Abstract
Effective knowledge management of highly sensitive information requires a meticulous design of access-control policies. These determine which organizational actors have access to what knowledge and under which circumstances. An effective design of such access-control policies requires not only a command of an apt representation formalism, but an in-depth understanding of the dynamic privacy needs of the organization. Acquiring these competencies is the central goal of any education or knowledge management process. We present a controlled experiment designed to examine the differences in novices’ competencies in using two ontology formalisms – Frames and OWL – while constructing access-control ontology-based policies. The two differ in the level of structuration, and the abstract thinking they require. The findings offer partial support for Bloom’s predictions. The results show that students performed relatively well in both formalisms with respect to the tasks of comprehending and implementing access-control policies. However, when it came to synthesizing new access-control policies, the students found the Frames formalism significantly simpler than the OWL where they failed miserably.
As of 2006,
Introduction
Access rules to information stored in Electronic Patient Records (EPR) requires an in-depth understanding of the dynamics of clinical practice, as well as the organization’s legal and ethical duty to protect patients’ privacy and guide against abuse of sensitive information [1, 2]. It is thus the responsibility of the privacy officer to develop an access control (AC for short) policies which restrict and hierarchize access to identifiable data of the EPR to anyone responsible for the diagnosis, treatment and continuity of care of the patient.
These AC policies should be flexible enough to allow for changes in the delegation of access if new circumstances arise. Such flexibility can be achieved by using an appropriate ontology formalisms (or languages). In this article, we consider two ontology formalisms: Frames [3] and Ontology Web Language (OWL for short) [4, 5], taught in a “knowledge representation and decision-support systems” course, with the intent of teaching students how to best maintain and extend a set of AC policies, formulated via these ontology formalisms (or languages), so to increase the patient privacy and the confidentiality of patient data, while being flexible enough to consider new cases (or scenarios) [6].
In recent years there has been a visible rise in the use of knowledge representation via ontologies in various domains (e.g., managing knowledge of build-to-order production network [7]). Alas, mastering ontology languages can prove to be rather challenging, particularly for inexperienced users, and especially when they are required to perform OWL-based tasks. The computer science literature has made it evident that for the vast majority of non-logicians, the existing OWL syntaxes are either too verbose, or too complicated to master. As a result, there are teachers who shun away from teaching OWL ontology to students [8]. While enlightening, these works tend to focus either on the manner in which these tools can be improved or made more user-friendly [8], or on measuring students’ comprehension of code writing [9]. Recent studies in the Interactive Sociotechnical Analysis (ISTA) approach [10] offer a more nuanced framework for studying system interaction. Their proposed socio-technical analysis specifies the important relationships among the technology, workflow, clinicians, and organizations in designing and implementing new technologies. The framework emphasizes the recursive and iterative nature of these relationships and their potential for producing unintended consequences[10].
An appropriate design and maintenance of AC policies reflects these mutual influences among the social subsystem (people, tasks, relationships), the technical subsystem (technologies, techniques, task performance methods, work settings), and their social and organizational environments. Moreover, system designers create policies to express conditions on the access to data. To reduce source clutter and improve maintenance, developers increasingly use domain-specific declarative languages to express these policies. In turn, administrators need to analyze policies relative to properties, and to understand the effect of policy changes even in the absence of properties [11]. Following the ISTA approach, we focused on novices’ actual use of two ontological languages, rather than uses that were planned or envisioned by designers or educators. Our first aim was to assess the level of cognitive difficulty posed by each of the ontological languages. The second was to test whether exposing students to two ontological languages, one notoriously simpler than the other (i.e., Frames), will improve the students’ overall performance in both languages, but particularly in OWL. The purpose was to devise a teaching method that would improve OWL’s teaching and learning, in a way that would allow novices to master these tools and use them in real-life settings. We thus devised a controlled experiment to examine how each of the ontological languages impact thinking in three cognitive levels: comprehending, implementing and synthesizing. Finally, we tested students’ high order thinking in both languages, to assess similarities and differences.
What is an ontology?
An ontology is a formal naming and definition of the types, properties, and interrelationships of the entities that interrelate in a particular domain of discourse [12]. Ontologies often include reasoning rules that support the processing of inferring new knowledge [13].
The SitBAC ontology
The SitBAC ontology enables the designing and implementing of access control policies in the healthcare domain [14]. The knowledge stored in the SitBAC ontology is comprised of “
On top of the SitBAC ontology, the SitBAC Framework (Fig. 1) [6] provides a real-time mechanism for processing new incoming data-access requests and infer an approve/deny response, in accordance with the data-requestor’s authorization to either view or edit data. The inference mechanism uses the Situations, (i.e., access rules) as an input for the decision making process, and is built on top of a Reasoner [16], which is a tool that enables classifying (i.e., mapping) new incoming data-access requests into existing ontology classes (in the described case, into Situations).
Frame-based ontology
Minsky [3] introduced the term Frame in the Knowledge Representation context. He defined a Frame as a data structure that can be used to represent classes with certain properties, called Slots. Frame was believed to simulate the way human beings store knowledge. The Frames language is considered as “easy-to-use” ontology formalism especially for novices, since it is comprised of a structured template that guides the users in their work (see Fig. 2). But while Frames offers suitable representation of the required knowledge, it does not support reasoning mechanisms, which are mandatory for the SitBAC framework. In other words, in the event of a new incoming data-access request, the SitBAC framework will not be able to generate an approve/deny output in real-time (Fig 1). To support reasoning, the Frame-based version of SitBAC requires adding additional knowledge in the form of rules (e.g., Jess rules [17]).
Web ontology language
OWL [4, 5] was introduced by the World Wide Web Consortium (W3C) as their standard ontology language. OWL is based on description logics [18]. It allows for a wider variety of semantic expressions. An OWL ontology [19] consists of Classes, Properties, and Individuals and can also contain Datatypes and Axioms (see Fig. 3). OWL supports the use of a Reasoner, which is, as mentioned before, a mandatory requirement of the SitBAC framework. In other words, in the event of a new incoming data-access request, SitBAC framework will generate an approve/deny output in real-time.
Building an ontology
The following describes a set of general steps that are used while building an ontology, with a focus on the SitBAC ontology: Identifying basic concepts relevant to the domain at hand. For instance, a data-requestor is a central concept in the healthcare domain, along with another central concept – the patient. Both are of SitBAC Entity type. The data-requestor has a role (e.g., physician), which is a concept as well, of SitBAC Refineable type. Identifying the relationships between the basic concepts (Entities and Refineables) defined above. For instance, we need to relate the “data-requestor’s location” to the “the patient’s location” via a SitBAC Relation type. Identifying and structuring complex concepts that represent real scenarios taken from the domain in question. In the SitBAC ontology it is expressed via Situations classes, hierarchizing constructs, which designate how these concepts should behave in a scenario of data access. For instance, a physician is allowed to view a patient’s Record who is hospitalized in the physician’s ward.
The first two steps are similar using either language (Frames or OWL). Both require users to identify all relevant concepts, award them with the correct properties (i.e., Refineables) and designate correct relationships between them. Alas the third step is performed differently using each of the languages. Whereas the Frames ontology is organized as a check-list of properties to be selected from structured menus, specified and defined, the OWL language is based on descriptions logic [18] which means that the user is required to write logical expressions designating each entities’ role and attributes and then stating the rules of their engagement using the connectors “and, or, not” along with “∀,∃” (i.e., universal and existential). So whereas the Frames ontology guides the user in the identification processes, the OWL ontology requires the user to demonstrate relatively high levels of proficiency in both the syntactic and semantic aspects of logical writing.
In addition, the OWL version of SitBAC uses Semantic Web Rule Language (SWRL) [20]. SWRL is based on OWL, and allows users to write rules to reason with OWL designated individuals and to infer new knowledge regarding these individuals. In SitBAC, SWRL is used to represent SitBAC Relations (see Fig. 4). SWRL is a member submission of W3C [20]. SWRL like OWL is logic-oriented: suppose one wants to represent in SWRL the following rule: “If a person has a male sibling, we can infer that he has a brother”. The corresponding SWRL rule is expressed as follows:
Cognitive processes using two ontological languages: Frames and OWL
Recent research in human-machine interface has focused more heavily on the cognitive processes of programmers and users, with the intent of enhancing programmers’ and users’ skills and competencies, as well as improving software’s usability and friendliness [21]. A model of programmer behavior must be able to account for five basic programming tasks: (a) composition: writing a program; (b) comprehension: understanding a given problem; (c) debugging: finding errors in a given program; (d) modification: altering a given program to fit a new task; and finally (e) learning: acquiring new skills and knowledge. While performing these tasks, programmers develop a complex multi-leveled body of knowledge that involves the appropriate use of concepts and techniques, as well as grammatical proficiency of the ontological language itself.
The distinction between syntax (sentence form) and semantics (sentence meaning) is fundamental to thinking about language. Grammar is perhaps the most distinctive feature of human languages; typically, we convey meaning in structured sentences. Hence, comprehension requires understanding not only individual words but also the syntactic frame in which the words are embedded [22].
Another part of that knowledge, called semantic knowledge, consists of general programming concepts that are independent of specific programming languages. A relatively higher level of semantic knowledge is required to solve problems in application areas such as statistical analysis of numerical data, stylistic analysis of textual data, or designing an Access Control model for an EPR system. Syntactic knowledge involves details concerning the format of iteration, conditional or assignment statements, valid character sets; or the names of library functions. It deals more with grammer and less with abstract concepts related to the problem at hand. It is therefore more precise, detailed, and arbitrary than semantic knowledge, which is generalizable over many different syntactic representations. The semantic knowledge is acquired largely through intellectually demanding, meaningful learning, including problem solving expository instruction. It thus encourages the learner to “anchor” or “assimilate” new concepts within existing semantic knowledge or “ideational structure.
Evidently, different ontological languages affect the syntactic and semantic representation of knowledge in different ways, so to demand different competencies of their users. Our study compared students’ competencies in using two types of languages – Frames and OWL.
Rector and Stevens [23, 24] have held more than a dozen workshops in which they have taught the OWL language to students. They identified two major classes of difficulties encountered by newcomers to OWL. The first are difficulties in understanding the exact meaning of OWL expressions due to its pedantic language [23]. These result in errors in understanding common logical constructs; mistaken use of “∀” (i.e., universal), rather than “∃” (i.e., existential) restrictions; erroneously thinking that “only” (allValuesFrom) implies “some” (someValuesFrom); and finally difficulties in understanding subclass axioms used for implication. The second class of difficulties stems from users’ overreliance on structured templates rather than open-world systems [23].
Since each of these ontological languages requires different cognitive levels of complexity, we asked, as educators, how can the teaching of these two languages improve the students’ proficiency levels? Benjamin S. Bloom developed an approach called “mastery learning [25] 1982), which today offers the basis for the competency-based education model [26]. Bloom’s taxonomy is a six-level classification system that uses observed student behaviors to infer the level of students achievement. Moving from simple to more complex, the taxonomy provides a framework for distinguishing between six categories in the cognitive domain. These are: knowledge, comprehension, application, analysis, synthesis, and evaluation. The categories are ordered from simple to complex and from concrete to abstract. The taxonomy also posits that mastery of each simpler category is a prerequisite to mastery of the next complex one [27]. The knowledge category refers to the ability to retrieve relevant knowledge from long-term memory (i.e., recognizing and recalling pieces of information). The comprehending category refers to the ability to determine the meaning of instructional messages, including oral, written, and graphic communication. It entails interpreting, translating or explaining pieces of information. Implementing refers to the ability to carry out or using a procedure in a given situation. The analyzing category refers to the ability of breaking material into its constituent parts and understanding how the parts relate to one another and to an overall structure or purpose. It requires one to differentiate, organize, and attribute. Finally, the synthesis category refers to the ability to put elements together to form a novel, coherent whole or make an original product [27].
The proposed use of the taxonomy, suggests ways to encourage students to increase the complexity of their critical-thinking skills and mastering of the more abstract tasks in code writing. Despite its usefulness, Bloom’s taxonomy, has been criticized for its lack of sufficient theory and validation. His critics argue that the taxonomy’s level is not always distinct, and that the underlying structural principle— increasing complexity— is naïve.
In the domain of computer science, Bloom’s taxonomy has been applied to guide course design and evaluation, structuring assessments and comparing the cognitive difficulty level of computer science courses. It also provides educators a friendly roadmap with which to evaluate students’ acquisition of new skills. In recent years, attempts have been made to relate Bloom to specific computer programming tasks. More specifically, to classify typical programming and software engineering tasks [9]. Following Bloom’s lead, we were able to design tasks to evaluate the students’ proficiency at three cognitive levels: comprehending, implementing and synthesizing. By so doing, we were hoping to explain students’ difficulties in mastering the tasks that demanded relatively high levels of thinking. We also contribute to current discussion in the literature by empirically testing Bloom’s predictions.
Research hypotheses
Reynares et al. [28] and Schwitter et al. [29] studied how different formalisms could potentially help users to first conceptualize the statements that they later formulated in OWL. Both studies found no difference in the performance of users who used different formalisms in the conceptualization stage. Yet, they did not examine, whether the learning of two languages will improve the overall performance in both. Following their lead, we devised a controlled study to test how students perform on tasks requiring three levels of learning, was devised. First, students’ comprehension of the two ontological languages, as well as their implementing and synthesizing capabilities, was measured. A set of questions that lay within the ‘comprehend’, ‘implement’ and ‘synthesize’ sub-categories of the cognitive dimension of the revised Bloom’s taxonomy, was devised.
We hypothesized the following: The overall performance in the Frames tasks will be higher than in the overall performance in the OWL tasks. The order in which the homework assignments is completed affects the students’ performance in the subsequent homework assignment. More specifically, the order in which the given set of questions is completed affects the students’ performance in the subsequent set of questions. In both, Frames and OWL ontologies, we expect to find a strong and positive relation between the scores in the relatively simpler tasks and the relatively complex tasks, so that the higher the score in the comprehending task, the higher will be the scores in the implementing and synthesizing tasks.
Research design
A controlled experiment was carried out as a crossover study, where a student’s performance on a task using Frames could be compared to the same student’s performance on the same task using OWL. Theoretically, this allows estimating the effect of OWL vs. Frames with greater precision as compared to parallel group experiments. In a randomized controlled crossover experiment with repeated measures, the subjects are randomly assigned to a different sequence of exposures and the same measures are collected multiple times for each subject. Usually, in a crossover design, all subjects receive the same number of exposures. In the described case, a semi-randomized controlled crossover experiment with repeated measures was carried out. The students were paired, and then assigned to group A and B respectively. The pair matching was done with respect to relevant extraneous variables such as: a) the number of semesters the students have studied; b) their current grade-point average (GPA); c) their achievements in the course thus far, and their class attendancerecord.
The experiment took place during 2013 as part of the course “knowledge representation and decision-support systems”, which is an advanced elective course in the Information Systems program at the University of Haifa. 32 students took part in the experiment, 12 graduate students and 20 undergraduate students in their last year of their studies. As noted, the students were divided into matched pairs, so that one student of each pair was associated with group A, while the other pair student was associated with group B. Each group comprised of 8 female students and 8 male students. The average grade for group A was M = 80.29, sd = 7.9 and for group B M = 79.9; sd = 7.1.
The course in question covered knowledge-representation methods including Frames and OWL ontologies. The SitBAC ontology with its two implementations (Frames and OWL) were presented to the class in two sessions, at each the students got a Protégé project file comprised of the SitBAC ontology, along with a set of Situations to be considered. In addition, they were given a detailed guideline document (one for each formalism: Frames and OWL) containing a full description of how to maintain and extend the SitBAC ontology, starting with adding a new required concept to the ontology (i.e., of type Entity, Refineable, or Relation), followed by orders for creating SWRL rules, and ending with a set of instructions for creating a new ontology Situation. The students were requested to use the guidelines throughout the SitBAC sessions and while preparing their homework. The instructor emphasized the importance of using the guidelines while working on the ontologies during both SitBAC sessions. Moreover, these sessions were carried out in a computer lab, thus allowing the students hands-on practice of every step of the work, with the presence of an instructor to guide them. During these sessions, the students practiced 1)
The students were given two homework assignments:
In accordance with Bloom’s suggested taxonomy, each HW contained three questions ordered in a raising order of complexity.
Comprehending question. The first question required the students to translate to natural language (Hebrew), three Situations represented via Frames ontology or OWL ontology. This task required the students’ a passive knowledge of both syntax (sentence form) and semantics (sentence meaning) of the ontology. In other words, the students were expected to express their comprehending by “translating” the Situations into Hebrew.
Implementing question. The second question presented the students with descriptions of three new data-access scenarios (written in Hebrew). The students were asked to create the appropriate ontology Situations for each given scenario. It is noteworthy that all required concepts for the appropriate Situations were already defined in the ontology. (E.g., with respect to the Situation presented in Fig. 2, possible existing concepts include Physician for Role and inPatient for patient’s Status.). This task required a higher level of both synthetic and semantic knowledge. They had to recognize the relevant concepts, their place in the hierarchical structure, their attributes and relations to other concepts. Then they had to create the appropriate Situations correctly in the ontology.
Synthesizing question. In the third and final question, the students were presented with descriptions of three new data-access scenarios and were asked to create for each of these scenarios an appropriate ontology Situation. However, this time, several concepts, essential for completing the task, were missing from the ontology. The students were expected to acknowledge that they are missing, create them, place them correctly in the ontology, and then author the appropriate new Situations. For example, one of the three scenarios described a data-requestor with the role of lab technician. This role did not exist in the ontology. Thus, the students had to create a Lab-technician concept and place it in the Role hierarchy in the ontology, prior to the creation of the new Situation in question. In another scenario, the students were required to redefine a SWRL rule in a case of a new SitBAC Relation. This task required the students not only to be able to break material into its constituent parts and understand how these parts relate to one another, but to generate new concepts. In the OWL homework assignment such abilities were heavily depended on the students’ proficiency in logical thinking.
Working on a scenario consists of a collection of
Students’ performance was measured in the following manner: Every student had to complete two homework assignments (HW1 and HW2). The maximum score, each student could have potentially scored was 90 and the minimum was 0. Each homework assignment consisted of three questions. The maximum that the students’ could have scored in the comprehending task (no matter what ontology language they used) was 29, in the implementing task, 50, and in the synthesizing 11 (see Table 1). All scores were normalized to yield a final score for each student.
Results
First hypothesis
We hypothesized that the overall performance in the Frames tasks will be higher than in the overall performance in the OWL tasks. The hypothesis was confirmed (see Fig. 5). Students’ scores in the Frame’s homework assignment were compared to Students’ scores in OWL homework assignment. Student’s mean score in the Frames homework assignment was M = 81.16, sd = 5.64. While the mean score in the OWL homework assignment was M = 75.75, Sd = 7.65. T-test revealed the difference between the two homework assignment to be significant [p < 0.001; t(31) = 3.2166].
Second hypothesis
The next step was to test whether the order in which the homework assignments were completed affected the students’ performance in the subsequent homework assignment. Bloom [25] posits that mastery of each simpler task is a prerequisite for mastering the next more complicated task [27]. Following his logic, we assumed that a relative high score in the Frames homework assignment will be followed by a relatively high score in the OWL homework assignment. We also predicted that a relatively high score in the OWL homework assignment will be followed by a relatively high score in the Frames homework assignment.
Students’ mean scores in Frames 1 and Frames 2 were similar irrespective of the order in which the homework assignments were carried out (see Fig. 8). The same is true for the OWL homework assignments (see Fig. 9). This finding indicates that there was not a spill over affect between the Frames and OWL homework assignments. Performances in the Frames homework assignments remained higher than performances in the OWL homework assignments throughout the experiment (see Fig. 6 and Fig. 7). In group A, the difference between the two assignments was found to be significant [p < 0.05; t(15) = 2.0406)], and so was found in group B [p < 0.01; t(15) = 3.4069)].
This finding can be explained by the fact that some of the easy-to-use features of Frame-based editors are not at all compatible with the OWL syntax or semantics, as Wang et al. point out [30]. Nevertheless, we proceeded to testing for differences in performance between students of different GPA scores. The students were divided into two groups based on their GPA scores. 60 to 80 comprised the ‘low’ achievers group (N = 15), and students’ whose GPA ranged between 81to100 comprised the ‘high’ achievers (N = 14). T-test confirmed slight differences between achievers and non-achievers in several of the tasks. In Frames, low achievers’ average score in the comprehending task was M = 88.28, sd = 0.11, compared with the ‘high’ achievers whose average score was M = 94.33; sd = 0.65). [P < 0.05; t(27) = –1.73)]. Slight differences were also found in the Frames synthesizing task. The ‘low’ achievers scored M = 68; sd = 0.8, and the ‘high’ achievers scored M = 74.78; sd = 0.87 [p < 0.05, t(27) = –0.43]. Slight differences were also found in OWL implementation task, where ‘low achievers’ scored M = 86.80; sd = 0.15; and ‘high achievers’ scored M = 92.14; sd = 0.06 [P < 0.05; t(27) = 1.22]. Differences were also found in the OWL synthesizing task. ‘Low’ achievers scored M = 22.42; sd = 0.17; and ‘high’ achievers’ scored M = 38.26; sd = 0.23 [P < 0.05; t(27) = 2.16].
Third hypothesis
Interested in the manner in which the students’ performed within each homework assignment, a correlation analysis to quantify the direction and strength of the linear association between the three task variables: comprehending, implementation and synthesizing, for each of the Frames and OWL homework assignments, was performed. Based on Bloom’s suggestions, we predicted that a strong and positive correlation between the three variables so that higher scores in the comprehending task will be associated with higher scores in the implementation and synthesizing tasks for both Frames and OWL. With regards to the Frames ontology, the findings reveal strong and positive correlations between comprehending and synthesizing, as well as between implementing and synthesizing. However, with regards to the OWL ontology, no such correlations were found (see Tables 3 and 4 respectively).
To assess differences in levels of complexity within each homework assignment, we performed a general linear model for repeated measures for Frames and OWL separately. In frames, the general linear model indicated significant differences between the comprehending task and the synthesizing task and between the implementing task and the synthesizing task (see Table 5).
In the OWL assignment, the general linear model indicated significant differences were found between every task to any other.
With regards to the Frames homework assignment, students’ scores in the comprehending (M = 91.38, sd = 0.95), and implementing questions (M = 93.8, sd = 0.62) remained relatively high and relatively similar. A considerable drop in the average scores was notable in the synthesizing task (M = 70.17; sd = 0.135), but the average scores remained relatively high. In this respect, the evidence supports Bloom’s prediction that mastering tasks of a lower thinking order is a perquisite for mastering tasks of a higher thinking order. Alas, let us remember that the Frames ontology is relatively structured. It consists of built in templates that guide the user in his efforts to designate concepts, attributes and relations. In this respect, the Frames ontology supports a scaffolding model of learning thus providing the novices the support needed for developing higher levels of cognitive thinking.
In the OWL homework assignments, students’ average scores dropped significantly with each growing level of complexity (comprehending, M = 94.61; sd = 0.67; implementing, M = 89.81; sd = 114; synthesizing, M = 30.97; sd = 0.216). This drop in scores, indicates a relatively sharp rise in complexity – one that perhaps indicates that mastering a relatively lower level of thinking is of little help in mastering a higher level of thinking (see Table 6).
The students participating in our study found the synthesizing task to be the most challenging in both ontological languages. However, in the OWL the synthesizing question proved even more challenging, perhaps because it required relatively high proficiency of both synthetics and semantics knowledge. For novices, under trained in logical thinking, generating new concepts and authoring the SWRL rules were indeed daunting tasks, one that was not mitigated by mastering the comprehending and implementing tasks. In this respect, our evidence offer limited support for Bloom’s prediction of mastery. When we analyzed the mistakes that the students’ made in the synthesizing task we uncovered the following errors. 1. Misplacing a newly-created class in the class hierarchy – an error that could indicate a lack of understanding of the SitBAC domain 2. Failing to update the covering axiom after adding a new class – an error that could indicate a lack of understanding of OWL semantics. 3. Erroneously disjointing designated classes. 4. Erroneously mapping a required Relation individual to the wrong Relation class in the ontology, or, incorrectly defining the respected SWRL rule. These errors indicate difficulties in fully mastering the OWL and SWRL semantics specific difficulties [23, 24] were also found in using Frames for complex synthesizing actions. For example, 1. Choosing the wrong Relation Template (e.g., Location_Relation instead of Organization_Relation). 2. Choosing the wrong hasRelation instance (e.g., isEqualTo instead of isIn). However, these errors are fewer than the ones found in using OWL, and are not strongly related to mastering the Frames ontology. Rather they are related to misunderstanding the particular SitBAC ontology and the logic operators.
Discussion
García and Vañó designate “knowledge management” as one of the specific challenges that business management will have to face in this century. Moreover, they claim that knowledge management is fundamental for organizational learning [31]. Designing a proper model for authorization and access control for electronic patient record (EPR) is essential to a wide scale use of EPR in large health organizations. Both flexibility and precision are essential perquisites for designing an access control policy that both protects the patient privacy and the confidentiality of patient data, while efficiently considering new cases [32, 33]. Specifying the access conditions and privileges for an EPR user is still a difficult task, especially given the urgency of data-access in the clinical setting. It is therefore important for students to familiarize themselves with a wide variety of conceptual tools that would allow them to pick and choose the most suitable solution.
This article reports on findings from a controlled experiment designed to study students’ performances using Frames and OWL ontological languages, while maintaining and creating data-access rules via SitBAC ontology. Inspired by Bloom’s educational approach, we devised a controlled experiment to test aspects of higher order thinking. These are fundamental not only to the mastering of ontological languages, but also to the development of students’ skills as private officers, and system analyzers.
Our findings offer partial support for Bloom’s predictions. The results show that students performed well in both formalisms with respect to the tasks of comprehending and implementing Situations. However, when it came to synthesizing new Situations (i.e., new organizational AC policies), the students found the Frames formalism significantly simpler than the OWL where they failed miserably. Despite undergoing a similar and detailed preparation in class, the students performed relatively well in the Frames synthesizing task, while failing in the corresponding OWL synthesizing task. Interestingly, both ‘high’ achievers and ‘low achievers’ managed the Frames synthesizing task, though with differing levels of success. However, in the parallel OWL task, both ‘high’ achievers and ‘low achievers’ failed.
The teaching method we employed proved more effective in the Frames ontology language than the OWL. Designed to accommodate the scaffolding model suggested in Bloom’s taxonomy, the Frames ontology proved much simpler for novices to master, then the OWL which does not. Even so, we contend that organizing the course’s curriculum and homework assignments in accordance with Blooms’ taxonomy can help both individual students and educators to determine the level of their work. This self-analysis, then, will allow students to support their own higher level thinking. As Athanassiou and his colleagues [26] suggest, students respond positively to instructor emphasis of Bloom’s taxonomy of cognitive development and subsequent instructor evaluation using the taxonomy. To improve students’ performance in the OWL ontological language, it is advisable to strengthen their logical thinking. We believe that a firmer basis in logical thinking would improve students’ performances, as they will be better equipped to handle the rather abstract semantics of the OWL language.
The described study was performed on a limited number of students, and though offers strong evidence as to the differences in students’ performances using the two ontological languages, could benefit from both a repetition of the study and a follow up study to determine students’ performances in a more mature stage of their educational career. In focusing on the cognitive competencies of novices in relations to the features of both ontologies, we offer both educators’ and practitioners’ insights as to how to improve novices’ performances in using both these tools. Developing novices’ nuanced understanding of the ontology’s syntax will not only improve the flexibility and sensibility of the control access policies, but will better attune them to the needs, urgencies and dilemmas of the medical environment in which these policies are implemented.
Footnotes
Acknowledgments
We would like to express our appreciation and to thank Prof. Mor Peleg for her guidance, support and assistance during the research execution and while writing the article.