Learning engineering has assembled a catalog of design patterns — each with an identified problem context, a validated solution approach, and an evidence base — that practitioners apply across diverse instructional contexts. These patterns span cognitive, structural, and adaptive dimensions of learning system design, and collectively they represent the applied translation of learning science into engineering practice. Understanding this catalog gives learning engineers a working vocabulary for design deliberation and a set of starting points that have been empirically vetted, reducing the risk of reinventing approaches that research has already evaluated. [LE-LS-GL-007]
Cognitive Load Management Patterns
The pattern family derived from Cognitive Load Theory constitutes perhaps the field's most widely applied and replicated set of design principles. The worked example pattern addresses the problem that novice learners attempting to solve unfamiliar problems from scratch must simultaneously manage the problem's content complexity and the overhead of strategy search — a combination that often exceeds working memory capacity and produces learning that is slow, error-prone, and poorly transferable. The solution is to present complete, annotated solution examples before introducing problem-solving practice, gradually fading support as competency develops. [LE-LS-AP-008] Extensive controlled research has validated this pattern across mathematics, physics, programming, and medical diagnosis, establishing it as one of the most robust findings in instructional design. [LE-LS-PP-004]
Companion patterns address related load management challenges. The split-attention reduction pattern specifies that when instructional materials require learners to mentally integrate information from multiple spatially separated sources (such as a diagram and a text description placed on different pages), the materials should be redesigned to place related information in close physical proximity. The redundancy reduction pattern identifies cases where including the same information in multiple modalities (spoken narration plus identical on-screen text, for instance) forces learners to process and filter redundant channels rather than learning from the content — and prescribes elimination of the redundant source. Both patterns have generated consistent learning gains in controlled studies and are standard features of contemporary multimedia learning guidelines. [LE-LS-AP-008]
Adaptive Sequencing and Mastery Patterns
The knowledge component decomposition and mastery gating pattern — originating in the Cognitive Tutor research program at Carnegie Mellon — addresses the problem of heterogeneous skill development within a learner population. Rather than advancing all learners through content at a uniform pace, this pattern specifies decomposing target skills into individual knowledge components, assessing each component independently through targeted practice items, and gating advancement to subsequent content on demonstrated mastery of each component. [LE-LS-AP-001] Bayesian Knowledge Tracing provides the probabilistic infrastructure that operationalizes this pattern in deployed systems, continuously estimating mastery probabilities from response data and triggering mastery-based transitions. [LE-LS-AP-011] The decade-long Cognitive Tutor research program documented the effectiveness of this pattern in algebra and geometry education, producing consistent learning gains in randomized controlled studies conducted in real schools. [LE-LS-AP-014]
The active practice prioritization pattern responds to a different design challenge: how to allocate limited learner time between content exposure (reading, watching, listening) and practice (problem-solving, applying, generating). Large-scale empirical research leveraging platform telemetry across seven online courses demonstrated that active practice is causally linked to learning gains in a way that passive content consumption is not — a finding robust enough to function as a design principle. [LE-LS-AP-012] The implication for learning engineers is that time allocation decisions should systematically favor practice over exposure, and that instructional designs that bury practice items at the end of long content sequences are likely underperforming relative to designs that intersperse practice throughout. Kenneth Koedinger's broader research program established that learning environments should be instrumented to measure practice behavior specifically, not just content completion, in order to monitor whether this pattern is being implemented effectively. [LE-LS-PP-005]