Live presentation

Asynchronous discussion


Machine Learning algorithms to increase children motivation in cognitive training videogames

Melina Vladisauskas

  • Ciudad Autónoma de Buenos Aires,
  • Argentina
  • Melina Vladisauskas ¹
  • , Laouen M. Belloli ²
  • , Diego Fernández Slezak ²
  • , Andrea P. Goldin ¹
  • 1 Laboratorio de Neurociencia, Universidad Torcuato di Tella - CONICET
  • 2 Laboratorio de Inteligencia Artificial Aplicada, Depto. de computación, FCEyN, UBA–CONICET

Mate Marote is an open source cognitive-training software aimed at children from 5 to 8 y.o. It consists of a set of computerized games specifically tailored to train executive functions (EF): a class of processes critical for purposeful goal-directed behavior, including working memory, flexibility, and cognitive control.

For the last ten years we have been using this software to measure and train children’s EF at their own schools. The interventions involve 4 sequential stages: (1) familiarization, (2) pretest (baseline), (3) training and (4) posttest. At present, all children start playing each training game at the same difficulty level. While the trials’ complexity rapidly adjusts to the child’s performance, starting in a non-challenging level may cause some children a motivational decrease, which is a known factor that can diminish the positive outcomes of a cognitive training intervention.

In the present study, we used Machine Learning to create an algorithm that classifies children in groups according to their performance in stage 2 in order to personalize stage 3. We trained and tested the algorithm with data from an intervention in which 73 6-year olds participated. This is the first time that we apply these computational models to our data, and preliminary results suggest that the algorithm fits the data neatly and can be applied in future interventions. Further studies will test whether this approach results in better cognitive training.