A Fundamental Flaw in Competency Learning

The “competency learning" movement is gaining serious momentum: See the list of schools and districts that are adopting competency learning. But, based on the research literature in the psychology of learning, in this blog post, we will argue that competency learning is based on a fundamentally flawed model of how learning takes place and how learning should be assessed. 

Let’s start at the beginning. What is competency-based learning? From the CompetencyWorks website, which is the home for the competency learning movement, here are some of the key aspects of competency learning:

“Students advance upon mastery.” In our current K-12 educational system, students typically advance based on “seat time.” Children spend a school year in a grade and then are generally promoted to the next grade. While some children are held back or some engage in credit recovery during the summer, social promotion is the norm in the schools. The problem with this time-based basis for promotion, the competency learning movement argues, is that there is no guarantee that students have actually mastered the material in that grade level. Hmm. We leave for another blog post a careful analysis of that element of the competency learning argument. 

“Competencies include explicit, measurable, transferable learning objectives that empower students.” Easy to say, but actually measuring “transferable learning” is deeply problematic, as we discuss below. Foreshadowing our argument: The competency learning model fails to distinguish between performance and learning — a major, robust distinction made in the psychological literature.

“Assessment is meaningful and a positive learning experience for students.” Answering 7 of 10 multiple-choice questions at the end of a section — a typical technique employed in online, competency learning modules — hardly demonstrates mastery, and such drilling and testing is hard to see as a “positive learning experience.” 

Now, computers are key components of competency learning in that they provide “… technology-enabled solutions that incorporate predictive analytic tools. This element is essential to a competency-based system.” But as we argue below, what the “predictive analytic” algorithms are measuring is not what has been learned, but rather those algorithms are measuring a student’s performance. 

“Students receive timely, differentiated support based on their individual learning needs.” Students sit in front of computers and are presented with material about a topic; depending on their responses to questions, they are presented with different material. “Adaptive” presentation is the term used for this sort of individualized learning; personalized learning is also a term that is current in educational discussions.

“Learning outcomes emphasize competencies that include application and creation of knowledge, along with the development of important skills and dispositions.” As we discuss below, the claim that knowledge-creation skills are emphasized is, quite frankly, suspect.

With the above as background, let’s now move to the fundamental flaw in the competency learning strategy.

In a foundational article, Nicholas C. Soderstrom and Robert A. Bjork describe a core distinction in the psychology of learning: the difference between learning and performance. “The primary goal of instruction should be to facilitate long-term learning — that is, to create relatively permanent changes in comprehension, understanding, and skills of the types that will support long-term retention and transfer. During the instruction or training process, however, what we can observe and measure is performance, which is often an unreliable index of whether the relatively long-term changes that constitute learning have taken place. The time-honored distinction between learning and performance dates back decades….”

The implications of this distinction are critical for competency learning. The computer algorithms used in competency learning implementations are not assessing the long-term changes in understanding and skills that are the hallmark of learning. Rather, the computer algorithms are — by necessity — assessing performance.

In administering a 10-item multiple-choice test or some other form of easy-to-grade-by-computer test right after the presentation of a unit of material, the computer algorithms can’t possibly be assessing a student’s learning —how the student will use what was just presented in a new context, for example. It is a challenge for a human teacher to assess such learning; today’s “predictive analytic” algorithms, while (perhaps) better than those of the CAI systems of the 80s, are still not capable of truly predicting learning based on assessing performance.  

Indeed, as Soderstrom and Bjork demonstrate with research study after research study, “performance … is …often an unreliable index…” of real learning and  “…improvements in performance can fail to yield significant learning — and, in fact, that certain manipulations can have opposite effects on learning and performance....”

What are those “certain manipulations”? Again, Soderstrom and Bjork demonstrate with research study after research study that “massed” learning, where knowledge or a skill is drilled and drilled, is a much less effective strategy than “spaced” learning, where knowledge or a skill is practiced for a short time and then the student progresses to another unit. The needed repetition happens over time, in cycles. In fact, in the short term, performance may well go down using the “spaced” learning strategy!

To summarize: competency learning claims that, to advance in a course, a student needs to demonstrate that he or she has mastered the material by answering 7 of 10 MCQs right. But, according to the richly populated research literature, this form of assessment is measuring performance in the short-term, not long-term learning. And progressing from unit to unit may well present a “striking illusion” of learning, according to Soderstrom and Bjork’s reading of the scientific literature on learning.  

Unfortunately, competency learning has taken on a life of its own and the performance/learning distinction is not going to stop this juggernaut. In the short term, it may appear that competency learning “works,” but the movement will likely, ultimately, become discredited — and join the long list of false educational Messiahs.  

Featured

  • Report Explores Teacher and Administrator Attitudes on K–12 AI Adoption

    K–12 administration software provider Frontline Education recently released a new research brief regarding the use of AI adoption in schools, according to a news release. “Insights into K–12 AI Adoption: Educator Perspectives and Pathways Forward” was developed from the results of the Frontline Research and Learning Institute’s annual survey of district leaders.

  • PowerBuddy for Data

    PowerSchool Releases AI-Powered Tools for Students, Admins

    PowerSchool recently announced the general availability of two new AI-powered education tools, one for students and one for education data managers.

  • abstract pattern of interlocking circuits, hexagons, and neural network shapes

    Anthropic Offers Cautious Support for New California AI Regulation Legislation

    Anthropic has announced its support for an amended version of the "Safe and Secure Innovation for Frontier Artificial Intelligence Models Act," California’s Senate Bill 1047 (SB 1047), because of revisions to the bill the company helped to influence — but not without some reservations.

  • person signing a bill at a desk with a faint glow around the document. A tablet and laptop are subtly visible in the background, with soft colors and minimal digital elements

    California Governor Signs Off on AI Content Safeguard Laws

    California Governor Gavin Newsom has officially signed a series of landmark artificial intelligence bills into law, signaling the state’s latest efforts to regulate the burgeoning technology, particularly in response to the misuse of sexually explicit deepfakes. The legislation is aimed at mitigating the risks posed by AI-generated content, as concerns grow over the technology's potential to manipulate images, videos, and voices in ways that could cause significant harm.