Definition
Cognitive load is the total demand placed on working memory and executive resources during the performance of a task. The framework — developed by John Sweller and collaborators since the 1980s — partitions load into three additive components: intrinsic load (demand inherent to the task itself, determined by the number of interacting elements that must be held simultaneously to make sense of the material), extraneous load (demand introduced by the way information is presented rather than the task itself, including poor instructional design, irrelevant detail, and split attention across non-integrated sources), and germane load (productive demand corresponding to the construction of schemas — the durable cognitive structures that allow chunking and reduce future intrinsic load).
Total load is bounded by working-memory capacity (Cowan's revised estimate of ~4 chunks, applicable when no rehearsal strategy is engaged). When total load exceeds capacity, performance degrades discontinuously: comprehension fails, errors multiply, and the learner falls back on whatever heuristic responses are available without engaging the integrative processing that intrinsic and germane components require. Critically, the failure mode is invisible to the loaded subject: subjective feelings of confidence and comprehension dissociate from objective performance, with high-load conditions reliably producing miscalibrated overconfidence.
Cognitive load can be measured behaviorally (secondary-task performance, reaction time variance), physiologically (pupillometry, EEG theta/alpha ratios, heart-rate variability), and via self-report (Paas mental-effort scale, NASA-TLX), with reasonable convergent validity across measures.
Context
Within Prismatic, cognitive load is the primary mechanism by which kompendium EA-1.x cognitive-overload techniques achieve their effects. Information saturation, time pressure, parallel-decision induction, and irrelevant-detail injection all operate by inflating extraneous load until total load exceeds capacity, at which point the target's epistemic-evaluation routines become unavailable and processing collapses to heuristic shortcuts the manipulator has pre-positioned to be exploited.
The Decision Engine treats cognitive-load proxies as explicit inputs to confidence calibration: session duration, decision density, document length, time pressure indicators, and concurrent-task counts all enter the load model, and decisions made under elevated load receive proportionally reduced weight. Platform interface design follows cognitive-load principles deliberately: integrated rather than split presentation, progressive disclosure, signposting of intrinsic complexity, removal of decorative extraneous elements.
The M5M doctrine (5-minute task chunks with 1-minute breaks) is grounded in cognitive-load research: load accumulates on timescales of minutes, recovery requires explicit rest, and load is most dangerous when it has been sustained long enough to compromise self-monitoring of one's own load state.
Interactive: The Bandwidth Container
Working memory holds about four chunks. Intrinsic load is task-inherent, extraneous load is presentation overhead, germane load builds the schemas that future-proof understanding. When the three stack above the ceiling, processing collapses to heuristic shortcuts — and the loaded subject does not feel it happen.
Demand inherent to the task
Demand from poor presentation / clutter
Productive load building schemas
See Also
- Working Memory Limit — Capacity bound
- Dorsolateral PFC — Neural substrate
- Anterior Cingulate — Conflict monitoring under load
- System 1 / System 2 — Load forces System 1
- Attentional Control — Allocates limited resources
- Kompendium — EA technique taxonomy
- Epistemická anihilace — Threat model
Connect & Contribute
Created by Tomáš Korcak (korczis) | Open Source under GHL