[Postdoc] Analysis of Traces in a Literacy and Numeracy Skills Application: Learning Analytics for Preschoolers
Type of position: post-doctoral position
Requirement: PhD in computer science
Duration: 11 months
Beginning date: September 2019
Contact: Sylvain Castagnos (sylvain.castagnos at loria.fr)
The KIWI team builds intelligent assistants and recommender systems to improve Human-Computer Interaction. It is a question of automatically modeling users’ needs, preferences or uses, and then helping them by customizing the service offered according to the context. This research theme is at the intersection of several major disciplines such as machine learning, and cognitive sciences. These research questions present numerous opportunities, among which: search engines and websites for personalized access to information for the general audience; tools for diagnosis and rehabilitation in the field of medicine (neurodegenerative diseases, autism, etc.); and e-learning platforms to offer educational resources and feedback tailored to the learning process of each user.
It is in the field of e-learning that this 11-month postdoctoral offer is part of. The KIWI team is a partner of the e-FRAN LINUMEN project alongside the L2PN and LISEC laboratories in Nancy, and the company LearnEnjoy in Paris. The goal of this project is to develop a digital tool to support emerging literacy and numeracy skills in preschool and elementary school children. This system should reduce cognitive inequalities related to the social origin of students and thus promote early school enrollment. In addition, in accordance with the recommendations of the new programs of the 2015 kindergarten, this device will allow students to develop skills related to the manipulation of digital tools and the discovery of their uses. In this context, an application for Android tablets has been developed. This application allows children of small, medium and large sections to realize a wide range of activities in literacy and numeracy with different levels of difficulty.
A study is underway in several schools to test the application in Lorraine (region in the North-East of France). Through this study, we are collecting all traces of interaction of the children during the accomplishment of the various activities, thanks to an event logger integrated into the application. For example, we can find the list of activities performed by each child, the description of the activities, the start and end dates, the answers given to each question within an activity and the associated timestamps, the number of mistakes made, etc.
The dataset built during this study will serve as a support for the successful candidate for the design of learning analytics tools to measure the progress made by children over time. Traces will first be processed by usage mining algorithms to eliminate noise, disambiguate and evaluate cases of uncertainty. The data, once processed, will be analyzed and transformed by machine learning models. The objective is to infer higher level information to understand the difficulties encountered by learners, to monitor progress and to allow personalized support by teachers or parents. By way of example, we can try to evaluate some cognitive functions of the child in a stochastic context, such as executive functions (inhibition, flexibility, update), attention, memory, comprehension, trust, preferences or mental effort required by the child. We will also be able to measure the student’s ability to move from a paper problem to his digital resolution. This inferred information will then be synthesized in graphical form and integrated into a dashboard.
In summary, several scientific questions motivate the collection and analysis of these digital traces, among which:
- the behavioral modeling: how to measure and predict trust, attention and effort?
- the explanatory factors: what are the factors explaining a more or less significant progression?
- the impact of the mode of interaction (e.g., validation button or drag-and-drop cards) on trust, attention and effort;
- the impact of the digital medium in relation to a paper medium;
- the identification of typical learner profiles: statistics of progress made by the child over time.
Achieving these goals requires strong knowledge in artificial intelligence and machine learning. Previous experience in the field of e-learning is desirable. Knowledge in cognitive science, child psychology, statistical neuroscience and / or human-computer interaction (UX design) would be a plus.
Interested candidates should contact Sylvain Castagnos as soon as possible by email (sylvain.castagnos at loria.fr) enclosing a resume and a cover letter. Applications end on May 30, 2019.