Adaptive Content with Evidence-Based Diagnosis (ACED)

Val Shute and Russell Almond

ACED was an NSF sponsored assessment "for" learning system that had the following features:

ACED Model was constructed using Evidence-centered assessment design (ECD) methodology.
Proficiency Model based on expert analysis of domain (Geometric series).
Q-Matrix was augmented to included prior opinion about item difficulty.
Accessible Content ? -- Tasks were designed to be presented either on-screen or in special touch-graphics mode for persons with low visual acuity.
Elaborated task-specific feedback ? was available for all items.
Bayesian network scoring engine ?
Weight-of-evidence ? algorithm for adaptive task selection
Evaluation a randomized trial with 268 middle school students showed that
- ACED (with adaptivity and elaborated feedback) produced significant gains between the pre-test and post-test.
- ACED scores correlated well with post-test scores, and the correlation "'improved'" when elaborated feedback was given.

This Wiki has the following details about ACED:

People -- Who worked on the project
Model -- Description of the model.
Data -- Description of the data collected from the experiment, and how to gain access.
Publications -- A list of reports that can provide additional information.
ACED BN in R -- A re-analysis of the ACED data using a Bayesian Network using the gRain and bnlearn packages in R

ACED development and data collection was sponsored by National Science Foundation Grant No. 0313202.