Abstract:

The scientific community widely acknowledges that student academic performance depends on factors such as shame, introversion, self-esteem, anxiety, motivation or associative memory. The challenge is to provide individual attention using ICT making possible the interaction and correction of these parameters.

The main goal is to present the development of an individualised learning mathematical formula (L) that is a combination (linear or not) of the mentioned parameters, and which has to be optimised in real-time: L = P1·f1(t) + P2·f2(t) + · · · + Pn·fn(t). Each fi(t) is a n-order quadratic function of time. In its simplest case (n=1), we will have a linear behaviour. Every fi(t) is weighted by a real number Pi which gives an idea of the importance given to the factor in the learning process.

In this work, and taking advantage of previous works [1][2][3], we have created the L normalised function (0 <= L <= 1) which depends only on associative memory f(1), anxiety f(2), self-esteem f(3), and motivation f(4), which are considered crucial parameters concerning student achievement. Then, the complete Learning Function L will be the sum of four terms (i=1, 2, 3, 4). Each term is the corresponding function fi(t) multiplied by its assigned weight Pi.

In the first step, associative memory and motivation can be suited to linear functions, meanwhile anxiety has been associated to a quadratic one, and self-esteem is dynamically adapted depending on every new value of L.

To start repetitive optimization calculations of (L) the first step to solve is how to obtain the initial weights Pi of those individualised parameters. The only way we found was to set approximate values based on individual students' tests, with the hope that the model would dynamically correct such values as students advance in the program.

The procedure is the following: Starting from initial packages of exercises, and based on the successive obtained results, the software modifies, for each parameter, not only its Pi weights but even the quadratic dependence on time, if needed. The new L value allows the system to adapt the next level of the exercises until the L function approaches the optimal value of 1.

Although the model has been shown to be effective with a small number of students, further research is needed to fully evaluate the effectiveness in a real-world educational environment, dealing with hundreds of students working with maths educational platform. We believe this approach has great potential to improve student outcomes by tailoring learning to individual needs and we are excited to continue exploring its possibilities.

References
[1] Roca Mató, M., Martorano Gomis, A. and Batlle Grabulosa, J. (2013). “Taking advantage of new technologies to minimise the impact of the factors that limit learning L2. Results from a real experience in a set of primary schools”. ICERI 2013 Proceedings. IATED Library.
[2] Martorano Gomis, A. Roca Mató, M., and Batlle Grabulosa, J. “A pilot based on a process of immersion of 1st year academic students in a problem-based methodology, dealing with a real industrial environment, and using an e-learning platform”, INTED 2014.
[3] Friedlander, Laura J., et al. "Social support, self-esteem, and stress as predictors of adjustment to university among first-year undergraduates." Journal of College Student Development 48.3 (2007): 259-274.

Keywords:

Personalized Learning, e-learning Projects and Experiences, Evaluation and Assessment, ICT.