Tolerating Approximate Answers about Student Learning
Presented by听Derek Briggs at the听Oxford University Centre for Educational Assessment on听May 24, 2018.听
The statistician John Tukey once wrote 鈥淔ar better an approximate听answer to the right question, which is often vague, than an exact听answer听to the wrong听question, which can always be made precise.鈥 Some prime examples of vague questions in education, for which only approximate answers are likely available: What are students learning? How much are they learning? Are they learning enough? These questions, which require considerable unpacking, stand in contrast to more precise questions that can be answered with greater confidence: 听How reliable are the scores on this test? Is a test score high enough to infer mastery of the content domain? Is the score predictive of success on other related tests? These latter questions, while not wrong per se, represent the sorts of things that psychometricians think that people should care about, rather than the sorts of questions they actually care about.听There is good reason to debate the appropriate role for psychometrics to play in contexts where there is a desire to make inferences about student growth.听 In this talk, I use recent and ongoing research on learning progressions in mathematics and science to illustrate how measurement provides a valuable frame of reference in our attempts to answer questions about student learning.听Yet I also emphasize the danger of overselling measurement as an outcome when theories of learning are nascent and scoreable items are in short supply.