AI model evaluation
How new AI tools should be evaluated when they enter assessment, learning, and instruction. The work asks what evidence a model must provide before its outputs can support educational decisions.
Research
Research that makes the basis of a decision visible.
The center studies assessment as a system of design choices, evidence, interpretation, and use. Its projects range from item development to the evaluation of new technologies.
Portfolio
Seven areas of current and continuing workA useful assessment does more than produce a score. It makes the meaning of that score defensible, the limits of the evidence visible, and the next decision clearer.
Test development
Designing items, tasks, and forms that produce dependable evidence for their intended use. This includes attention to content representation, accessibility, scoring, and form assembly.
Validity studies
Examining whether interpretations and decisions are supported by evidence. Studies may compare measures, examine alignment, or test assumptions behind a proposed use.
Cognitive diagnostic modeling
Using fine grained models to describe what learners know and where instruction can respond. These models aim to make diagnostic information more useful than a single total score.
Teacher evaluation
Studying growth and value added measures with attention to fairness, stability, and responsible use. The work considers both statistical behavior and the policy setting.
State science assessments
Supporting science assessment programs as standards, item designs, and reporting expectations change. The focus is on evidence that fits the intended construct.
Peer review
Independent technical review that helps agencies strengthen assessment systems and explain their choices. Reviews can address design, analysis, documentation, or reporting.
Project example. A 2012 literature review examined how student characteristics, including computer familiarity and test anxiety, related to performance on computer based tests compared with paper tests. The review was designed to help assessment leaders think about who may be affected when a test changes mode.
How a project takes shape
Frame the decision
Define the interpretation or action the evidence must support, then identify the assumptions that need attention.
Build the evidence
Choose methods that fit the question, whether that means test design, psychometric analysis, a literature review, or an evaluation study.
Make it usable
Report findings in language that supports good decisions and states clearly what the evidence does not establish.