Supporting learning and information sharing in natural resource management with technologies for electronic documents.
"Sustainability is better seen as a measure of the relationship between the community as learners and their environment rather than an externally designed goal to be achieved," (Sriskandarajah, Bawden, & Packham, 1991).
In Australia the water crisis is worsening. There is a major problem with drought and our use of natural capital is not sustainable. The community is crying out for sustainable solutions. This has lead to one of the biggest consultation exercises in natural resource management (NRM) undertaken in Australia involving government, industry, the community, Co-operative Research Centres (CRCs), universities and Commonwealth Scientific & Industrial Research Organisation (CSIRO). All these stakeholders took part in the formulation of the business case of the CSIRO Healthy Country Flagship Program (CSIRO, 2003).
The Healthy Country Flagship Program's aim is to achieve sustainable NRM through informed participation by engaging community and stakeholders' groups. A prerequisite to efficient, constructive participation is that community and stakeholders groups have access to different knowledge sources, are more closely attuned to the different issues and viewpoints, and are sufficiently equipped to understand (and maybe resolve) complex issues (salinity, ecosystem stability, erosion, grazing, nutrients, etc.).
Traditional support for knowledge sharing and learning approaches has mostly focused on documents sharing. These approaches have been successful only in very specific and constrained environments where the task people are engaged in is well defined, the people are collaboratively working towards a common goal and people are within a similar practice. In contrast, decision making in natural resource management takes place among various communities (social, cognitive and political), with different practices (farmers, tourist operators, state and federal regulatory agencies, etc.), engaged in ill-defined, complex tasks with conflicting goals using various information types (databases, documents, decision support systems, ecosystem models, etc.). In particular, from our research in this area, we have noted that:
* Little knowledge generated in science is directly impacting the practices and decision making within the communities. NRM plans and strategies need to be continually adapted to reflect new scientific knowledge.
* Land managers are central to achieving sound management of land, water and vegetation resources and to addressing critical issues such as salinity, yet they do not always have the required information to make sound decisions about the management of natural resources.
* Local and indigenous knowledge is not always taken into account. This leads to reduced community ownership of local problems, and little adoption of new methods and policies (Productivity Commission Report, 2003).
* There is little understanding of other stakeholders' views and issues.
It is clear that support for knowledge sharing is key to searching for sustainable NRM solutions. We have taken the approach that by improving the access of the stakeholders to relevant information, supporting idea sharing, model exploration, information annotation (i.e., with local knowledge), and providing an editing mechanism to capture new knowledge, we will alleviate some of the issues identified above.
Supporting Learning and Knowledge Sharing
As global competition based on knowledge intensive products/services rapidly increases, many organizations are seeking ways to harness knowledge through business strategies and Information & Communication Technology. Computer networks, Internet and Intranet, e-mail, bulletin boards, groupware, workflow, news groups, data warehousing, decision support systems, Lotus Notes etc. have already become important media for knowledge creation, sharing, and transmission (Liebowitz, 1999; Macintosh, 1994; O'Leary, 1997). These tools are core to knowledge management. Research in technologies for supporting learning and knowledge sharing often uses a combination of:
* The data/document approach using databases and document repositories with associated data mining and search engine facilities.
* The knowledge-based approach uses ontologies representing knowledge models (Gandon, 2001; Decker, Erdmann, Fensel, & Studer, 1999), cases representing past experiences (Simon & Granbastien, 1995), lessons learned representing current practices (Alem, 1998).
* The people finder approach with associated yellow pages, expertise finding systems (McLean, Vercoustre, & Wu, 2004; Craswell, Hawking, Vercoustre, & Wilkins, 2001), peer helper technologies (McCalla, Greer, Kumar, Meagher, Collins, Tkatch, & Parkinson, 1997). The expertise finding capability has been coupled with lessons learned corporate memory (Alem & McLean, 2003).
* The collaborative approach using computer supported co-operative work, video conferencing and mind mapping technologies. Work by knowledge models (a conceptual model of the domain and a meta model describing the terminology structure) have been used for supporting collaborative work (Kethers, von Buol, Jarke, & Rudolf Repges, 1998).
* The community centered approach uses interaction through online communities of practices (COPs) and communities of interest (COIs), chat rooms, or bulletin board technologies (Preece, 2002; Brown, van Dam, Earnshaw, Encarnacao, Guedj, Preece, Shneiderman, & Vince 1999; Walker and McCown, 2003).
The limitations of the database, document and knowledge based approaches include operating in a very constrained environment where the tasks people are engaged in are well defined, with people who are collaboratively working towards a common goal within a similar practice (automotive engineers, aerospace engineers, offshore oil operators). They commonly work within one organization whose leaders are supportive of knowledge sharing.
Furthermore, the knowledge based approach is often very labour intensive. Building the ontology, maintaining it, and manually annotating documents requires a great deal of work. There is a need for a more cost-effective (light weight) approach. Also, as far as we know, the knowledge based approach has dealt mostly with representing and exploiting ontologies and lessons learned models. We do not know of any work using this approach that represents and exploits more physical models (of ecosystems, for example).
Finally, the community-based approach is often restricted to supporting one specific practice, for example, farmers (Walker, Cowell, & Johnson, 2001) or health practitioners (Preece, 2002). As far as we know, little has been done in linking the community centred approach with the data/document and knowledge based approach.
It is also significant that the impact of knowledge sharing is generally not evaluated. Important questions need to be addressed, such as whether knowledge sharing led to learning, better community understanding, and better environment management.
These needs suggest that supporting learning and knowledge sharing in NRM requires an integrated approach, combining a community-centred approach with data/document and knowledge-based approaches, supporting not only document sharing, but also idea and view sharing and collective exploration of ecosystem models in an open, evolving and networked information environment.
Our Proposed Research Framework
Our aim is to support individual participation as well as group participation, in a knowledge sharing and learning environment where relationships between participants are being developed (through collaboration models and social networks), resources are being used (documents, web resources, models), new information/knowledge is created and shared (collaborative design of NRM plans and strategies, collaborative design of biophysical models) and learning is supported (e-learning).
Our framework (Figure 1) incorporates the three dimensions: information (documents), discussion (forum) and knowledge models (domain knowledge, user models, biophysical models, simulation models), all dynamically evolving through users' participation (as suggested by the large arrow).
Central to our approach is the notion of designing lightweight models and leveraging on these models for supporting learning and knowledge sharing. This approach is more cost effective than the knowledge intensive approach such as that of (Gandon, 2001). A knowledge model in a form of a knowledge map is composed of a set of topics (concepts), and associations between the topics (is-a can-contribute-to, can-reduce, etc.). The knowledge map is used as the basis for designing a structured and semantically rich navigation space and is also used for generating advanced queries of the document collection. We can associate with each concept a set of query words and other search operators that are to be used to retrieve external (i.e., web) information when the user is examining this concept. The query can be created manually, by an expert or the participants themselves, or by the system, through analysis of the knowledge map. We argue that the combination of model-driven navigation and an advanced search facility provides access to relevant information to a specific topic in the context of the topic's relations and associations with other topics. In doing so, the environment supports learning about the domain.
[FIGURE 1 OMITTED]
For our research we examined water quality issues in the Wet Tropics coastal region in North Queensland (Douglas Shire). We created a knowledge map for this domain that includes the following high level concepts: land use, land practices, effect on pollution and water quality issues. For example (see Figure 2), horticulture is a type of land use in the Wet Tropics; use of chemical and irrigation are current practices in horticulture; herbicides, pesticides and fertilizers are types of use of chemicals; the use of chemicals contributes to contaminant runoff which then is a water quality issue.
Consider a user is interested in finding out about the impact of horticulture practices in the Wet Tropics. The user can navigate the map, find related water quality issues as well as open relevant documents related to the use of chemicals. Relevant documents are made accessible and presented in a form that increases the user's awareness about how information relates to each other.
We also promote the notion that the environment should be open and able, where appropriate, to take advantage of external information. One example is simply using external web resources that are well known sources of domain specific knowledge such as Department of Sustainable Environment (DSE), Catchment Management Authority (CMA), Department of Primary Industry (DPI), etc. Thus information delivered to a particular user in response to a query may not just come from information captured within the environment, but also from external web sites and databases. This helps to ensure that the information gathered and presented to the user is up-to-date. Another example is to use a people finding tool, such as that described in (McLean et al., 2004) that automatically extracts evidence of expertise for individuals and can be used to recommend relevant people as well as information. In this way, we hope to encourage further discussion amongst the participants.
[FIGURE 2 OMITTED]
Our knowledge sharing environment extends classical portal architectures in the following three ways:
* We add an explicit knowledge model level to support better access to and understanding of information.
* We open the portal to external web sources that are dynamically added to the portal.
* We offer a participation component that elicits people expertise, roles, networking and involvement in discussion.
The next session describes our initial implementation, in the context of our engagement with Douglas Shire community.
Our Initial Implementation
The project objective is to electronically capture and deliver information to targeted members of the DS community (mostly MAS and DSC) in a way that helps understanding of water quality issues and water quality monitoring data. Our research hypotheses were 1) access to information and data provides insights and hence may lead to better decisions and 2) sharing of existing information and knowledge among the various members of the DS community will support shared understanding and hence may lead to improved participation.
Our initial implementation is aimed at delivering relevant information to a user through leveraging domain knowledge encoded in the knowledge network. For the network we gathered information from domain users and domain literature and used a topic map representation (2). Our knowledge map contains over 150 entities and 250 relationships (3). The top level map is shown in Figure 3(a) and a sample graphical representation of part of the actual map is shown in Figure 3(b). We also harvested and indexed document data from a set of around 25 websites amounting to about 15 Gigabytes of data. Each entity had a default query that could be used against this document set; however, for many entities we manually crafted queries that were more specific. We provided some textual description for some of the entities which gave a short summary of the concept and the system generated a concept description that not only used this text but also described how a particular entity was related to other entities. Finally we also manually attached documents, people, and database records to certain entities where we were confident that the attached document was highly relevant. This provided an initial information space in which the users could browse and is illustrated in Figure 4. In this figure the left-hand panel shows the knowledge map and is primarily used as a navigation panel. The top right panel provides a natural language description of the current node and its relationship to other entities in the network. This description is generated automatically. The central panel shows all the data objects that have been manually attached to this particular node. These data objects can be documents, web pages, database records, people, images, and in the future video. The bottom right panel shows the results of the query for this entity on the harvested data in a standard search page format. If the user decides that a particular result is of high relevance he can attach that document to the entity by clicking on the image link next to the result. The user can also refine the search query in this pane.
[FIGURE 3 OMITTED]
A user can also annotate a particular entity by right-clicking on the current concept name in the navigation panel, or by right-clicking on the link of an attached document on the attached-document panel. This allows the user to select an option to annotate the item and he can then provide a comment or note. In this way, we can support the capture of local knowledge and the collective creation of new knowledge.
We also provide a map-based entry to the browser. There are 12 water quality stations in Douglas shire at various river sites. These stations are measuring water quality as part of another project in which CSIRO is participating. We are able to access web services that allow us to query the location of the sites. We use this to dynamically create a map of Douglas Shire showing each water quality station that is active. We then map the locations to relevant concepts in the knowledge map. If, for example, a water quality station is situated in the middle of a sugar cane farming area, we allow the user to select the "sugar cane" concept from a drop-down list of relevant topics when he clicks on the water quality station on the map (see Figure 5).
[FIGURE 4 OMITTED]
Figure 5 also show the main screen of the editor. The editor mode allows users to add, edit or delete new concepts and new relationships, link concepts by a relationship, add/remove contacts and add/remove attached documents.
In a collaborative environment with many users, the idea is that this open, web-based information space will support the capture of knowledge from different users. Although we have not implemented linked discussions yet, we hope that the browser can provide an entry point into informed discussion about natural resource management issues.
Evaluation of the Prototype
There are three dimensions of our approach that could be evaluated:
* Learning: how effective the approach is in terms of improving end users' level of understanding of water quality issues and their relations to land uses and land practices.
* Capture: how effective the system is in terms of supporting the authoring of the Knowledge Map: adding/removing concepts, adding/removing links, adding/removing information resources (documents, data, spatial maps).
* Sharing: how effective the approach is in terms of supporting document sharing, models sharing, experience sharing.
[FIGURE 5 OMITTED]
The evaluation we are reporting on here examined the first dimension only. The evaluation of our approach has been performed with 16 students from the Department of Arts at Monash University. The evaluation consisted of the following steps:
* Assess the student's computer skills, domain knowledge and spatial awareness (using student profile questionnaire).
* Provide time for the student to become familiar with the knowledge portal.
* Ask the student to answer five review questions.
* Assess the student's overall learning experience using the knowledge portal (using an exit-questionnaire).
Then, for each of the tasks and for each student:
* Assess the student's prior knowledge (using pre-knowledge evaluation questionnaire).
* Assess the student's post-knowledge (using post-knowledge evaluation questionnaire).
* Capture the student's log.
The five questions were:
Q1: Explain the nature of the impact of pig-digging on water quality in Douglas Shire.
Q2: What factors affect water quality in the Douglas Shire, and how do they relate to each other?
Q3: What practices minimize horticultural impact on the environment?
Q4: List causes of water system degradation in Douglas Shire.
Q5: What are "sea grass meadows" or "sea grass beds," and what are the consequences of sea grass meadow degradation?
The students were asked to answer the five questions in a different order (Figure 6) to compound the factor related to learning about the system.
An expert was asked to assess learning gain based on student's pre- and post-answers to questions using the following score: 0 no change between pre- and post-answer, 25% positive change, 50% positive change, 75% positive change, and 100% positive change. Figure 7 shows the students' overall learning gain per question.
The learning gains of students with low initial knowledge level are shown in Figure 8.
Figure 9 compares the average learning gain for the three different groups: students with a high level of domain knowledge (series1), all students (series2), and students with a low level of domain knowledge (series3). The figure shows that students with a high level of domain knowledge learned less than students with a low level of domain knowledge, and students with a low level of domain knowledge have similar learning gains to the average of the whole group. We compared the distribution of learning gains per question by using the two-tailed paired t-test. Statistically significant differences between the two groups occurred when the result of the t-test was less then 0.05. We obtained: t-test (Low, All) = 0.173477 and t-test (High, Low) = 0.016521. Thus, we can see that the High group has learning gains that are statistically different from the Low group, but the Low group could not differentiate the learning gains from the whole group average.
[FIGURE 7 OMITTED]
[FIGURE 8 OMITTED]
[FIGURE 9 OMITTED]
In this paper we have proposed a research framework for supporting learning and knowledge sharing by making use of technologies for electronic documents. The backbone of such an environment is an open, dynamic and evolving knowledge map composed of documents, knowledge models (ecosystems models, data models) as well as people. Users are able to navigate through electronic resources via the evolving knowledge map and to gain greater understanding of the relationships, issues, relevant information sources and people, thus enabling them to hold more informed discussions. We have presented our initial implementation and evaluation in the context of a case study in Douglas Shire: a sugar cane area in far North Queensland. Initial analysis of our evaluation data indicates that our environment supports learning for different profiles of learners. Learners with a low level of domain knowledge have better learning gain than learners with a high level domain knowledge.
Our next step is to further analyse our evaluation data and explore the various factors that may explain learning gains. Our second step is to assess with members of the community in Douglas Shire, the extent to which our technology is effective in terms of supporting capture and sharing. But as stated by Huysman & DeWit (2002):
One should not fall into the known trap of assuming that it is the use of these technologies that stimulates people to communicate and share knowledge. The first thing to be addressed is the question of how to stimulate a need to share knowledge among a group of people. It is only when this need is satisfied that physical and electronic spaces are used for knowledge sharing purposes.
When such a need exists, we can envisage other tools that may extend the knowledge sharing space. These tools include a people finder tool that helps to locate people with the required expertise together with some evidence of the expertise. A facilitator can use this tool to solicit an expert to engage in a current discussion. Whereas a person can currently be manually attached to a concept, a people finder tool uses the document and data, both within the portal and pointed to by the portal, as evidence for people's expertise and finds people automatically. We can also envisage support for collaborative creation of new models as a result of the discussion. As we place our emphasis more on collaborative understanding and sharing rather than information access and creation, we can imagine tools that support group awareness, measure the level of participation, analyse the participation process, and even visualise credibility based on expertise, reputation, contribution and social network analysis.
Student 1,2,3,4 Q1Q2Q3Q4Q5 Student 5,6,7 Q5 Q1Q2Q3Q4 Student 8,9,10 Q4Q5 Q1Q2Q3 Student 11,12,13 Q3 Q4Q5 Q1Q2 Student 14,15,16 Q2 Q3 Q4Q5 Q1 Figure 6. Order of questions
We would like to thank Anne-Marie Vercoustre and Michael Flower for their contribution to this work. Special thanks to Professor Jim Peterson, Director of the Centre for GIS at Monash University, Xiaoye Liu, and the 16 students who participated in the evaluation of our learning environment.
(1) Commonwealth Scientific & Industrial Research Organisation Information and Communication Technology
(2) See http://www.topicmaps.org
(3) We concentrated our effort in creating links around certain concepts.
Alem, L. (1998). Learning in the workplace: Initial requirements of a lessons learned centred corporate memory. Proceedings of the Intelligent Tutoring Systems: 4th International Conference, San Antonio, Texas.
Alem, L., & McLean, A. (2003). Technologies for electronic documents to support learning. Paper presented at the AI-ED Workshop, Sydney, Australia.
Brown, J., van Dam, A., Earnshaw, R., Encarnacao, J., Guedj, R., Preece, J., Shneiderman, B., & Vince, J. (1999). Human-centred computing, online communities and virtual environments. ACM SIGCHI Interactions, 6(5), 9-16.
Craswell, N., Hawking, D., Vercoustre, A., & Wilkins, P. (2001, April). P@NOPTIC Expert: Searching for experts not just for documents. Proceedings of Ausweb 2001, 21-25.
CSIRO (2003). Healthy Country, A Case for CSIRO Investment. Internal Report.
Decker, S., Erdmann, M., Fensel, D., & Studer, R. (1999). Ontobroker: Ontology based access to distributed and semi-structured information. (R. Meersman, Z. Tari, & S. Stevens, Eds.). DS8: Semantic Issues in Multimedia Systems. Berlin: Kluwer Academic Publishers.
Gandon, F. (2001). Experience in ontology engineering for a multi-agents corporate memory system. Proceedings of the Workshop on Ontologies and Information Sharing, IJCAI 2001, Seventeenth International Joint Conference on Artificial Intelligence, 119-122.
Huysman, M., & DeWit, D. (2002). Knowledge Sharing in Practice. Norwell, MA: Kluwer Academic Publishers.
Kethers, S., von Buol, B., Jarke, M., & Rudolf Repges, R. (1998). Lessons learned from cooperative terminology work in the medical domain. Proceedings of MedInfo 1998, 628-632.
Liebowitz, J. (1999). Knowledge Management Handbook. Boca Raton, FL: CRC Press.
Macintosh, A. (1994). Corporate knowledge management: State of the art review. Proceedings of the International Symposium on the management of industrial and corporate knowledge (ISMICK94), 26-27.
McCalla, G., Greer, J., Kumar, V., Meagher, P., Collins J., Tkatch, R., & Parkinson, B. (1997). A peer help system for workplace training. Proceedings of Artificial Intelligence in Education, IOS Press.
McLean, A., Vercoustre, A., & Wu, M. (2004). Combining structured corporate data and document content to improve expertise finding. Submitted for publication.
O'Leary, D.E. (1997). The Internet, Intranets and the Al Renaissance. Computers, 30(1), 71-78.
Preece, J. (2002). Online communities: Designing usability and supporting sociability. Hoboken, NJ: John Wiley & Sons, Inc.
Productivity Commission (2003). Industries, land use and water quality in the Great Barrier Reef catchment. Research Report, Canberra.
Simon, G., & Grandbastien, M. (1995). Corporate knowledge: A case study on the detection of metallurgical flaws. Proceedings of ISMIK'95, France, 43-52.
Sriskandarajah, N., Bawden, R.J., & Packham, R.G. (1991). Systems agriculture: A paradigm for sustainability. Association for Farming Systems Research-Extension Newsletter, 2(3), 1-5.
Walker, D.H., Cowell, S.G., & Johnson, A.K.L. (2001). Integrating research results into decision making about natural resource management at a catchment scale. Agricultural Systems, 69, 1-2, 85-98.
Walker, D., & McCown, B. (2003). A community of practice approach to organizational learning on engaging with communities for improved natural resource management. CSIRO, Internal Report.
LEILA ALEM AND ALISTAIR McLEAN
CSIRO ICT (1), Australia
|Printer friendly Cite/link Email Feedback|
|Publication:||International Journal on E-Learning|
|Date:||Jan 1, 2005|
|Previous Article:||Data mining technology for the evaluation of learning content interaction.|
|Next Article:||Designing agents for feedback using the documents produced in learning.|