ADDIA
ADDIA

ADDIA
ADDIA
 : 
Plataforma d’aprenentatge competencial personalitzat mitjançant l’ús d’intel·ligència artificial
Plataforma d’aprenentatge competencial personalitzat mitjançant l’ús d’intel·ligència artificial

A Project coordinated by IIIA.

Web page:

Principal investigator: 

Collaborating organisations:

DIDACTIC LABS

DIDACTIC LABS

Funding entity:

Generalitat de Catalunya
Generalitat de Catalunya

Funding call:

INNOTEC
INNOTEC

Funding call URL:

Project #:

ACE014/20/000039
ACE014/20/000039

Total funding amount:

198.183,00€
198.183,00€

IIIA funding amount:

97.285,00€
97.285,00€

Duration:

01/May/2021
01/May/2021
31/Dec/2022
31/Dec/2022

Extension date:

The project ADDIA is a joint project between the company Didactic Labs and the IIIA-CSIC. It aims to address a complex but high-impact problem such as the ability to personalize learning for students. It is a task that, performed by a person is of enormous complexity and requires a very high volume of time and for this reason, it is not usually done. Achieving this, however, would mean having students better prepared for the world of work and more involved in their education.

To meet this need, the following challenges will be addressed in the project:

Challenge 1: Pedagogical data modelling and data mining.

To carry out the project it is necessary to create an isolated environment of an experimental nature to develop, optimize and evaluate the system as well as to define the properties and constraints of the different entities (mainly students and activities). Didactic Labs has a huge database of assessments and class plans that will need to be addressed and refined. It will be necessary to apply transformations of some data, define patterns and models, detects sub-sampling or generate synthetic data with SMOTE type algorithms to obtain a meaningful data set and to avoid the problem of a Cold Start.

Challenge 2: Automatic identification of homogeneous groups of students.

It is well known that learning occurs much more efficiently in the community and when the learning group is relatively homogeneous with respect to the competencies to be acquired. Identifying groups of students who are similar in this respect will be key to the success of learning process management automation.

Challenge 3: Recommendation of tasks based on competence deficits.

The basic principle of task-oriented learning is that in solving a case study the skills needed to solve it are achieved more solidly. In the IIIA, algorithms have been developed that, based on a given task, propose teams of students who can solve it. However, where task-based collaborative learning can have a deeper impact on improving students' competency is determining what task to assign to a given group of students so that they can all improve some of their skills. We want to find algorithms that make optimal workgroups such that, once they are assigned tasks based on skills deficits, they maximize the learning of the whole group. In addition, these methods must guarantee a series of restrictions/rights such as diversity, ensuring that teams have members with the required level of competence and others who do not have it, equal opportunities, maximizing the increase in competence of all students in a balanced way and gender equality, among others.

Challenge 4: Recommendation system with an explanatory and dialogic capacity.

To recommend the actions to be taken by the teacher to correct the shortcomings detected, both for individual students and groups of students with similar needs, we will use the technology of recommending systems. Instead of entrusting this task to a recommended ground (a single technology), we propose the creation of a multi-recommended system that adds the vision of a set of specialized recommenders. The proposed mechanism, however, wants to go beyond the simple recommendation. It should be a mechanism that should accompany the teacher in making decisions. Achieving this requires a system that is capable not only of recommending actions but of being able to justify, explain and discuss them. A pedagogy team will be in charge of evaluating the accuracy of the system during the development phase.

The project ADDIA is a joint project between the company Didactic Labs and the IIIA-CSIC. It aims to address a complex but high-impact problem such as the ability to personalize learning for students. It is a task that, performed by a person is of enormous complexity and requires a very high volume of time and for this reason, it is not usually done. Achieving this, however, would mean having students better prepared for the world of work and more involved in their education.

To meet this need, the following challenges will be addressed in the project:

Challenge 1: Pedagogical data modelling and data mining.

To carry out the project it is necessary to create an isolated environment of an experimental nature to develop, optimize and evaluate the system as well as to define the properties and constraints of the different entities (mainly students and activities). Didactic Labs has a huge database of assessments and class plans that will need to be addressed and refined. It will be necessary to apply transformations of some data, define patterns and models, detects sub-sampling or generate synthetic data with SMOTE type algorithms to obtain a meaningful data set and to avoid the problem of a Cold Start.

Challenge 2: Automatic identification of homogeneous groups of students.

It is well known that learning occurs much more efficiently in the community and when the learning group is relatively homogeneous with respect to the competencies to be acquired. Identifying groups of students who are similar in this respect will be key to the success of learning process management automation.

Challenge 3: Recommendation of tasks based on competence deficits.

The basic principle of task-oriented learning is that in solving a case study the skills needed to solve it are achieved more solidly. In the IIIA, algorithms have been developed that, based on a given task, propose teams of students who can solve it. However, where task-based collaborative learning can have a deeper impact on improving students' competency is determining what task to assign to a given group of students so that they can all improve some of their skills. We want to find algorithms that make optimal workgroups such that, once they are assigned tasks based on skills deficits, they maximize the learning of the whole group. In addition, these methods must guarantee a series of restrictions/rights such as diversity, ensuring that teams have members with the required level of competence and others who do not have it, equal opportunities, maximizing the increase in competence of all students in a balanced way and gender equality, among others.

Challenge 4: Recommendation system with an explanatory and dialogic capacity.

To recommend the actions to be taken by the teacher to correct the shortcomings detected, both for individual students and groups of students with similar needs, we will use the technology of recommending systems. Instead of entrusting this task to a recommended ground (a single technology), we propose the creation of a multi-recommended system that adds the vision of a set of specialized recommenders. The proposed mechanism, however, wants to go beyond the simple recommendation. It should be a mechanism that should accompany the teacher in making decisions. Achieving this requires a system that is capable not only of recommending actions but of being able to justify, explain and discuss them. A pedagogy team will be in charge of evaluating the accuracy of the system during the development phase.

2021
Ariadna Quattoni,  & Xavier Carreras (2021). Minimizing Annotation Effort via Max-Volume Spectral Sampling. Findings of the Association for Computational Linguistics: EMNLP 2021 (pp. To appear). Association for Computational Linguistics. [BibTeX]  [PDF]
Pablo Aramendía
Contract Engineer
Phone Ext. 431836

Jesus Cerquides
Scientific Researcher
Phone Ext. 431816

Jordi Sabater-Mir
Tenured Scientist
Phone Ext. 431856

Carles Sierra
Research Professor
Phone Ext. 431801