An Exploration of the Challenges and Possibilities of Multidimensional Visualization in the Context of Visual Learning Analytics
Rafael M. Martins1, Marcelo Milrad1, and Italo Masiello2
1Dept. of Computer Science and Media Technology / 2Dept. of Pedagogy and Learning Linnaeus University
1. Problem Definition and Value
Data science, or the process of analysing large repositories of data in search of useful patterns and insights, has matured into a well-defined discipline with disruptive results in many different knowledge domains [1].
When used in the context of education and learning, it is usually known as Learning Analytics (LA), a field which has been gathering more attention recently due to the recognition of its potential to improve our understanding of the theory and practice of learning [2]. Visualization supports analysts to understand, interact, and influence both the process and the results of the analysis of complex data. When interwoven with a data science workflow, it is known as Visual Analytics (VA), or the "science of analytical reasoning facilitated by interactive visual interfaces" [3]. The importance of applying VA in combination with LA techniques has been identified in recent works [4], giving rise to the emerging field of Visual Learning Analytics (VLA). However, as noted by some authors [5], the field of VLA is still in its infancy, and its applications are mainly limited to the use of simple statistical charts, such as bar plots and scatter plots, in processes that are usually not deeply supported by solid educational theories. Based on our experience with applications of multidimensional VA in other domains, we can identify a large potential for improvements in VLA when it comes to using sophisticated visual abstractions applied to larger and more complex data sets.
2. Objectives
The activities of this seed proposal are divided into two main objectives, to be performed in parallel:
O1. A theoretical one, where we will strive to understand the current state of the art in visual learning analytics, its most relevant open challenges, and the potential areas where multidimensional visualization can contribute. Our main question here is: What are the important open challenges in visual learning analytics with multidimensional data?
O2. A practical one, where we will strive to familiarize ourselves with the large-scale, multidimensional data which will be provided by our industrial partners, and obtain initial insights about its structure and potentials. Our main goal here is to answer the following question: What kind of valuable information about the learning processes can be extracted from the raw data by using multidimensional visualization?
Ideally, among the open challenges identified, we will find the most relevant ones and obtain a better understanding of the domain and the solution space of possibilities to be derived from the data. This should lead us then to the final objective:
O3. Conceptually join the lessons learned from the first two objectives, mapping the abstract research challenges into concrete, novel methods of visual learning analytics, which will then be consolidated into the first draft of a KK proposal for external funding.
3. Expected Results
Each of the previously defined objectives will result in one partial deliverable: [O1] the theoretical objective will result in a draft of a paper describing the current state of the art in multidimensional visualization applied to learning analytics, in the format of a systematic literature review; and [O2] the practical objective will result in a visual analytics prototype on top of the data that will be provided to us by the industrial partner. The prototype will be influenced by the literature review and will function as a catalyst for the dialogue between all the different parties involved in the seed project, informing the decisions regarding [O3] the writing of the final KK proposal for external funding.
4. Consortium
The main applicants are Rafael M. Martins (Senior Lecturer in Computer Science and Media Technology), who will lead all the planned activities; Marcelo Milrad (Professor in Computer Science and Media Technology), who will advise on the integration of digital technologies and learning; and Italo Masiello (Professor of Educational Technology), who will advise on the applications of learning analytics. Alisa Sotsenko Lincke, (PhD Candidate in Media Technology) will advise on learning analytics, based on her experiences with the company and the data, and another research assistant (N. N., to be defined) will be involved mainly with tasks of programming and software development for the prototype. The industrial partner who will provide us with data and domain knowledge is Hypocampus (Göteborg), with a potential contribution from IST (Växjö).
6. Activities & Time Plan
The planned activities should be executed from March to August 2019. There will be two progress/planning meetings per month with the internal consortium members, of two hours each, and one meeting per month with the industrial partner, also of two hours each. Other details:
● March-April: [O1] Define and refine the methodology of the literature review, perform the initial pilot search; [O2] Obtain the data and bootstrap the visual analytics prototype; [O3] Write the initial draft of the KK proposal.
● May-June: [O1] Iterate on the literature review, following the previously-defined methodology;
[O2] Iterate the main features of the visual analytics prototype; [O3] Finish and submit the KK proposal.
● July-August: [O1] Finish and submit the literature review paper; [O2] Finalize the visual analytics prototype.
7. Budget
The total budget of the seed proposal is 200.000 SEK, co-funded by DISA and the consortium (50% each), which will be used for (1) funding the working hours of the members, and (2) funding the meetings with the industrial partner. In details (values in SEK):
From \ To Rafael Marcelo Italo Alisa N. N. Total
DISA 75.000 - - 10.000 15.000 100.000
Own funds 15.000 50.000 25.000 10.000 - 100.000
References
[1] Hey, Tony, Stewart Tansley, and Kristin Michele Tolle.The Fourth Paradigm: Data-Intensive Scientific Discovery. Microsoft Research, 2009.
[2] Baker, Ryan Shaun, and Paul Salvador Inventado. Educational data mining and learning analytics.
Learning analytics. Springer, New York, NY, 2014. 61-75.
[3] Thomas, James J., and Kristin A. Cook. A visual analytics agenda. IEEE Computer Graphics and Applications 26.1 (2006): 10-13.
[4] Kay, Judy, and Susan Bull. New opportunities with open learner models and visual learning analytics.
International Conference on Artificial Intelligence in Education. Springer, Cham, 2015.
[5] Vieira, Camilo, Paul Parsons, and Vetria Byrd. Visual learning analytics of educational data: A systematic literature review and research agenda. Computers & Education 122 (2018): 119-135.