to create and map representations of
their work for planning, enacting and reflecting on Metafora learning activities (see Fig. 1). The main feature of
this tool is the use of cards and connectors to present a plan for future work or to create a diagram of work
completed for reflection. The cards contain visual symbols and titles, as well as space to insert free text (see Fig.
1) The symbols and the titles represent different stages and processes related to inquiry learning (e.g.,
experimentation, building models, making hypotheses), attitudes taken towards the group work (e.g., being
critical, being open) and cards that allow access to different resources within the Metafora tool box (e.g.
LASAD, microworlds, etc). The connectors represent relational heuristics (“is next”, “needed for” and “related
to”) to explicate how the various cards are related in the given plan.
Although it is built as a stand-alone web application, it is most effective as an embedded tool within the
Metafora platform, acting as an entry gate and pivot to the other tools. Students can create and modify plans for
facing various challenges in math or science. The students can also invoke other tools, including microworlds
and discussion tools, and utilize them through specialized resource cards that are part of the visual language.
With the planning tool, students describe how they will tackle their current challenge using the visual language
as a guide and then move together through the various planned stages, enacting activities and noting when
activities are started and completed. Thus, the plan is also a visual representation of the groupsachievements
and current status.
Microworlds integrated in the Metafora system
Metafora provides five microworlds that are fully integrated in the Metafora platform. These microworlds serve
as an arena for inquiry and constructionist work. (1) eXpresser: a microworld designed to support students in
generalizing rules based on the structure of figural patterns of square tiles. In eXpresser, students construct
animated models comprising patterns of repeated building blocks of tiles. (2) The “3d Math” Authoring Tool: a
3d programmable environment inside which users may graphically represent and manipulate 3d objects that
they either find ready-made in an embedded library or construct themselves when using Logo procedures and
commands. (3) The Physt 3d Authoring Tool: a 3d programmable environment that allows teachers (i.e. “the
Pedagogical Designers”) to create 3d game-like microworlds (e.g. the 3D Juggler microworld), for simulating
phenomena defined by Newtonian Laws. (4) Sus-City: a game template for non-technical users (teachers and
students) to construct and play their own “Sustainable City” games. The game design is based upon two types of
user intervention: a) adding content on the template, i.e., the city terrain, city sites and site properties and b)
defining the initial set of values for the player and the threshold values which indicate violation of the system
and end of the game. (5) PiKi: a microworld that addresses kinematics through a serious game with a pirate-
based theme. Other microworlds like Geogebra can be integrated with the Metafora system in a less integrated
way, but still allow productive collaborative inquiry based activities.
Discussion tools and referable objects
Metafora provides discussion tools to allow general communication and collaboration, but also aims specifically
to support the L2L2 process by allowing discussion and argumentation spaces to integrate artifacts created in
other tools. Two discussion tools serve different purposes. First, the chat tool offers a quick and ever-present
space for students to gain each other’s attention and share informal thoughts in situ. Second, LASAD (Loll et
al., 2012) offers a structured approach to discussion through argumentation graphs (see Fig. 2), which have been
shown to improve discussion and argumentation skills (Scheuer et al., 2010). Both the chat functionality and the
LASAD system are customized to display and offer links to referable objects that reside within other tools.
These referable objects are artifacts shared from other tools that can be viewed (text or thumbnail images) as
components of the discussion, but can also be accessed in the context of the original tool through return links
(see an example in Fig. 2). This need emerged from early experimentation with the system and was supported
by previous related research (e.g. Stahl, 2006).
Figure 1. Screenshot of the Metafora platform with several learning tools opened (see tabs on the upper border).
The planning tool is shown in the center (started activities are marked in yellow, finished activities in green; the
arrows are connectors symbolizing the relations between the visual cards).
Figure 2. A discussion map in LASAD with embedded referable objects from a microworld (PiKi)
By using referable objects, students can include planning cards and/or microworld objects in their discussion
without the need of anaphoric or deictic language. These direct references allow continuous dialog that is
explicitly linked with and contextualized by the students work in other tools. In this way, referable objects
allow students to more easily and naturally engage in L2L2 activities such as offering help to one another, and
reflecting on ideas and hypotheses in an ongoing process of negotiation of new meaning for created artifacts.
Analysis and visualization
Each software tool stands as an independent learning tool and can offer its own automated analysis of student
work through individual analysis components. Analysis ranges from low-level activity indicators (such as
indicating the creation or modification of artifacts) to high-level analyses (such as identification of whether a
student is struggling). The intelligent components of the tools that create these various analyses report them to a
centralized analysis communication channel for the entire Metafora platform. A cross-tool analysis agent then
monitors this channel and offers higher-level analyses of student work. Defining and creating these high-level
analyses is an ongoing effort based on prototypes and Wizard-of-Oz experimentation. The theory behind this
work and first implementation steps is described in more detail in (Dragon et al., 2011). This analysis
information is used to offer both direct feedback to students (through a notification system) and useful summary
information to both students and teachers (through visualization tools that filter and aggregate information). The
specifics of the information that should be displayed, to whom and when, are currently under investigation.
L2L2 in Metafora and its significance for CSCL research and practice
The Metafora system is conceptualized and implemented as a full-fledged web-based application, with the
platform and diverse learning tools running in a web browser. Thus, it is easily accessible and built to integrate
third-party web tools to support complex L2L2 scenarios. Our current primary set of learning tools described in
the earlier section is specifically designed with the L2L2 principles in mind and with a high degree of semantic
interoperability, i.e., seamless transition between the different tools via referable objects and the potential for
cross-tool analyses. The growing maturity of the system has already been demonstrated with extensive
experimentation of the pedagogical scenarios in classrooms (see, e.g, Metafora public deliverable1). This
practical application and the empowerment of teachers will continue. In the future, the system will also be used
as a research platform for a variety of complex learning scenarios. The ongoing data analysis of experimental
data will provide insights into the nature of L2L2 and how the analytic system of Metafora can be enhanced to
support students. The automated analysis of Metafora extends the automated work of collaborative learning
scenarios that has been developed in earlier work (see, e.g., the work on ARGUNAUT (McLaren, Scheuer, &
Mikšátko, 2010) and that of Rosé and colleagues (Rosé et al, 2008)). Metafora pushes the envelope on prior
automated analyses by providing analyses across a variety of learning tools (Dragon et al., in press).
We argue that Metafora’s unique contribution to the CSCL agenda is its ability to afford and explicitly
represent the group work as a collaborative artifact (in the planning/reflection tool) and as such expose it as a
subject of group discussion. Planning and reflecting activities make students focus on the meta-level task of
understanding how their group succeeds or struggles in planning and enacting their work. The planning tool -
used as a gate to all the tools integrated in the Metafora toolbox - plays a crucial role in this process and as such
is the most significant tool developed in the project. This tool allows students to elevate their thoughts and
discussion beyond the content of their task, and motivates reflection on how they work together and how, as a
group, they can succeed in their learning objectives. We recognize this higher-level student effort as Learning
To Learn Together (L2L2), a collaborative learning process involving several key competencies that can be
practiced and recognized within the Metafora platform. Defined earlier in this document, the L2L2 learning
performances can be characterized by a set of learning behaviors such as willingness to share, give feedback and
reflect, distribute tasks and roles. To this end, the Metafora system allows a smooth interplay between dual
interaction spaces (e.g. Mühlpfordt & Stahl, 2007) of the microworlds, the planning/reflection tool and the
LASAD discussion tool. Referable objects allow students to make cross-tool reference to objects, and shared
resources represented as artifacts allow students to seamlessly move between planning, enacting, and reflecting.
These dual interaction spaces serve as an appropriate arena for students’ sharing artifacts and ideas along their
collaborative work and as such support their L2L2 behavior.
Preliminary insights from our studies (reported in project deliverables1) show that the students tend to
use the planning tool for reflecting upon their work and concretizing their next steps accordingly. Discourse
analysis of the groups’ oral discussions (while working with the planning tool) reveal a clear picture of
collaborative meaning-making processes over the scientific concepts symbolized in the cards. Moreover,
discussions around elements of the visual language and their possible meaning in the context of the groups’
work supported processes of L2L2 such as role and task re-allocation, mutual engagement and reflection.
Structure of the demonstration
The demonstration will be divided into three parts. In the first part (15 minutes) we will present the pedagogical
concepts behind our work and introduce our pedagogical approach to L2L2. Next we will devote 10 minutes to
introduce the Metafora tool. In the third part (30 minutes) we will invite the participants to work in groups and
solve a challenge in science (a problem related to kinematics) with the use of the planning tool and a
microworld in Metafora. Participants will be asked to work collaboratively on their plan and devise a solution
with the use of the microworld. In the final part of the demonstration (15 minutes) we will conduct a reflective
discussion on the affordances of the Metafora tool and its aim to support L2L2.
Claxton, G. (2004) Teaching children to learn: beyond flat-packs and fine words Burning Issues in Primary
Education No. 11 Birmingham: National Primary Trust.
Dragon, T., McLaren, B.M., Mavrikis, M., Geraniou, E. (2011). Scaffolding Collaborative Learning
Opportunities: Integrating Microworld Use and Argumentation. In Ardissono, ed.: Advances
in User Modeling: UMAP 2011 Workshops (pp. 18-30), Girona, Spain, July 11-15, Revised Selected
Papers. Volume 7138 of Lecture Notes in Computer Science., Girona, Spain.
Dragon, T., Mavrikis, M. McLaren, B.M., Harrer, A., Kynigos, C., Wegerif, R., & Yang, Y. (in press).
Metafora: A web-based platform for learning to learn together in science and mathematics. To be
published in a special edition of IEEE Transactions on Learning Technologies.
Fredriksson, U. and Hoskins, B. (2007) The development of learning to learn in a European context. Curriculum
Journal Vol. 18, No.2, pp. 127 - 134.
Higgins, S., Wall, K., Baumfield, V., Hall, E., Leat, D. and Woolner, P. with Clark, J., Edwards, G., Falzon, C.,
Jones, H., Lofthouse, R., Miller, J., Moseley, D., McCaughey, C., and Mroz, M. (2006) Learning to
Learn in Schools Phase 3 Evaluation: Year Two Report . London: Campaign for Learning. Available
Loll, F., Pinkwart, N., Scheuer, O., McLaren, B.M. (2012). In: How Tough Should It Be? Simplifying the
Development of Argumentation Systems using a Configurable Platform. Bentham Science Publishers.
McLaren, B.M., Scheuer, O., & Mikšátko, J. (2010). Supporting collaborative learning and e-Discussions using
artificial intelligence techniques. Intern. Journal of Art. Intelligence in Education (IJAIED) 20(1),1-46.
Mühlpfordt, M., & Stahl, G. (2007). The integration of synchronous communication across dual interaction
spaces. Proceedings of the 8th International conference on Computer supported collaborative learning.
Rosé, C., Wang, Y.-C., Cui, Y., Arguello, J., Stegmann, K., Weinberger, A., & Fischer, F. (2008). Analyzing
collaborative learning processes automatically: Exploiting the advances of computational linguistics in
CSCL. International Journal of Computer-Supported Collaborative Learning, 3(3), 237-271.
Scheuer, O., Loll, F., Pinkwart, N., McLaren, B. (2010). Computer-supported argumentation: A review
of the state of the art. Intern. Journal of Computer-Supported Collaborative Learning 5(1) 43102.
Stahl, G. (2006). Group Cognition: Computer Support for Building Collaborative Knowledge (Acting
with Technology). illustrated edition. The MIT Press.
The Metafora project is co-funded by the European Union under the Information and Communication
Technologies (ICT) theme of the 7th Framework Programme for R&D (FP7), Contract No. 257872. We thank
our colleagues in the project for the fruitful discussions and cooperation to support L2L2.
1 Metafora public deliverable D3.1 - The scenarios, the microworlds and a descr. of the research design (2012) . Via http://www.metafora- &id=33&Itemid=50
... The EU-funded Metafora project (ICT-257872) enabled the development of a system and of an educational environment aimed at promoting L2L2 (de Groot et al. 2013). The Metafora system comprises (1) a visual tool for planning and reflecting on group work, (2) microworlds for experiencing phenomena and exploring problem spaces, (3) a space for argumentation, and (4) a module for observing group work and possibly intervening by sending messages. ...
Learning to learn together (L2L2) is a complex competence requiring that all the group members are able to coordinate, regulate and plan the learning task by balancing issues of individual ability, motivation and expectations through constant dialogue. In this paper we report on a project to define the complex competence of L2L2 and to support it with a set of web-based tools and associated pedagogy, the Metafora Project. The system we develop embodies our theory of L2L2 and the results of our design-based research suggest that this system can succeed in making key elements of L2L2 explicit in the talk and actions of groups of learners.
... The EU-funded Metafora project (ICT-257872) enabled the development of a system and of an educational environment aimed at promoting L2L2 (de Groot et al., 2013). The Metafora system comprises (1) a visual tool for planning and reflecting on group work, (2) microworlds for experiencing phenomena and exploring problem spaces, (3) a space for argumentation, and (4) a module for observing group work and possibly intervening by sending messages. ...
Full-text available
In this paper, we identify Learning to Learn Together (L2L2) as a new and important educational goal. Our view of L2L2 is a substantial extension of Learning to Learn (L2L): L2L2 consists of learning to collaborate to successfully face L2L challenges. It is inseparable from L2L, as it emerges when individuals face problems that are too difficult for them. The togetherness becomes a necessity then. We describe the first cycle of a design-based research study aimed at promoting L2L2. We rely on previous research to identify collective reflection, mutual engagement and peer assessment as possible directions for desirable L2L2 practices. We describe a CSCL tool: the Metafora system that we designed to provide affordances for L2L2. Through three cases in which Metafora was used in classrooms, we describe the practices and mini-culture that actually developed. In all contexts, groups of students engaged either in mathematical problem solving or in scientific inquiry and argumentation. These cases show that L2L2 is a tangible educational goal, and that it was partially attained. We show how the experiments we undertook refined our view of L2L2 and may help in improving further educational practice.
Conference Paper
The design of CSCL tools has long been a subject of research in the learning sciences community. To this end, theories like dialogic learning and argumentation led to new understandings that see the social context as means to organize learning through collaborative meaning-making. The Metafora and the Collaso learning environments aimed at developing technological tools to afford the smooth integration of inquiry and argumentation to foster learning to learn, L2L, and learning to learn together, L2L2 (in Metafora) and (collaborative) inquiry-based learning (in Collaso). Using three examples of learning activities in one of these environments revealed preliminary understandings and insights on the way tools’ design influences group learning in the 21st century. Our preliminary observations show that the two learning environments achieved similar goals despite their different designs. This may shed light on the relevance of educational design for technological environments, suggesting also a closer look at classroom’s enculturation and teachers’ work when using CSCL co-located in the classroom in order to assess such design work (This paper is partially based on work done by the author and others which was published in Schwarz et al. 2015. The author wish to thank to MinCet Team especially Aviran Mor and Yogev Levy for their dedicated work on Collaso. Many thanks also to Dr. Gil Amit for opening the doors of the Ashkelon College to run our pilots).
Full-text available
Teaching to argue is challenging. Classic face-to-face approaches do not scale up for large groups due to resource limitations (teacher time), but have shown to be effective. As a consequence, there have been attempts to convey argumentation skills via educational software. Even though some of these systems have shown their suitability in their original domains of application, the systems typically do not generalize – there has been little carry over to other domains. This chapter reviews existing approaches, their technological strengths and weaknesses, and proposes a generic architecture to overcome the latter. Based on this architecture, the LASAD (Learning to Argue – Generalized Support Across Domains) framework has been developed. The goal of this framework is to simplify the development of argumentation systems based on some well-defined configurations. In this chapter, we describe the flexibility of the LASAD framework and demonstrate how it can be configured to emulate the existing argumentation systems Belvedere and LARGO.
Full-text available
This paper presents Metafora, both a platform for integrated tools as well as an emerging pedagogy for supporting Learning to Learn Together in science and mathematics education. Our goal is to design technology that brings education to a higher level; a level where students not only learn subject matter, but also gain a set of critical skills needed to engage in and self-regulate collaborative learning experiences in science and math education. To achieve this goal, we need to understand how educational technology can bring students' attention to, and promote these higher level skills. We first discuss the core skills that students need as they learn to learn together. We then present a platform and pedagogy to support the acquisition of the critical skills. Finally, we present an example use of our system based on results from pilot studies. This example demonstrates interaction with the platform to highlight potential benefits and limitations of our approach to promoting the associated skills.
Conference Paper
Full-text available
This paper presents our research efforts to support students' collaborative process when learning mathematics and science as they interact with microworlds and engage in discussions and structured arguments. The system provides students with an environment to explore challenging problems and encourages them to collaborate using a discussion tool to argue and share their rationales and insights using specific examples from microworlds. The challenge of providing useful analysis in such a situation is to recognize, across the learning environment as a whole (both microworld and discussion tool), situations where students need support, and then to make the learners aware of these situations in a productive manner. We present a use case that demonstrates how students work within the system and how we envision the system will provide support. We conclude that the analysis and support that we propose has the potential to enhance the benefits of a combined system and offer more support than a system focused on the individual tools separately.
Conference Paper
Full-text available
Dual interaction spaces--that combine text chat with a shared graphical work area--have been developed in recent years as CSCL applications to support the synchronous construction and discussion of shared artifacts by distributed small groups of students. However, the simple juxtaposition of the two spaces raises numerous issues for users: How can objects in the shared workspace be referenced from within the chat? How can users track and comprehend all the various simultaneous activities? How can participants coordinate their multifaceted actions? We present three steps toward integration of activities across separate interaction spaces: support for deictic references, implementation of a history feature and display of social awareness information.
Full-text available
Argumentation is an important skill to learn. It is valuable not only in many professional contexts, such as the law, science, politics, and business, but also in everyday life. However, not many people are good arguers. In response to this, researchers and practitioners over the past 15–20 years have developed software tools both to support and teach argumentation. Some of these tools are used in individual fashion, to present students with the “rules” of argumentation in a particular domain and give them an opportunity to practice, while other tools are used in collaborative fashion, to facilitate communication and argumentation between multiple, and perhaps distant, participants. In this paper, we review the extensive literature on argumentation systems, both individual and collaborative, and both supportive and educational, with an eye toward particular aspects of the past work. More specifically, we review the types of argument representations that have been used, the various types of interaction design and ontologies that have been employed, and the system architecture issues that have been addressed. In addition, we discuss intelligent and automated features that have been imbued in past systems, such as automatically analyzing the quality of arguments and providing intelligent feedback to support and/or tutor argumentation. We also discuss a variety of empirical studies that have been done with argumentation systems, including, among other aspects, studies that have evaluated the effect of argument diagrams (e.g., textual versus graphical), different representations, and adaptive feedback on learning argumentation. Finally, we conclude by summarizing the “lessons learned” from this large and impressive body of work, particularly focusing on lessons for the CSCL research community and its ongoing efforts to develop computer-mediated collaborative argumentation systems.
Full-text available
In this article we describe the emerging area of text classification research focused on the problem of collaborative learning process analysis both from a broad perspective and more specifically in terms of a publicly available tool set called TagHelper tools. Analyzing the variety of pedagogically valuable facets of learners’ interactions is a time consuming and effortful process. Improving automated analyses of such highly valued processes of collaborative learning by adapting and applying recent text classification technologies would make it a less arduous task to obtain insights from corpus data. This endeavor also holds the potential for enabling substantially improved on-line instruction both by providing teachers and facilitators with reports about the groups they are moderating and by triggering context sensitive collaborative learning support on an as-needed basis. In this article, we report on an interdisciplinary research project, which has been investigating the effectiveness of applying text classification technology to a large CSCL corpus that has been analyzed by human coders using a theory-based multi-dimensional coding scheme. We report promising results and include an in-depth discussion of important issues such as reliability, validity, and efficiency that should be considered when deciding on the appropriateness of adopting a new technology such as TagHelper tools. One major technical contribution of this work is a demonstration that an important piece of the work towards making text classification technology effective for this purpose is designing and building linguistic pattern detectors, otherwise known as features, that can be extracted reliably from texts and that have high predictive power for the categories of discourse actions that the CSCL community is interested in.
Full-text available
An emerging trend in classrooms is the use of networked visual argumentation tools that allow students to discuss, debate, and argue with one another in a synchronous fashion about topics presented by a teacher. These tools are aimed at teaching students how to discuss and argue, important skills not often taught in traditional classrooms. But how do teachers support students during these e-discussions, which happen at a rapid pace, with possibly many groups of students working simultaneously? Our approach is to pinpoint and summarize important aspects of the discussions (e.g., Are students staying on topic? Are students making reasoned claims and arguments that respond to the claims and arguments of their peers?) and alert the teachers who are moderating the discussions. The key research question raised in this work: Is it possible to automate the identification of salient contributions and patterns in student e-discussions? We present the systematic approach we have taken, based on artificial intelligence (AI) techniques and empirical evaluation, to grapple with this question. Our approach started with the generation of machine-learned classifiers of individual e-discussion contributions, moved to the creation of machine-learned classifiers of pairs of contributions, and, finally, led to the development of a novel AI-based graph-matching algorithm that classifies arbitrarily sized clusters of contributions. At each of these levels, we have run systematic empirical evaluations of the resultant classifiers using actual classroom data. Our evaluations have uncovered satisfactory or better results for many of the