Improving the Interface of Empirical Modelling Tools - Dissertation Example

Introduction Empirical Modeling (EM) is an ongoing research method originally initiated in the early 1980s by Meurig Beynon at the of Warwick. The early stages of EM research include ARCA, a definitive notation for interactive Cayley Diagrams [1]. ARCA is a programming notation intended for interactive specification and manipulation of combinatorial graphs. It incorporates variables for denoting primitive data elements and diagrams [2]. The research led to the development of the Evaluator of Definitive Notations (EDEN). In 1987 Edward Yung introduced the notion of Modelling With Definitive Scripts (MWDS). Experiences developing models with the principle of EM exposed many limitations of traditional construal computation, as they were found to be insufficient to capture the richness of the activity. Hence, principles differentiating the work from the mainstream were introduced. These principles have influenced the development and interpretation of tools and models through the present day. The term ‘state-as-experienced’ distinguishes the state in EM from the notion of state as understood in mainstream computing fields. A construal is a particular understanding of a situation; it is always subject to future re-interpretation and alteration. Three elements represent the EM principle in developing models: observables, dependency and agency. Tools developed for building EM models are tightly related to these three concepts. Several other tools have also been developed to support the environment of Empirical Modeling activities. Most of them were concentrated on mapping the three principles mentioned above to achieve the goal. Due to the differences between EM activity and mainstream computer modeling, traditional programming is insufficient to support the definitive notation environment of EM. Hence, there are significant amounts of work involved in developing the environment with traditional programming. Tools like tkeden, dtkeden, and webeden have shown their ability and usability in developing EM models. Most of them use very basic interfaces, as insufficient work has been put into developing a user-friendly interface along with good functionality and environment. The aim of this research is to review the tools used in supporting EM activity throughout the research of Empirical Modelling. The research examines the weaknesses related to the user interfaces of each tool and improves the user friendliness of EM tools, providing more easily accessible information regarding the models current state. The research will mainly focus on improving the interface of tkeden. A supporting tools modeller is developed to have more insight into the current state of the models. For example, in current versions of tkeden there are no real time displays for the definition of observables and dependency. Digging into the source code can only spot agency representation. It requires a massive amount of time to understand a large scale model built by others in current versions of tkeden because dependencies and agencies have to be identified individually from a source code that may contains tens of thousands of lines of codes. Related Work Throughout the Empirical Modelling research several tools have been developed to provide a definitive notation environment for developing models. Each of them serves different purpose, but most are based on the EDEN language. The most well known tool used in EM research is tkeden; many models have been built using tkeden. It is a definitive system that can manage definitions based on the EDEN language and is implemented with tk/tcl. Tkeden has had massive contributions throughout EM research and led to many variations based on EDEN. Abstract Definitive Machine Meurig Beynon, Mike Slade and Edward Yung designed the Abstract Definitive Machine (ADM) in 1988. It is designed as a computational model that is based upon definitive representations of state. ADM is a tool for animating the LSD specifications to represent concurrent action by parallel redefinition and reflects context dependence of agent actions by the use of scripts. Cadence Cadence is an environment that is based on the DOSTE interpreter. Candence IDE, the Warwick Games Design (WGD) library and EDEN are examples of modules that can communicate with DOSTE. DOSTE itself is conceived as a special kind of operating system, and has been deployed in this way [3]. In Cadence, DOSTE captures current state via a combinatorial graph: a family of nodes connected by directed edges. There are also labels nodes in the graph. The DOSTE state is constantly being updated, so the structure of the graph is subject to change. A node can be interpreted both as an observable, and as an observational context for other observables. Cadence is still an ongoing research project, and more functionality is under development. One such example is networking extensions based on distributed web enabled architectures. Although Cadence has more improved interfaces compare to tkeden, it still lacks the functionality for modeller to actually see the graph of nodes representing current states of DOSTE. Web Eden In recent research several versions of web-enabled EDEN tools have been developed. Web Eden is a tool to support Empirical Modelling involving interactive environments to mediate understandings of an external domain. It introduced a new paradigm for open source development that blends with the learning experience, making it possible for server and client machines to share the computational load in interpreting a model. It also overcomes the problems of efficiently interpreting many EDEN models concurrently by enabling distributed processing and load-balancing over many EDEN virtual machines. The frontend interface is developed using Flash and pushing interactions to the backend via an XML protocol. EDEN code and dependency maintenance are carried out in a modified instance of TkEden that communicates changes with the frontend user interface. While Web Eden was a big step forward for distributed definitive systems, it has several drawbacks. The EDEN models are sent to the server for interpretation, so that load-balancing between a network of EDEN engines is not implemented. It also relies on Adobe Flash for visualization, which leads to cross-platform issues [4]. Js-Eden Js-Eden is a recent development of a browser based EDEN interpreter. Js-eden implements the EDEN engine with JavaScript. Presenting the user-interface using state-of-art web technology provides a user-friendly interface. It contains a list of observables and definitions for each of these observables. This provides easier access to finding the observables in large-scale models. Looking at the observable table easily identifies dependencies and agencies. Any changes to observables will reflect in the user interface immediately. Although dependencies still have to be manually identified, it is a great improvement compared to other variation of EDEN interpreters. The Dependency Modelling Tool (DMT) DMT is a tool for graphical display of dependencies between EDEN observables. It was developed under the Java environment, and takes file definitions of observables recorded in the model and visualized dependencies of each observable provided. It is a useful tool to present the visualization of dependencies in the model, yet it still required manual input of the observables definition. Hence it does not provide any real time update for the change of dependencies in the model. Research Process and Methods Several tools have been accessed and used throughout the course of CS405. Tkeden was the main tools used, as there are many models and resources available which were originally built using tkeden. It has been proven that tkeden is a very successful definitive system to support the principle of Empirical Modelling. But the drawbacks are also significant, and the major problem of tkeden is the user interface. It only contains a very simple console allowing users to input their definitions. Still, users can search for previous inputs by going through the history. There are no records that can save any codes, so users have to use n separate text editor to save any import codes. The efficiency of constructing a model has been affected significantly. Definitions and dependencies of observables cannot be identified easily. Series of commands are required to retrieve the definitions of an observables and its even harder to identify dependencies between observables. There are also several external tools that have been briefly reviewed. For example, GeoGebra is software for learning and teaching mathematics. The idea is very similar to the definitive system. Models built using Geogebra contain variables, dependencies and agencies. The interface of Geogebra is much better than tkedeen. It has a list of variables for the model, along with definitions of each variable and their dependencies in a separate window. Changes applied to each variable will be updated dynamically in the interface to provide real-time information of the models current state. Ideas from Geogebra can greatly affect this dissertation project to improve tkedens user interface. The general methods will be used in this dissertation project would be doing literature review of several tools that has similar functionality as tkeden. Reviewing their interface and compare it with tkeden. Also, external tools that were not built to support Empirical Modelling principles will also be included in the literature review process because to improve the interface, not only tools that within the EM filed need to be considered but also external tools that serves for different purpose will also be considered. Examples like Geogebra, Scratch etc are a very good starting point. Project Management Overview In this dissertation project, the aim is to provide an improved user interface of tkeden that can present real time information of observables. The project will be split into four stages. The first stage is to research and review the ideas from tools that have been developed based on the EDEN language. This stage reviews the ideas of these tools and seeks the possibility of integrating them into tkeden in order to provide a user-friendly interface that presents more information about observables automatically. Research is conducted into the core of EDEN interpreters, for their potential ability to achieve this goal. For example, the symbol table in EDEN interpreters has been found but still requires a significant amount of work to make use of the symbol tables. In the second stage, design tools will be implemented based on the research done in stage one. This will be built using the EDEN language, which is more like an add-on for tkeden. The third stage implements the tools designed in stage two and reviews the implementation to seek potential improvement of the current design. Research will be required after the implementation to compare the implementation with currently available tools, identify any weakness or disadvantage of the implementation and improve the design by adding new ideas or de-bugging current implementations. In the final stage, testing and report writing to document the process of this research project will be carried out. Time Table This dissertation also includes an add-on for tkeden, for which some of the software development methodology will be used. Prototyping will be the first choice to conduct the development. Several prototypes will be developed and user feedback on these prototype are very important because only users can decide if it is a good interface and what it is they need. Hence feedback will be gathered from several member swithin the department as well as PHD students that are currently working on research for Empirical Modellings. The user will spend most of their time using tkeden or similar tools for building EM models. There are also workshop sessions in July where many students will explore and develop models with EM principles. Hence feedbacks from these users are also important. The overall aim for the first prototype of the software will be set before July so that the prototype can be tested in the workshop session to gather feedback. In the first stage, large amounts of work involving literature reviews and gathering ideas and inspirations from tools within EM field are required. The estimate time scale for stage one will be 4-6 weeks. The deadline for this stage is set to be around the end of March. A software requirement should be produced before the deadline in order to proceed to the second stage. In the second stage, a design of the software architecture based on the software requirement produced from stage one will be produced. The estimated time scale for this process will be 4 weeks. The deadline for this is will be the end of April. An implementation of the prototype according to the design drawn from stage two should be produced in the third stage. The estimated time scale for this process will be 6-8 weeks and there may be delays due to the May/June exams. The target deadline will be mid to late June. The testing and debugging process should be commenced immediately after the prototype is developed. This is to meet the date set for the EM workshop session in July. In stage four, reviewing feedback gathered from users will be used to develop improvement of the design and if required re-design the software architecture. The time scale for this will be by the end of August is because there may be several repeating iterations from the earlier stages. Once this stage is finished, the rest of the time will be for report writing and debugging the final implementation. Conclusion To conclude, in current Empirical Modelling research, several tools have been built to support the principles of EM. The lack of support for the interface of tools to identify observables, dependencies and agencies in a model motivated this research project. The objective of this research project is to provide an environment that can easily identify observables and their relationship in a model, so modelers can focus on building their models in the EM approach without having to worry about maintaining a list of observables and their relationships in the model. This also provides easily understandable information for users to analyze and improve any models that are already built by others. This research may have a great impact in the ongoing Empirical Modeling research. Read More