Cognify

Cognify is a WIP, intending to be an openly extensible education platform which optimizes learning for the individual using machine learning.

Primarily, the application will used a modified form of spaced repetition consisting of a scheduler (for when to study specific subjects) and an optimizer (meant to fine tune the scheduler for a specific person’s needs)

The scheduler leverages a modified form of the DSR (Difficulty, Stability, Retrievability) model, which is used to predict memory states. The optimizer uses machine learning to learn a specific user’s memory patterns and finds parameters that best fit study performance on quizzes/flashcards/etc.

The base DSR scheduling system is also not based heuristically (in the way most education platforms’ schedulers are), and congregates data across all users to create a general DSR model.

To do this, we’re largely leveraging the following papers:

• A Stochastic Shortest Path Algorithm for Optimizing Spaced Repetition Scheduling
• Optimizing Spaced Repetition Schedule by Capturing the Dynamics of Memory

But the scheduler is only a small part of the application. We also intend to use what’s known in education and general psychology to try and maximize performance on subject tests and long term memory tests rather than pure app retention/app usage. To do this we’ll be using color psychology, interleaved practice, elaborative encoding, emotional regulation, and many more research based influences towards maximizing memory.

On extensability, we aim to build off the groundwork many note-taking applications have made in leveraging open source developers to add to the application.

The project is currently still in the early stages of development.