This page answers some interesting questions about the work we do in the Optimal Learning Lab
How do I use the algorithms you have developed to do my own tutoring ?
|The are five ways the work on this website might be applied.|
- Individual - After the tutor is fully developed it will be available to individuals. The user will be able to enter their own learning set, load sets from a main web page, and share sets with other individual users.
- Courseware - Classroom testing is currently underway in Chinese and Spanish to deploy a system for vocabulary practice. This system will provide a central location for tutor distribution, secure login, and data collection. Detailed reports on learner performance will be available to teachers.
- Application - Fixed content tutors might be mounted at any web location. For instance, a personnel web page might have a tutor that helps new employees memorize the names of their coworker's. Perhaps a geography website wants to add a world countries and capitals memorization tool. Perhaps a game designer wants a tool that allows game players to quickly memorize the key mappings on the game controller.
- Research - Because the tutor includes meticulous recording of learner actions, and because the model allows one to characterize quantities such as forgetting or the rate of encoding, the tutor provides a setting in which to test other procedures to enhance learning. Current plans will allow multiple choice responses, picture content, and experimental and assessment sessions that have experimenter determined schedules rather than algorithm based schedules.
- Teacher Aids - Discussions have begun on how the tutor might be configured for teacher use. The idea is that the teacher would record learner actions during learning (say on a touchscreen computer) and then the computer would display dynamic recommendations for instruction based on a model of student learning.
How does the algorithm decide what is "optimal" ?
|The algorithm for scheduling practice uses a mathematical model of learning to predict when new practice should occur for recall to be optimal later. This model accounts for:|
- When prior practice occurred
- How many prior practices occurred
- Spacing between prior practice was
- Whether prior practice occurred as testing or passive study
- Duration of prior practices
- An individuals history of success or failure with tests
- What type of practice occurs
How does the algorithm decide when an item has been learned well enough ?
|The model can be used to compute what the expected future recall probability is for every item in the set of items being learned. If that recall probability meets a prespecified criterion then an item can be removed from the learning set.|
Similarly, if one merely wants to maximize improvement in probability correct as a function of time given a set of items, the model can be used to compute how many practices each item should get to maximize future recall probability for the amount of time spent learning.
How does the algorithm decide when to introduce items the learner has not yet seen ?
|New items are introduced only when previously learned items are well learned. This is so for two reason:|
- Introducing new items is sometimes necessary to maintain the optimal spacing. Therefore when all items have been introduced, spacings of practices become too narrow and efficiency of the system decreases slowly.
- Introducing too many new items degrades overall performance by making the task too difficult.