Can We Have Our Cake and Eat It, Too?
Several states are exploring the possibility of a through-course assessment, meaning that a student will be assessed multiple times a year and those scores will be rolled up to a summative score at the end of the year. My former colleague, Brian Gong, provides a nice summary of why traditional approaches to through-course assessment have failed https://www.nciea.org/blog/state-testing/why-has-it-been-so-difficult-develop-viable-through-year-assessment.
However, I will argue that his claim about the purpose of the summative assessment needs to change when considering a through-course assessment. Rather than asking the summative assessment to determine “what a student knows and can do at the end of the year,” we need to alter our expectation of the summative assessment to a single score. The original purpose of the accountability test was to determine what percentage of students in each school, district, and state met expectations of their state standards. This information was to be used to determine where additional supports (or sanctions, under No Child Left Behind) were needed. Later waivers of NCLB provisions required states to provide instructional feedback from these assessments. However, instructional feedback is not the primary purpose, nor are summative assessments well designed to provide it.
If we switch our thinking to say that all we need from a summative assessment is an estimate of student understanding of grade-level content, or a theta score in psychometric-speak, more options are opened.
The option I believe needs more exploring is one that involves the use of a computerized adaptive test. Once items that cover the entire content of a school year have been placed on a scale, the summative purpose can be met simply be determining where on the scale the student falls at the end of the year.
The problem Brian stated about determining an appropriate scope and sequence that works across a state still exists. However, with the help of our educators, I believe this is a challenge that can be overcome. Much work has been done on learning progressions over the past 20 years, and the work is starting to drive curricular decisions. Many of the differences in scope and sequence that exist today are not as drastic as one might think. Overlapping content assessed at each administration can help alleviate those concerns.
For example, if a state has ten standards, A thorough J, the first assessment could measure standards A-D, the second C-F, the third E-H, and the fourth G-J. Each test would be shorter than a traditional summative test, although the four tests combined would ultimately be longer. All standards would still be assessed—meeting the Federal requirement for full alignment—but across the year instead of in one administration.
Each administration could therefore provide strong instructional feedback to educators as it focuses on a smaller number of standards. The student’s ability level, theta score, would be saved as a starting point for the subsequent administration. Thus, every student would have a test personalized to where they were in the learning progression, and no restrictions would be placed on how much a student could grow from one administration to the next. The length may therefore vary across students with the dual goal of assessing the appropriate content for that period and ensuring an accurate theta estimate with a high level of reliability.
The theta score of the final assessment would then become their summative score which would then be categorized into a performance level. There would be a myriad of ways a student could reach a proficient score, but ultimately, if we separate our accountability purpose from the instructional one, a through-course model becomes more feasible.