Profile cover photo
Profile photo
Joseph Jay Williams
695 followers -
As a faculty member at National University of Singapore, I research learning & health technology using HCI, psychology & machine learning.
As a faculty member at National University of Singapore, I research learning & health technology using HCI, psychology & machine learning.

695 followers
About
Posts

Post has attachment

Post has attachment

Post has attachment

Post has attachment

Post has attachment

Post has attachment

Post has attachment
Add a comment...

Post has attachment

Post has attachment
tiny.cc/williamstalk

Quality explanations (e.g. of concepts, how to solve problems) are important components of online problems and lessons. But instructors have limited time to author multiple explanations, and even less to collect data about which ones are helpful to students.

This talk presents AXIS, the Adaptive eXplanation Improvement System, an example of a plug-in tool that makes it easy for course teams to dynamically add explanations to online lessons & problems in EdX and Canvas, and automatically identify which ones are effective. The plug-in prompts students to answer reflective questions. This helps the students learn by explaining (Williams & Lombrozo, 2010), while crowdsourcing a pool of student answers/explanations that can be provided to help future students learn. 

Explanations from the pool are presented to help future students, with the frequency of presentation being proportional to an instructor/researcher's rating of explanation quality. As students themselves rate how helpful explanations are, a statistical machine learning algorithm (a Bayesian version of Thompson Sampling) automatically increases or decreases how often explanations are presented, based on how highly they are rated by students. This ethical experimentation combines the rigor of randomized experimental comparisons with real-time use of the data to help students.

In the context of online math problems, this approach is shown to produce explanations that increase learning and satisfy the instructor's needs (Williams et al, 2016). 

The talk also explains how the MOOClet formalism enabled this tool, and how MOOClets can support other tools for instructors to adapt any kind of text: explanations, emails, motivational messages, and tips for studying. Ongoing work is using these tools to adaptively personalize explanations and messages. Instead of assuming a single best explanation, new tools enable instructors to personalize different explanations to different students.

A description of AXIS is at tiny.cc/axisoverview. For more information about using AXIS, sign up at tiny.cc/useaxis
Add a comment...

Post has attachment
Wait while more posts are being loaded