Med_L&T_med

Educational data mining (EDM) is emerging as a field of research that combines computational methods, humanities, and research approaches to understand how students learn. New tools and methods for interactive computer-assisted learning—intelligent tutoring systems, simulations, games—have paved the way for collecting and analyzing student data, discovering patterns and trends in that data, making new discoveries, and testing hypotheses about how students learn. Data collected by e-learning systems can be aggregated across a large number of students and may contain numerous variables that data mining algorithms can explore to build models.

EDM focuses on developing methods to analyze the unique types of data generated by educational institutions and using these methods to better understand students and the environments (learning situations) in which they learn.

EDM focuses on developing new tools and algorithms to discover data patterns. EDM develops methods and applies techniques from statistics, machine learning, and data mining to analyze data collected during teaching and learning. EDM tests learning theories and informs educational practices.

The purpose of the project is to explore a general technical and pedagogical framework to support decision-making in the context of technology-enhanced learning (TEL). That is, to support the decisions made by TEL stakeholders (teachers, tutors, students, institutions, communities of practice) based on learning and teaching data. A teacher is a TEL stakeholder when they prepare instructional materials (for example, by organizing resources for a lesson sequence or designing a learning scenario), when they deliver instruction (for example, by adapting instructional materials or supporting students as the learning framework is implemented), and when they assess learning outcomes . A tutor is an AAT actor when they support and supervise the learning process. A student is an AAT actor when they take a course (for example, when they compare themselves to other students, find the best resources, self-regulate their learning, etc.). An institution is an OAL actor when it manages teaching and learning processes at the institutional level (for example, by offering a curriculum, developing the faculty, managing administrative workflows, etc.). A community of practice (for example, a community of mathematics teachers) is a stakeholder when it creates and shares content, instructional scenarios, problem-solving tools, or best practices. We refer to the decision-making role in OTL situations as an OTL stakeholder.

We believe that decisions regarding AAT can be improved by providing AAT stakeholders with relevant analyses and reports on the educational data collected, such as data on student motivation, actions, and knowledge; and data on the actions and/or practices of other teachers from a community-of-practice perspective.

The medical field provides an ideal setting for exploring this new research framework. Indeed, we can identify several learning situations that utilize an emerging pedagogical approach (the flipped classroom) with traditional multiple-choice tests (MCTs), such as PACES, or an innovative pedagogical approach with new testing systems (serious games, simulators), such as LOE or TELEOS. It is also a large-scale experimental field involving thousands of students and dozens of instructors.

In both cases, we will develop process analyses for various stakeholders in the AAT sector to address a number of questions. The analysis process will be integrated into the Undertracks platform. These analyses will be designed to be shareable and reusable.

To examine the reusability of the analysis processes, some of them will be applied in other similar contexts. In particular, the analysis process for understanding student progress and the co-design process with tutors could be reused in the context of “C2i Level 1” because both use the same learning platform (Learneos) and the tutors’ objectives are more closely aligned.

Research Objectives

The overall objectives of this long-term collaboration are as follows:

1. Predict students' future learning behavior by creating student models that incorporate detailed information such as students' knowledge, motivation, metacognition, and attitudes;
2. Identify or refine the domain models that characterize the content to be learned and the optimal instructional sequences;
3. To study the effects of the various types of educational support that can be provided by learning software;
4. To advance scientific understanding of learning and learners by developing computational models that integrate models of the student, the subject matter, and pedagogy.

The short-term objectives are as follows:

1. Explore and design an analytical process to understand student progress based on PACES data. Propose an indicator to measure this progress.
2. Reuse this analysis process in other similar contexts.
3. To explore and co-design, together with PACES instructors, an analytical process to help them develop an effective instructional framework for creating useful multiple-choice questions—that is, questions that take into account both students’ needs and institutional constraints.
4. Propose a set of indicators to feed into a computational model of the student. Each indicator will involve an analytical process. The indicators analyze the student’s behavior, knowledge, and cognitive strategies based on the LOE learning platform. Test the indicators in the context of the TELOS project.

Why collaborate?

The Undertracks platform, developed by the MeTAH team, aims to share experimental data, operators (algorithms for data analysis and visualization), and educational data analysis processes. The team is designing and implementing a web platform to store structured educational data and operators (currently written in Java or C++). In particular, they are designing a graphical interface and a processing engine that enable the combination of data and operators. However, while this platform is designed to facilitate the sharing of experimental data analysis among researchers, it does not provide functionality for end users such as distance education stakeholders. This project allows us to go further and study the properties of platforms for end users. In addition, several innovative operators (analysis and visualization algorithms) and processes must be designed to meet the aforementioned research objectives.

The MeTAH team also manages the TELEOS platform, which was developed in partnership with the TIMC, the LIP, and the orthopedic department at Grenoble University Hospital. TELEOS is a simulation-based intelligent tutoring system (ITS) that was implemented to provide a complementary learning dimension to the orthopedic surgery training process. In fact, in addition to the declarative knowledge that resident surgeons must master, another part of their training requires repeated practice. The knowledge involved in this part is perceptual-motor, meaning it is often tacit and empirical. Furthermore, the tasks associated with this type of knowledge are also poorly defined, as different strategic approaches can be applied to perform a given operation, and no precise method can be defined in advance to meet the validation criteria for related tasks. This creates a gap in the learning process that is difficult to bridge using traditional teaching methods. The TELEOS learning environment aims to provide the missing intermediate phase of learning.

The Themas team is deeply involved in the new PACES curriculum and is co-developing the LOE platform, a serious game designed to teach concepts related to epidemiology.

The Laboratorium of Epidemiology© (LOE) immerses learners in a large-scale, persistent, distributed simulation combined with a game-based scenario. It was designed collaboratively and used by researchers, instructors (hereinafter referred to as tutors), and students as part of both an educational project and a research project. The educational project is part of a biostatistics course in the medical school. The research project, which we have named “Laboratorium,” enables repeated data collection campaigns that are not one-time events in the lives of students or tutors and represents an attempt to reduce data collection biases and produce well-documented databases.

The game is based on a computer simulation of various institutions (including hospitals) and on role-playing. Students take on the role of public health physicians and are placed in a professional scenario that would otherwise be inaccessible to them, involving the outbreak of a disease in several hospitals. Students must design and conduct an epidemiological study and write a scientific paper to be presented at a conference. The primary learning objective concerns the statistical analysis of a medical database. However, students are given an assignment that contextualizes statistical problems. The assignment involves designing a diagnostic tool for VTE (venous thromboembolism) for hospital use. By working on this problem, students will learn statistics, understand the role they play, and, more generally, the function of statistics in public health. The game is fully integrated into the medical school’s standard curriculum. It lasts four months, including eight four-hour classroom sessions.

Both frameworks (PACES and LOE) generate educational data. Automating these processes and speeding up the collection and analysis of this educational data will be crucial for understanding complex phenomena and generating new insights into student behavior and pedagogical phenomena.

Contacts

MeTAH Team (LIG Laboratory): Vanda Luengo

Themas Team (TIMC Laboratory): Pierre Gillois

Publications supported by the project

Toussaint, B.-M., Luengo, V., Jambon, F., and Tonetti, J.: From Heterogeneous Multisource Traces to Perceptual-Gestural Sequences: the PeTra Treatment Approach. In: Conati, C., Heffernan, N., Mitrovic, A., Verdejo, M. F. (eds), Proceedings of the 17th International Conference on Artificial Intelligence in Education (AIED 2015), Madrid, Spain. LNCS, vol. 9112, pp. 480–491. Springer, Heidelberg (2015)

Toussaint, B.-M., Luengo, V.: Mining surgery-phase-related sequential rules from vertebroplasty simulation traces. In: Holmes, J.H., Bellazzi, R., Sacchi, L., Peek, N. (eds.) Proceedings of the 15th International Conference on Artificial Intelligence in Medicine (AIME 2015), Pavia, Italy. LNCS, vol. 9105, pp. 32–41. Springer, Heidelberg (2015)

Toussaint, B.M., Luengo, V., and Jambon, F.: Proposal for a trace processing framework for the analysis of perceptual-gestural knowledge—The case of percutaneous orthopedic surgery. Proceedings of the 7th Conference on Computer Environments for Human Learning (EIAH 2015), Agadir, Morocco, June 2015