I learn and then I forget, that's life, that's life [Thesis]

When adaptive learning spacing algorithms help us forget less

Student on unicycle advancing in his learning.

Children have short memories, but they remember quickly.

Victor Hugo - Writer (1802-1885)

It can be hard not to get overwhelmed when learning new skills. We barely have time to master a new concept before it's already time to absorb another and another. But how do we stay productive and avoid one concept chasing another?

The majority of existing curricula and school systems promote learning modes that solicit short-term memory. This learning allows for rapid progress but does not allow for consolidation of learning at the level of long-term memory. According to constructivists, one piece of knowledge builds on another and so on. So if we forget as we learn, how can we have solid knowledge that is not shaky? But the reverse is not much better. How can we make proper progress in acquiring knowledge if we freeze up revising the basics of it?

A real dilemma that Benoît Choﬀin proposes to develop in his thesis entitled "Adaptive Learning Spacing Algorithms for Optimizing Long-Term Mastery of Knowledge Components".

Simple as Hello

Some manuscripts featuring codes, matrices, and equations may intimidate some novices. Benoît Choﬀin invites us into this world in a delicate and logical way with a style that allows us to gradually assimilate and understand the different notions and problems addressed.

The narrative of this manuscript unfolds before our eyes in such a limpid manner that we would forget the complexity of the subject and the mass of work required to complete this thesis.

Benoît Choﬀin demonstrates a critical mind on his field both about his peers' work and his own and never forgets to propose a solution or an improvement to the identified weaknesses. Thus, at the end of the thesis, he proposes a number of strengthening and improvement measures that can be implemented on the presented tools.

Knowing what you want

"If a pupil or student were asked today whether revisions spread out over time are beneficial to learning, it is a safe bet that they would respond favorably. Few learners question the benefits of periodically revisiting previously acquired knowledge. However, few actually practice this strategy to consolidate their long-term knowledge and more choose learning strategies that promote short-term memory.

In fact, despite the recommendations of many cognitive scientists, school curricula still include too few incentives for systematic knowledge review and consequently for long-term memory.

As early as 1967, Pimsleur noted the almost total absence of such incentives in textbooks or in the training of teachers and trainers. Knowing that learning new knowledge often builds on older knowledge and requires a time investment as well as significant cognitive efforts, improving long-term memorization and permanence of knowledge is an essential issue in any learning.

Periodic review of knowledge is better known in cognitive psychology as spaced repetition. More formally, the spaced repetition strategy involves breaking up the learning of the same set of information (for example, vocabulary words in a foreign language) into smaller learning sessions spaced over time. Spaced repetition improves long-term memory of this information compared to learning in a single "massed" session: this benefit of spaced repetition is called the spacing effect. Note here that this strategy is better than "massed" learning for long-term memorization even with equal learning time.

However, while the benefits of spacing compared to massed learning are clearly established, the question remains as to how, in practice, to space out one's revisions.

In order that the revision of past knowledge does not interfere too much with the acquisition of new knowledge, it is necessary for learners to not plan more revision sessions than necessary while determining the timing of these sessions as best as possible. This is especially important because the right amount of spacing between initial acquisition and revision largely determines the magnitude of this effect."

Memory Refresher

The results of this thesis consist of improvements in tools that promote the integration of students' acquired knowledge into their long-term memory. These tools called learning spacing algorithms allow the automation of review and test suggestions for students to consolidate their knowledge. However, these tools do not adapt, or adapt too little, to their users' profiles.

The author's work has led to the creation of tools and models for the design of an adaptive and personalized learning spacing algorithm. To do so, Benoît Choﬀin reconsiders the mathematical representation of the learner by marrying statistical models and cognitive psychology models. Thus, the author is able, through this method, to envisage, taking into account the student's learning and forgetting, a learning curve for given skills, time and student.

The integration of the tools developed by the author in the Moodle Plume plug-in allows its users, the implementation of personalized worksheets for each learner according to his needs, his knowledge to be reviewed or assimilated as well as his learning stage. The author mentions the possibility of using this tool in a classroom to generate worksheets for each student to work on their own weaknesses or to help self-learners in their learning.

At your fingertips

Benoît Choﬀin's work contributes to our survival in a world where everything goes so fast and where it is so easy to forget and be overwhelmed when learning. The tools developed help to potentiate learners' learning time by allowing them to procrastinate our ability to forget.

What about you? How do you go about not being overwhelmed when learning?

Good reading!

Thesis presented and defended on January 28, 2021. Work carried out at the Interdisciplinary Laboratory of Digital Sciences (CNRS) within the doctoral school Sciences and Technologies of Information and Communication (STIC) : ED 580 (Université Paris-Saclay) (Orsay).

Sources

For more information on the Moodle Plume plugin

https://hal.archives-ouvertes.fr/hal-01764376

Thesis:

Benoît Choffin. Adaptive learning spacing algorithms for optimizing long-term mastery of knowledge components. Learning [cs.LG]. Université Paris-Saclay, 2021. French. ⟨NNT: 2021UPASG001⟩. ⟨tel-03216648⟩