Thursday, November 15, 2012

Heuristic teaching challenge

I am excited. Our company is taking a huge step in the direction of completely individualized learning, and we are hiring a lead server engineer to that end. Here's the listing, just in case you want to help us find that person:

This position is SO important to me that I want to discuss how the person that takes it is going to change the learning experience for our students. My hope is that we can have an open discussion about this project so that it improves everything for everyone. Certainly work of the type described here is being done in different areas, but since this is going to be so unique to our programs, I feel that it is completely okay to speak of this openly.

Some background on our company: we have a series of educational products that teach vocabulary and critical reasoning. They are amazingly effective and they do actually adapt to students' knowledge. But they could be so much better.

Some background on me: I've been making video games for 20 years. I believe that we can take gaming concepts and apply them to other fields, like education. That's why I'm here at FastPath.

Artificial Intelligence (AI) within video games can create situations that make for truly memorable and engaging experiences. If a game's AI is too easy, the game is boring to a player. If the AI is too hard, the game is considered unplayable. Neither of these are acceptable in the gaming market. However, when the AI is neither impossible or too simple, it can be called "just right," the experience is then considered "fun." This can also be called "pleasurable frustration."

The challenge of creating an experience filled with "pleasurable frustration" is that a players' ability to play or interact with a game continually changes, whether it's because they are gaining experience or maybe because they are having a bad day.

This can be adjusted for in gaming by adding "handicaps" such as in a foot race where you give someone a 10 second lead, or on a carnival midway when the barker gives you a couple of extra softballs to knock over the milk jugs. But those are done with people that can continually assess the situation.

The same can be said for having a teacher that understands their student well. They can take what they know about the student and use that to adapt their curriculum accordingly. For instance, they may be a soccer player so their physics discussions surround making a goal in a vacuum, or they may be sick and they need a few reminders in order to keep their mind focused upon the current topic.

In a video game, creating a dynamic experience that adapts to the players' performance is a challenge, but it is a common practice. Some games adjust the likelihood that you will get hit by a laser bolt, while others may make the enemies move slower so you can catch them easier.

In an automated online educational product, measuring a students' knowledge and wisdom is a challenge. Currently we use an assessment test that creates a score based upon a series of questions with different difficulties. The performance of a student in that assessment determines how far we move a student forward. This is a very valid way to customize a program, but it is not perfect.

Our challenge is to create an educational program that constantly adapts to an individual students' performance from every conceivable angle. This will create an educational version of "pleasurable frustration" that will also result in "fun."

We have a plan in place that, more than anything, needs a special person to make it a reality. I don't think I'll find the right person without making this plan somewhat public. So here goes my attempt at explaining my current take on it:


We take the data from all of our products and add it to a single database. This means we can then edit everything and preform analytics across programs.


This is where, I believe, we can start grading and sorting the data about how students perform in a way that we can build patterns of performance from. For example, we can compare the data for everyone that missed question 1 and begin predicting what other questions they may miss or get right.

To accomplish this, we'll need to build some serious analytics tools and some creative data storage concepts.


All data must be served up and measured / tracked based upon A/B testing so we can ascertain that we are providing students the best learning experiences. The types of things we could choose to test: audio files, questions, answers, button sizes, button colors, page layouts, etc.


Once these pieces are in place, we can track and measure performance of each user, find trends within groups of users, change the level of difficulty to determine if a student is ready to jump forward, and find many other ways to provide students a completely unique learning experience that is heuristically determined to be the "best" content for them at that moment.

If this sounds interesting to you, and you want to make it happen, then we need to talk. Now.

No comments:

Post a Comment

Follow by Email