This paper reports on a qualitative study about the application of tag clouds for supporting meta-cognition in self-directed and incidental learning. It analyses the use of a personal tag-cloud visualization of the tags that are used at a public social bookmarking service. The study focuses at the types of metacognitive control based on reflection notes of the learners. These notes were analyzed regarding the contents of the reflections as well as regarding their meta-cognitive type. The study has two important outcomes. Firstly, a personal tag cloud can stimulate reflection on the tagging activity of a learner. Secondly, reflecting on the tagging activity is not built into the design of a tag cloud.
This paper focuses on tag clouds that are based on the personal tagging information of
a learner. For supporting self-directed and incidental learning tag clouds hold some
potential to stimulate reflection on concepts and learning processes, even though such
visualisations do not show valid information in terms of approved domain knowledge.
A tag cloud can stimulate reflection because the visible tags are meaningful to the
learner and allow associations to the actual learning experiences. This paper reports
on a qualitative study about using tagging visualisations to stimulate reflection about
self-directed or incidental learning activities. This article analyses if the visualisation
of a learner's tagging activity in the form of a tag cloud can stimulate meta-cognition
of self-organised learners. The present study analyses the types of reflection that are
stimulated by a personal tag cloud that is based on a mash-up of different information
that is provided by the social bookmarking service delicious.com [
Prior research [
In contrast to formal education, self-directed learning is defined by a high degree
of learner control in weakly structured knowledge domains [
The concept structures in personal learning environments vary for each user and
can range from a few tags to a couple of hundreds. This freedom for the learners
requires that visualizations are independently designed from conceptual structures or
knowledge domains. Related approaches [
The present paper analyses if tag clouds can stimulate and support reflection in
self-directed and incidental learning, which depends on the ability to reflect on actions
[
Lave & Wenger [
The present study uses the ReScope system [
The design of the ReScope tag cloud is based on the contextual dimensions concept and process of the contextualization dimensions of situated learning. 1. Concepts: The overall tag usage of each tag is indicated by its font-size. 2. Process: The most recent usage of tags is indicated through color codes.
Because this information is entirely based on the tags that are used by a learner, it is expected that it can serve as an anchor for reflecting about the personal tagging habits and the related knowledge structures. 4
This study has two main objectives: (a) identify if and how tag clouds can stimulate meta-cognition of learners regarding their self-directed and incidental learning on the Web2.0; and (b) identify design factors for the underlying use case for reflection support. Within this scope this study addresses the following questions. ─ Does the ReScope tag cloud stimulate meta-cognitive activity? ─ Does the reflection on the tag cloud confirm contextual factors suggested by the contextual framework for situated learning?
According to the relations of contextual dimensions and factors, it is expected that the notes can be linked to the context factors rhythm, value, and connection. Rhythm refers to all types of structuring a process. Value covers anything that is related to the added value of concepts and conceptual structures for the learner. The factor connection refers to the structure of the concepts as well as the relation of concepts to resources. The factors “integration” and “community building” are not expected to occur, because ReScope does not visualize connections between peers. 5
The participants were invited through promoted on several mailing lists, web-logs, web-sites of related research projects, and on two European conferences in the domain of computer assisted learning. The participation was entirely voluntarily. The interface of the ReScope system included information texts that asked the participants to write down their thoughts about their reception of the tag cloud in the notes field and to label important words with a hash ('#'). This technique is also known as hash tagging.
Nine weeks after the initial announcement of the system the system's usage has
been analyzed. “Noisy” statements like as “hello” or “xxx” were removed prior to the
analysis. The remaining notes were labeled with the three types of meta-cognitive
control of Ertmer & Newby [
Over the period of nine weeks, 76 users have registered with ReScope. According to the domains of the registered e-mail addresses, participants from 13 countries were identified. In average, the respondents were 36 years old; the youngest participant was 22 and the oldest was 58 years old. The majority looks at the official tag cloud at delicious.com (91%), although 73.5% of the respondents state that they look at their tag cloud only sometimes or rarely. Nine participants left 43 notes about their tag cloud. After cleaning the noise from the notes, 35 notes remained that describe reflective activity of their authors.
The first notes of five users reflect their surprise over the structure of their personal tag cloud and expressed the need to adjust the tags. An example for these notes is given by User A.
“I have too many occurrences of singleton tags #clean-up”.
The note of User A also shows the use of a hash tag. This feature has been used by six participants in 14 notes. The provided hash tags were beneficial for labeling the notes for the analysis. The following two examples illustrate the use of hash tags. In the first note, User F stated.
“need to cut out #duplicate_tags” A more complex example of hash tags is given by the statement of User D.
“Writing on my #blog (++space) I realised I hadn't used #tag podcast very consistently. #consistency important, so I #edited the post.”
All messages were labeled according the three types of reflection for metacognitive control: evaluate, plan, and monitor. One note was labeled with evaluation and planning because the note stated first the result of a self-evaluation and then defined an objective for future tagging. 17 notes are labeled as evaluative reflection, six notes as planning, and 13 notes as monitoring.
The results from the model centric labeling show that all three types of reflections were identified with the notes (see Table 2). The majority of the notes were labeled as evaluative reflection. A fifth of all notes and two fifths of those that were tagged as evaluative reflection referred to cleaning the tagging structure.
Additionally a content based labeling has done on all annotations. After this procedure a note could have many labels assigned. This step brought up three key categories: 23 notes were reflections about the semantic structure of the tag cloud, six notes were reflections on organizing the tagging process or on aligning it with other tasks, and nine notes were reflections on system features. Furthermore, two notes were labeled as reflections on social interaction and one note was labeled as a personal remark. Four notes were classified as follow-ups to prior notes that marked steps in a process. Within the notes on the semantic structure, three sub-categories were identified. The participants reflected in seven notes on the need of cleaning up the tag structure, in eight notes about the consistency of how the tags are used, in two notes about emerging tags.
The present study focused on the potential of tag clouds for supporting reflection on
the personal tagging of self-directed learners. It combines and extends the prior
research on visualizations of learner activity based on knowledge domain independent
information [
The results show that already the rather simple visualization stimulate reflection are conforming the models of meta-cognitive control. The notes of the content based labeling are equally distributed across all types of reflection (Table 2). This suggests that the tag cloud supports meta-cognitive processes in self-directed learning of all levels similarly. Based on the present results it can be argued that tag cloud visualizations can serve as a reflection support tool for self-directed learners.
The results of the content labeling can be interpreted as instances of the context factors for situated learning. By mapping the identified labels on the context factors, it becomes possible to address the second research question. The design of the tag cloud focused at the context dimensions process and concept. Therefore, it is expected that the participants' notes are related to the context factors rhythm, value, and connection (Table 1). Notes that were labeled only as semantic structure but not to any of the sub-categories were mapped to the connection factor, because these notes reflect the relations between the tags and the relations between tags and resources. Notes of the categories process related and emerging topics were mapped to the rhythm factor.
Finally, the notes of the categories clean-up and consistency were mapped to the value factor, because these notes refer to the value of the tags for the personal information management. Table 3 shows the results after remapping of the contextual factors. The results show that the reflections related to the three factors were almost equally stimulated by the tag cloud. This finding supports the expectations regarding the relevance of the contextual factors for reflection. 8
Reflection is the active reasoning that has been described as the learner's assessment and validation of experiences in problem solving and social interaction against conceptual structures and strategies. This process is part of developing consciousness about the relations between actions, beliefs, and concepts. Therefore, reflection is an important meta-cognitive process related to learning. This article analyzed if a tag cloud visualization of a learner's free form tags can stimulate such meta-cognitive processes. This study has two important results. 1. A personal tag cloud can stimulate reflection on the tagging activity of a learner. 2. The concepts of situated learning can be applied for developing tools for supporting self-directed learners.
The findings of this study indicate that targeted solutions for supporting metacognition do not need to be dependent on pre-structured domain knowledge. This opens opportunities for developing new forms of supporting self-directed and incidental learning in personal learning environments. 9 19. O’Reilly, T. (2005) What is web 2.0? Design pattern and business models for the next generation of software. Retrieved 01/12/2008 from http://www.oreillynet.com/pub/a/oreilly/tim/news/2005/09/30/what-is-web20.html 20. Schön, D. A. (1983) The Reflective Practitioner: How Professionals think in
Action. London, Maurice Temple Smith. 21. Smith, G. (2008) Tagging, people-powered metadata for the social web. Berkley,
CA: New Riders. 22. Wenger, E. (1998) Communities of practice; learning, meaning, and identity.
Cambridge: Cambridge University Press. 23. Wexelblat, A. & Maes, P. (1999) Footprints: history-rich tools for information foraging. Conference on Human Factors in Computing Systems. 17, 270-277. 24. Wilson, S., Liber, O., Johnson, M., Beauvoir, P., Sharples, P. & Milligan, C. (2006) Personal Learning Environments: Challenging the Dominant Design of Educational Systems. In Innovative Approaches for Learning and Knowledge Sharing at the EC-TEL 2006 (eds. E. Tomadaki & P. Scott), 173-182.