AI for Education
Personalizing education using deep learning
Personalized education is proven to produce learning gains for the
average student to the order of two standard deviations. The
challenge is to apply recent breakthroughs in deep learning to make
personalized education accessible to everyone — providing these
benefits on a global scale.
Student Achievement Score
Harnessing recent breakthroughs in deep learning, we present an
algorithm that can personalize education using only historic data.
This algorithm learns dependencies in content, models how student’s
learn — and subsequently present the right content at the right time.
The result is more engaged students and higher proficiencies.
In this example, we use Recurrent Neural Networks (RNNs) to
recommend content for each user. The RNN family of models have
important advantages over previous methods in that they do not
require the explicit encoding of human domain knowledge, and can
capture more complex representations of student knowledge states.
The RNN represents the latent knowledge state, along with its
temporal dynamics. Then, as a student progresses through a course,
it utilizes information from previous time steps, to make predictions
regarding future performance.
The model takes the content, answers, response times and an array
of contextual information as an input and process it through a
number of different network layers containing millions of neuron-like
connections to inform the predictions.
The prediction is subsequently used to analyze which piece of
content that won’t be too easy nor too difficult, is most likely to keep
the student engaged over time, and will result in the highest increase
in proficiency over multiple time steps.
During the fall of 2017, we will apply deep neural networks on one of
the world’s largest educational products, and we will be delighted to
share this case study with you at SXSWEdu.
With knowledge, passion, and creativity we will showcase how the
use of deep neural networks can increase student engagement and
proficiency, as well as reduce churn.