Automated instructional design for CSCL: A hierarchical task network planning approach
Introduction
The design and orchestration of effective and well thought out Collaborative Learning (CL) scenarios is a challenge that the Computer Supported Collaborative Learning (CSCL) community has faced for several decades (Dillenbourg, 2002; Kobbe et al., 2007). The goal of creating a scenario in the context of group learning is to properly structure the interactions among peers, thereby increasing the learning gains of individuals. Such a scenario, defined as a Unit of Learning (UoL), is a delimited piece of education or training, such as a course, module, or lesson, in which the elements describing “what should to be taught” are structured according to different pedagogical approaches to define “the way in which participants (students and teachers) should interact.”
Development, adaptation, and customization of UoLs for CSCL requires careful instructional design (ID). Thus, some researchers developed authoring tools that use diagrams of graphs (e.g., Cool Modes (Pinkwart, 2003)) and CSCL Design Script Patterns (e.g., COLLAGE (Hernández-Leo et al., 2006), and CHOCOLATO (Isotani et al., 2010)) to better support the ID. However, the authorship of UoLs for CSCL with these tools is difficult if the goal is to develop, adapt, or customize units taking into account the individual characteristics of students. This is because the designer must decide what activities should be defined, what groups of students should be formed, what learning goals must be achieved, what roles must be played, what tools and materials must be used, and so on.
For example, to develop a UoL in which four students acquire knowledge about the mathematical concept “derivative” through exercises, an instructional designer (in the analysis phase) identifies what will occur in the CL scenario by setting instructional/learning goals. In setting of these goals, the designer defines the purpose of elements to be obtained as the desired stages of learning development (individual goals), skills, and attitudes to be achieved when the run of UoL is finished. In this example, the designer sets “the individual goals for all student as the acquisition of knowledge about derivative at level restructuring”, and “the acquisition of attitude positive interdependence” as the purpose of UoL (instructional/learning goals).
Next, the designer (in the design phase) defines how the learners attain these goals through the group formation, the selection of tools and learning materials for students, and the definition of learning plans. In the group formation, the designer may want two groups with two students or one group with four students. This decision depends on the availability of the students, teacher or the characteristics of individual students (e.g., stage of learning development related to the concept “derivative”). In the selection of learning material, the designer performs finding the proper exercise in external repositories (e.g., exercises that none of the students have previously seen). To define the learning plans in the CL scenario, the designer employs an authoring tool that uses CSCL Design Script Patterns to decide how to apply any pattern (e.g., COLLAGE or CHOCOLATO). For example, the designer applies the “jigsaw” pattern if the selected exercise can be divided into two parts and the students have experience in CL, the “pyramid” pattern if the selected exercise is difficult and the students don’t have experience in CL, or the “distributed cognition” pattern if the students have experience using the knowledge related to “derivate” or they have cognitive skills. Alternatively, the designer can decide to define the learning plan by employing an authoring tool that uses diagrams of graphs (e.g., Cool Modes).
Finally, the designer (in the development phase) arranges the concrete learning plans and develops all media and any supporting documentation that will be used in the CL scenario. The designer must define roles according to the ability of a student to perform a role and the behavior the role a player performs. In this example, if the designer decides to use the pattern “distributed cognition,” for a group with all students, he will set roles “instructor” and “learner”, and he will define of transmission/reception messages for each student. At this stage of the scenario it becomes clear what learners should participate, what roles they play, what behavior they perform, what learning materials they learn, and what educational benefits they are expected to acquire. Next, the designer will develop learning environments to enable interaction among students in the CL scenario. The learning environment is represented as an arrangement of tools for the members of the group. In this example, the designer can define an environment in which the student with the role “learner” will use a tool “simulation” to participate in the exercise, and the student with the role “instructor” will use the tool “monitoring learning process.” The tool “making a report paper” is set for all of the students, the tool “discussion channel” (like a chat system) is set for the student with the role “learner,” and the tool “monitoring discussion” is set for the student with the role “instructor.”
In Artificial Intelligence, the automated planning field studies the automatic generation of action sequences, called plans. The execution of plans leads to the satisfaction of certain goals. Therefore, there is a direct relationship between ID and automated planning, in which plans are used to generate a UoL that defines an individualized teaching–learning process for CSCL. In automated planning, HTN planning is a technique that uses hierarchical tasks and methods to represent domain-specific strategies. The methods define multiple ways to deconstruct the tasks into sub-tasks until getting a consistent and coherent plan. The ID of UoLs using CSCL Design Script Patterns can be documented as hierarchical tasks and methods. This knowledge focused on the rationale of the ID process of elements to be included into UoLs.
In this paper, we present a model that formalizes the ID for CSCL as a HTN planning approach. In the second section of the paper, we present basic information about ID using CSCL Script Design Patterns. In the third section, a brief description of HTN Planning is summarized. In the fourth section, we present our approach as a model that formalizes ID for CSCL as HTN planning. In the fifth section, we present an example that shows the formulation of a planning problem and its results. In the sixth section, we present related works and compare them with the results obtained by this research. Finally, we present conclusions and discuss future work.
Section snippets
Instructional design using CSCL Script Design Patterns
CSCL represents a multidisciplinary paradigm within Technology-Enhanced Learning, in which computers are employed to enhance various educational aspects of group learning (Stahl et al., 2006). One of the main concerns for CSCL is how to improve social interactions, an essential element of group learning. The goal in this context is to increase the probability of reaching success in CL scenarios by providing students with a set of instructions that promote fruitful collaboration through learning
Hierarchical task network planning
In HTN planning, the generation of action sequences (called plans) that leads to the satisfaction of certain goals by agents is represented as sets of tasks (task network), and methods that decompose non-primitive tasks into sub-tasks until reaching a level of primitive tasks which can be solved by operators (Garrido et al., 2008, Ghallab et al., 2004). Our courseware web service uses the HTN planner JShop2ip (Java Simple Hierarchical Planner Order 2 for Instructional Planning), a version of
Instructional design for cscl as htn planning
Fig. 6 shows our approach to automated ID for CSCL as HTN planning. This model is based upon the classical ITS model (Kaplan and Rock, 1995; Woolf, 2010). Where the student model and the domain model are defined in external repositories, the pedagogical model is defined in planning domain definition as ID strategies. To start the ID, the designer (through an authoring tool) defines the initial CSCL scenario as a planning problem definition, and (optionally) can also define the information of
Research application
To evaluate the validity of our approach, we developed a CSCL courseware generator employing the reference model detailed in previous section. This courseware generator obtains personalized UoLs for CSCL. These units are adapted from the students’ characteristics through of a set of plans of sections, plans of practice sessions, plans of learning sessions, and plans of interactions. The plans of sections are obtained through the sequencing of learning goals, the selection of a pattern related
Related works
Since the first approach for integration of automated planning in ITSs many researches employ different techniques to obtain plans tailored to each student (Peachey and McCalla, 1986). For example, the Generic Tutoring Environment (GTE) developed by Van Marcke (1998) allows to declaratively define the pedagogical model as a set of tasks and methods, and them use it for the planning process. The pedagogical model defined in GTE was evaluated and utilized on different ITSs such as NOBIT
Conclusion and future research
The design of flexible and personalized well thought out collaborative learning (CL) scenarios (i.e. UoLs) is a complex task due to the various variables that need to be considered during the designing process. To address this challenge, we proposed an approach to formally model ID using HTN planning techniques. We established a set of guidelines that convert the knowledge of ID such as CSCL scripts, design patters and best practices into HTN planning knowledge. Thus, we can use the HTN
Acknowledgment
We thank the CNPq (Process: 400481/2013-8, 310204/2011-9 and 550449/2011-6) for providing support for this research. The authors also wish to express their gratitude to the reviewers and editors of Expert Systems with Applications who spent their time and effort to improve the quality of this work.
References (43)
- et al.
Planning and scheduling in an e-learning environment. A constraint-programming-based approach
Engineering Applications of Artificial Intelligence
(2008) - et al.
A semantic web-based authoring tool to facilitate the planning of collaborative learning scenarios compliant with learning theories
Computers & Education
(2013) - et al.
Using planning techniques in intelligent tutoring systems
International Journal of Man–Machine Studies
(1986) - et al.
Automating educational processes implementation by means of an ontological framework
Computer Standards & Interfaces
(2014) - et al.
Incorporating assessment in a pattern-based design process for CSCL scripts
Computers in Human Behavior
(2009) - et al.
Web Collage: An implementation of support for assessment design in CSCL macro-scripts
Computers & Education
(2013) - Dillenbourg, P. (2002). Over-scripting CSCL: The risks of blending collaborative learning with instructional design. In...
- et al.
Designing integrative scripts
- Erol, N. (1990). Epos collaborative learning environment. DELTA Project EPOS D7002 1C 90–02-02, ECOLE, Milano:...
- Ghallab, M., Nau, D., & Traverso, P. (2004). Automated planning: theory & practice. Access online via...
Ontological support for a theory-eclectic approach to instructional and learning design
Ims learning design support for the formalization of collaborative learning patterns
Generating cscl scripts: From a conceptual model of pattern languages to the design of real scripts
E-learning Design Patterns Book
COLLAGE, a collaborative learning design editor based on patterns
Journal of Educational Technology and Society
Deployment of ontologies for an effective design of collaborative learning scenarios
Cited by (5)
Hierarchy aware distributed plan execution monitoring
2016, Expert Systems with ApplicationsCitation Excerpt :In relation to the analysis of green supply chain management (GSCM) under uncertainty carried out by Tseng et al. (2014) relative to close-loop and open hierarchical structures, our work employs probabilistic modeling of hierarchical interactions among asymmetrically clustered participants based on communicating Markov Decision Processes instead of relying on fuzzy set theory and analytical network process (ANP) which can be affected by decision makers subjectivity. With respect to the Hierarchical Task Network (HTN) planning approach proposed by Challco et al. (2014) to optimize tasks for instructional designers in Computer Supported Collaborative Learning (CSCL) based on units of learning, our work focuses on hierarchical sharing of knowledge derived from event based units of information as part of monitoring an existing plan being executed. Compared to the work of Quispe and Galan (2014) on mobile ad hoc networks (MANETs) routing protocols which highlights the benefits of hierarchical routing strategies, our paper also leverages hierarchical information gathering and forwarding.
A systematic review of ai applications in computer-supported collaborative learning in STEM education
2022, Artificial Intelligence in STEM Education: The Paradigmatic Shifts in Research, Education, and TechnologyComputer-based systems for automating instructional design of collaborative learning scenarios: A systematic literature review
2016, International Journal of Knowledge and LearningFuzzy logic-based modeling in collaborative and blended learning
2015, Fuzzy Logic-Based Modeling in Collaborative and Blended LearningA Multi-Task combinatorial optimization model based on genetic algorithm and its application in college education curriculum planning
2015, International Journal of Emerging Technologies in Learning