Introduction

1There is some evidence to show that spatial decisions made collectively tend to be more effective than decisions made by an individual decision maker (e.g., Joerin et al., 2009; Simao et al., 2009). It is suggested that the spatial planning/decision making should involve the use of a collaborative/participatory approaches, where individuals with different backgrounds can be brought together in a group to solve a particular problem (e.g., Bailey et al., 2003; Kyem, 2004; Bugs et al. 2010). The Multicriteria Spatial Decision Support Systems (MC-SDSS) can play a key role in the collaborative spatial decision making (Jankowski and Nyerges, 2001a). These systems integrate Geographic Information System (GIS) capabilities (spatial databases and analyses) and Multicriteria Decision Analysis (MCDA) techniques to support a user or a group of users in making spatial decisions (Malczewski, 1999; 2004). The basic rationale behind the efforts to integrate GIS and MCDA is that the two distinctive areas of research can complement each other. While GIS offers unique capabilities for storing, managing, analyzing, and visualizing spatial data for decision making, MCDA provides a rich collection of techniques and procedures for structuring decision problems, designing, evaluating, and prioritizing alternative decisions. Although GIS-MCDA approaches have traditionally focused on the MCDA techniques for individual decision making, substantial efforts have recently been made to integrate GIS with MCDA for group/collaborative/participatory decision making (Simao et al., 2009; Jankowski and Nyerges, 2001a; 2001b; Malczewski, 1996; Feick and Hall, 2002; Bailey et al., 2003; Kyem, 2004).

2Web 2.0, a new trend in the World Wide Web development, shifts the Web to a participatory and fully interactive platform. It allows two-way communication; that is, a read-write Web by means of which people are contributing as well as consuming information. Integration of GIS-MCDA into the Web 2.0 platform offers an effective MC-SDSS for public participation, providing appropriate analytical tools and platforms to directly involve the public in the spatial planning process. Web 2.0-based MC-SDSS tools shift multicriteria decision making from a closed, place-based (fixed time and location), synchronous procedure to an open, asynchronous, distributed, and active decision making process. Over the last decade or so, significant research efforts have been made to integrate GIS and MCDA methods into the Web environment (Simao et al., 2009; Sikder and Gangopadhyay, 2002; Voss et al., 2004; Hall and Leahy, 2006; Karnatak et al., 2007; Rao et al., 2007; Jankowski et al., 2008; Boroushaki and Malczewski, 2010a; Meng and Malczewski, 2009). There are, however, two critical challenges with the Web 2.0-based MC-SDSS. First, the systems have been built based on models containing collaborative knowledge (e.g., community/collaborative knowledge, GIS-MCDA knowledge, etc.) targeted at human consumption and interpretation, where only the human is able to interpret the knowledge (Lassila and Swick, 1999). Exchange of any collaborative knowledge elements relies on the decision maker’s common sense to interpret knowledge elements and share them. It is not possible to automatically reason and interpret the knowledge embedded in such human-consumable GIS-MCDA models, making it hard for software agents to process, reason, share and reuse the collaborative GIS-MCDA knowledge on the Web. Second, the conventional collaborative GIS-MCDA models represent the subjective knowledge of one or a few individuals; they are not a shared and consensual model accepted by a group or community. Participants (decision makers) may have different views of the same decision problem area when participating in decision making process and require a unified, common, consensual, and shared reference for collaboration. Therefore, we suggest that the use of conventional GIS-MCDA in collaborative decision making can be enhanced by focusing on a Web 3.0-driven MC-SDSS.

3This paper proposes a Web 3.0-based MC-SDSS for collaborative decision making. The term Web 3.0 was first coined by John Markoff (The New York Times) in 2006 and many have tried to define Web 3.0 since then. Specifically, the MC-SDSS approach presented in this paper is based on the definition proposed by Wahlster and Dengel (2006): Web 3.0 = Semantic Web + Web 2.0. According to this definition, Web 3.0 is the integration of Semantic Web (Web of meaning) technologies with the principles of Web 2.0 (Web of people). The vision of Web 3.0 here is to synergistically integrate the emerging Semantic Web and Web 2.0 community-based methods (Wahlster and Dengel, 2006), where the participatory information on the Web 2.0 platform can be stored and represented in such a way that a computer can understand it as well as a human. Semantic Web mainly relies on an ontology for providing a formal (machine-readable and-understandable), semantic, shared (consensual) knowledge framework. Ontology captures some consensual knowledge and conveys a shared understanding of a domain that is agreed among a number of individuals or experts (Gaŝević et al., 2009). The proposed approach integrates the participatory/collaborative GIS-MCDA knowledge (the community approach to Web 2.0) in an ontological framework as the backbone of Semantic Web. The remainder of the paper is organized as follows. Section 2 focuses on Web 3.0-based collaborative MC-SDSS including collaborative GIS-MCDA ontology and the collaborative GIS-MCDA decision rule. Section 3 demonstrates the implementation of the proposed approach. It describes a case study of site selection problem in Teheran, Iran. Finally, Section 4 provides concluding comments.

Web 3.0-based collaborative MC-SDSS

Collaborative GIS-MCDA ontology

4The term “ontology” has different connotation in different disciplines. Guarino and Giaretta (1995) distinguish between “Ontology” (an uncountable noun with uppercase initial) as a philosophical discipline and “an ontology” (a countable noun with lowercase initial) as a computational discipline. In the philosophical field, Ontology refers to the subject of existence, i.e., the study of the categories of things that exist or may exist in some domain. It deals with the questions of what entities exist or what is, how these entities are related, and how can entities be categorized. In the computational discipline, an ontology refers to as a formal (machine-readable), explicit specification (declarative representation) of a shared (consensual) understanding and conceptualization (an abstract, simplified view of the world) of a particular domain (Gruber, 1993; Gaŝević et al., 2009). Collaborative GIS-MCDA ontology (computational discipline) aims at: (1) representing formally and semantically the knowledge elements of collaborative GIS-MCDA in a machine-understandable framework, and (2) providing a semantic GIS-MCDA knowledge skeleton for analyzing and solving spatial decision problems. It integrates the domain knowledge associated with three distinctive areas of applied research: MCDA, GIS, and collaborative decision making (see Figure 1).

5The ontology represents the domain GIS-MCDA knowledge elements including: (1) decision goals that characterize an improvement from the present state toward a desirable state, (2) decision constraints, (3) a set of evaluation criteria, (4) decision makers’ preferences, (5) the set of decision alternatives which represent different decision options, and (6) alternative evaluation outcomes (alternative ratings and orderings). The hierarchical nature of ontologies (meaningful hierarchy of concepts built on a generalization/specialization relationship) allows for modeling the MCDA concepts in a hierarchical structure (Kwon and Kim, 2004; Mahmoudi and Muller-Schloer, 2009; Sadeghi-Niaraki, 2009). These elements are organized in a hierarchical structure with the top level corresponding to the ultimate goal of the decision at hand. A goal essentially describes an improvement from the present state toward a desirable state. The decision criteria can be structured based on the top-down development approach (Noy and McGuinness, 2001), which starts with the definition of the most general concepts in the domain and subsequent specialization of the concepts. Two types of criteria concepts can be defined: objective and attribute. An objective is a statement about the desired state of the decision problem under consideration. It indicates the directions of improvement of one or more attributes. For any objective, several different attributes can be defined, providing complete assessment of the degree to which the objective might be achieved. Attributes are measurable characteristics expressing the degree to which the associated objectives are achieved for a particular decision alternative (Jankowski and Nyerges, 2001a). In the spatial context, an attribute concept describes a measurable quantity or quality of a geographic entity or a relationship between geographic entities.

6Participants may consist of a single or a group of decision makers (e.g., government, community, public, etc.) participating in the decision making process. An important task of the participants is to identify and assign their individual preferences with respect to the decision problem by determining the relative importance (weights) of criteria (objectives and attributes) against which the alternatives are evaluated (see dotted lines in Figure 1). Preferences reflect the values and interests of participants with respect to the decision criteria. A variety of comments including suggestions for designing new alternatives, the proposals for eliminating one or more pre-defined alternative parking locations, identifying some concerns with respect to a particular location, etc. can be input into the ontology by participants. The ontology stores the individual’s geo-referenced comments based on the conceptual model “Argumentation Map” proposed by Rinner (2001). It structures the comments, comment’s locations, and their many-to-many relationships. An individual identifies a location and makes a comment on it. A comment can refer to multiple locations, a location can be referenced by multiple arguments, a comment can be logically related to a number of other comments, and locations can be spatially related. Two types of orderings are generated for the alternatives: individual and group orderings. Individual ordering of alternatives are obtained based on the individual preferences. The individual ordering can be aggregated to generate group orderings. There is a consensus degree for each of the alternative orderings indicating the agreement among the individuals, as manifested in the individual solutions, with respected to the group solution (Boroushaki and Malczewski, 2010a).

7A formal (machine-processable) specification of the collaborative GIS-MCDA ontology needs to be developed using the Ontology language. The most common and popular ontology language is OWL (Ontology Web Language) from the World Wide Web Consortium (W3C). OWL builds on the RDF (Resource Description Framework) Schema, an XML-based vocabulary, specifying the concepts and their relationships, expressing the formalism with clearly- and well-defined semantics over which automated reasoning can take place (Brickley and Guha, 2004; Noguera et al., 2010). In OWL ontology, concepts and semantic relationships are represented by means of OWL classes and properties, respectively. OWL classes are created using the formal specifications that an instance needs to satisfy them to be the member of a particular class. OWL properties are used to state the semantic relationships between instances or from instances to data values.

Decision analysis

8Using the OWL GIS-MCDA ontology, the system provides a multiple-user-single-model (a single and shared GIS-MCDA model) framework for decision analysis. It uses a decision rule involving two stages: (1) the decision rule for modeling the individual decision making, and (2) the decision rule for combining individual preferences to produce group solution. The group decision rule takes the form of combining the individual preferences into a group preference in a way whereby the best compromise (the preferred alternative) can be identified (Boroushaki and Malczewski, 2010a).

Individual decision rule

9The decision rule for individual decision making involves combining the relevant spatial data (criterion values) and individual’s preferences to provide an overall assessment (ratings /ordering) of the decision alternatives. Let us assume that the decision problem involves a set of potential alternatives Ai (i = 1,2,..,m) and a set of criteria (or attributes) Cj (j = 1,2,…,n) on the basis of which the alternatives are evaluated. Further, suppose that there is a group consisting of individuals (or decision-makers, stakeholders, interest groups). Let Ik represent an individual involved in collaborative/participatory decision making (k = 1, 2, …, g). Each individual can specify his/her preferences with respect to criteria by assigning a weight of relative importance. Given the set of criterion values for each alternative location and corresponding decision-makers’ preferences, one can use weighted summation method to generate an overall evaluation rating for each decision alternative. This method computes the individual rating, R(Ik, Ai) by summing the weighted normalized criterion values associated with the i-th alternative. Corresponding to the individual ratings, individual orderings, O(Ik, Ai), for each of the alternatives are obtained.

Group decision rule

10The individual solutions can be aggregated into a group solution using a group/collective decision rule (e.g., Malczewski, 1996; Jankowski and Nyerges, 2001b; Feick and Hall, 2004). There is evidence to show that a combination of MCDA for individual decision making with voting techniques provides an effective tool for collaborative decision making in the GIS environment (Jankowski and Nyerges, 2001a). Here, a vote aggregation method involving the Borda count (Borda, 1781) is used as the collective decision rule. Given the individual orderings, O(Ik, Ai), one can derive the individual preference set based on the pairwise comparisons. For each pair of alternatives, Ai and Ap, the i-th alternativegets 1 point, if Ai is preferred over Ap; and 0.5 if an individual is indifferent between the two alternatives. The group overall score for the i-th alternative is calculated by summing the individual preference scores for that alternative.

11Figure 2 illustrates the conceptual framework for the group decision rule using an example. As shown, a group of four individuals input their preferences (criteria relative importance) and the system generates individual solutions (orderings) for five candidate sites according to the individual preferences. Once all the individual orderings have been determined, the group ordering for each of the alternatives is obtained.

Prototype implementation: A case study in Tehran, Iran Study area

12The rapidly changing pattern of urban growth of Tehran, Iran has led to the overuse and shortage of basic urban infrastructure. The public parking shortage in the city has emerged as an area of serious concern. In the recent years, urban planners and municipality departments have taken effective measures to increase the number of public parking facilities in different districts of the city. Tehran is divided into 22 municipal districts, each with its own administrative center (see Figure 3). The proposed system has been used for solving parking site selection problem in the center of the District # 22. A set of 24 candidate sites for locating parking have been identified in the study area. Participation in the parking planning process was open to the general public on a 24/7 (24 hours, 7 days) base. A total of 30 volunteers participated throughout the parking site selection process.

System description

13The architecture of the proposed Web 3.0-based collaborative MC-SDSS is shown in Figure 4. The system is developed based on the thin client approach (Peng and Tsou, 2003), where the user interface components (a Web browser) run on the client (user) machine but the application logic (GIS and decision analysis) and data elements (GIS data and OWL ontology) remain on the server. The OWL collaborative GIS-MCDA for parking site selection problem has been developed using Protégé-OWL editor. It is an extension of Protégé that enables users to load and save OWL ontologies, edit and visualize classes, properties, and define logical class characteristics as OWL expressions, and execute reasoners such as description logic classifiers (Horridge et al, 2004). The evaluation criteria have been determined based on the previous studies and interviews with the experts (e.g., Jiaxi, 2003; Matkan et al., 2009; Farzanmanesh et al., 2010; Boroushaki and Malczewski, 2010b; Talebi, 2010). The criteria (objectives) include: (1) proximity to roads, (2) proximity to recreational services, (3) proximity to administrative centers, (4) proximity to commercial centers, (5) proximity to educational services, (6) proximity to health centers, (7) population density (i.e., adjacent population to a candidate site), (8) the size of a candidate site, and (9) cost of land acquisition. For the proximity objectives, two spatial attributes “Number of neighboring features” and “Average distance” are specified. The former attribute indicates the number of target features of the same type (e.g., administrative centers) within a particular distance from a potential parking candidate site. The latter indicates the average distance between the parking candidate site and target features.

14Attribute value table, also referred to as attribute-alternative matrix, contains the normalized attribute values of the alternatives. The procedures for generating the spatial attribute values (e.g., the average distance) follow the major functionality of GIS, where the relevant data are acquired and stored in GIS databases, and then the data are manipulated and analyzed (the overlay and proximity analyses) to obtain the values. These values can be considered as output of GIS-based data processing and analyzing. Given the individual preferences in the OWL ontology and the normalized attribute values of alternatives in the attribute value table file, the decision analysis component computes individual orderings by integrating the data on alternatives and individual preferences into an overall assessment of alternatives. Then, the component combines the set of individual orderings into a compromise (group) solution and visualizes them on the Google Maps.

15The user interface of the collaborative MC-SDSS (see http://collaborativesdss.com/​SDSS/​) consists of 6 tabs including: “Instruction”, “Individual preferences”, “Individual solution”, “Group solution”, “Map-based argument”, and “Questionnaire”. The system captures individual preferences, and geo-referenced comments and their corresponding geographic locations via user interface and store them into OWL ontology.

The GUI (Graphical User Interface) description

16Individual preferences: Each participant inputs his/her preferences regarding the relative importance (weight) of evaluation criteria (objectives/attributes). The individual assigns weights explicitly to the criteria by distributing 100 points, which is the sum of weights assigned to all of the criteria. The specification of criteria preferences is aided by visualizing them in hierarchical structure (see Figure 5). Given the individual of criterion weights, the system checks the consistency, compatibility and integrity of the weights. In general, two consistency rules are applied to the weights assigned by the individual: (1) the sum of the all the weights assigned to the objectives must be 100 and (2) the weight of an objective must be equal to the sum of its attribute weights. The system checks if the weights meet these rules, otherwise it highlights the error indicating the criterion names. This page also provides the capability allowing the users to make comments on a particular criterion by clicking its name. The participants may offer valuable information about the decision criteria used in the parking site selection. For example, some users may argue a criterion is not suitable for the parking site selection domain, while others may suggest that the same criterion should be included in the decision problem. They may propose to eliminate the redundant criteria, combine two or more criteria, or decompose a particular criterion into a number of attributes to facilitate the measurement process.

17Individual and group solution: Once the individual preferences have been specified, the system stores them in the ontology, and computes and represents the alternative orderings on the Google Map according to the individual preferences (see Figure 6). By clicking the “result”, the orderings can be visualized on the map, where the alternative locations with orderings are displayed using numeric Google Maps markers. Clicking each of the markers, the user is able to see the rating of the alternative. The map is dynamically updated in response to changes in the criterion weights. Similar to the individual orderings, the individual can examine the group rankings within the group solution tab, showing the score and ranking of each alternative location based on the preferences of all the participants who have finished the site selection procedure. Based on the individual rankings, the group rankings of alternatives is determined and visualized on the map (see Figure 7).

18Map-based argument: It is a map-centered communication/argumentation tool to support geo-referenced discussion (Rinner, 2008; Simao et al., 2009), where individuals can hold conversations in the form of posted messages on the map. This tool allows for graphical submission, compilation, tracking of geographic proposal via annotated map. Clicking on the map, the individuals reference their contributions about different dimensions of the decision problem to the geographic locations (see Figure 8). They can deliberate and exchange information regarding the parking decision problem using the Google Maps. This tool provides the users with the ability to share his/her opinion about the current alternatives, to propose inclusion of one or more particular location as a new feasible alternative or exclusion of alternative(s) from the option set, to collaborate on design and refinement of the alternatives, and to help give voice to social, health, environmental, economic, and safety concerns related to a particular place. The system retrieves all the geo-referenced posts of whole individuals from the ontology and represents them to the currently logged-in user on the map.

19Questionnaire: The questionnaire has been used for eliciting the individual’s opinions toward using the system for parking site selection problem in Tehran (see Figure 9). The questions provided in the questionnaire ask the participants to rate various characteristics of the system, reflecting how usable the system is. In other words, the answers to the questions provide the system usability evaluation by examining whether an application works and has met its design goals according to the user’s needs (Bugs et al., 2010). Six questions have been provided for the users to answer. The answers to the questions 1 to 5 are in the form of a 5-point rating scale. The last question asks the individuals to describe what they like the best and the least about the system.

Conclusion

20This paper has presented a Web 3.0-driven MC-SDSS for collaborative spatial decision making, exploring a prototype development of the collaborative MC-SDSS using the Web 2.0 and Semantic Web approach, in which the participatory GIS-MCDA knowledge has been organized in a formal ontology as the core of the Semantic Web. The approach adopts the ability of ontology to formally, semantically, hierarchically, and explicitly formulate the underlying knowledge required for collaborative GIS-MCDA in a machine-understandable way. The ontology integrates GIS, MCDA, and participatory decision making knowledge elements associated with the decision problem domain in a shared and unified framework. It contains individual/participant data including individual profile, criteria preferences (relative importance of criteria), geo-referenced arguments, and comments on criteria.

21The use of ontology for developing MC-SDSS offers some advantages over the conventional approaches including the capabilities of reusing and sharing of decision knowledge, automated reasoning and cooperating with an inference engine (Sadeghi-Niaraki, 2009). The collaborative GIS-MCDA knowledge representation using the ontological framework is machine-understandable. Consequently, it can be semantically and intelligently interpreted, processed, shared, and reused by the Web-based applications. It provides a semantic description of the concepts and relationships that can exist in a spatial decision problem domain, thus supporting interoperability between the heterogeneous GIS-MCDA sources and enabling GIS-MCDA applications to meaningfully communicate relevant knowledge. Also, the ontology provides reusable GIS-MCDA building blocks, which many specific GIS-MCDA applications can use as pre-developed knowledge modules. Furthermore, ontology can facilitate the collaboration by providing a unifying MCDA decision knowledge skeleton that can be used as a common and shared reference for a collaborative process (Gaŝević et al., 2009).

22The proposed Web 3.0-based MC-SDSS is flexible and easy-to-use. It allows the users to explore the decision problem in an interactive and recursive way, change and refine their preferences and perform alternative evaluation in a real-time manner. It enables participants to generate the orderings individually and then the individual orderings can be combined into a group ordering by means of the group choice rule. The system also supports geographically referenced arguments and deliberations by providing visual access to public geo-referenced comments and debates in the decision domain. A variety of arguments including suggestions for designing new alternatives, the proposals for eliminating one or more pre-defined alternative parking locations, identifying some concerns with respect to a particular location, etc. can be stored in the ontology.

23In spite of the advantages and new opportunities that Web 3.0-based MC-SDSS creates for facilitating public collaboration and interpreting, processing, communicating, sharing, and reusing the collaborative GIS-MCDA elements, there are some practical and methodological problems with this approach. One of the main challenges in using Semantic Web techniques for developing collaborative MC-SDSSs is in the area of ontology development. The major challenge in developing the GIS-MCDA ontology is knowledge acquisition and maintenance - the collection of concepts and relations in collaborative GIS-MCDA domain, achieving consensus on them among the domain experts and other interested parties, and frequent updates due to the dynamics of the knowledge structure of the domain and its unpredictable changes over time (Gaŝević et al., 2009). The other challenge is to provide facilities to access and reason information/knowledge in several ontology languages, allowing the construction and access to the collaborative GIS-MCDA knowledge independently of the native language of content providers and users.