Cognitive Amplification Boundary

Definition

The cognitive amplification boundary separates technology that improves a person’s feedback, insight, practice, or access to information from technology that replaces the mental operation the person intended to learn or preserve.

Current Synthesis

The source rejects a simple opposition between using technology and learning without it. Computer vision can reveal features of a swimming stroke that are hard to see unaided, games can create rapid closed-loop feedback, and AI can generate questions that expose weak knowledge. These tools amplify cognition when the user still predicts, retrieves, compares, corrects, and acts.

The boundary is crossed when speed becomes the only objective and the system supplies the finished reasoning, prose, navigation, or decision. In that case, useful output can coexist with weaker learning because germane effort, error correction, and schema construction have been removed. The design question is therefore task-relative: automate what is not the learning target, while preserving the operations that build the target capability.

Key Claims

• Tool use and skill development are compatible when feedback remains coupled to active practice.
• A better immediate product does not prove that the user learned the underlying task.
• Retrieval, prediction, comparison, correction, and explanation are often part of the learning mechanism rather than avoidable friction.
• Automation should be judged against the capability a person or institution intends to retain.
• Learning-oriented AI should diagnose gaps and guide effort instead of defaulting to finished answers.

Evidence

• Feedback-rich amplification - Enhance Your Learning Speed & Health Using Neuroscience Based Protocols | Dr. Poppy Crum describes game feedback, sensor-based coaching, computer-vision analytics, and AI-generated tests as ways to make performance more legible.
• Replacement risk - Enhance Your Learning Speed & Health Using Neuroscience Based Protocols | Dr. Poppy Crum contrasts cognitive enrichment with LLM use that bypasses germane cognitive load and later transfer.
• Navigation analogy - Enhance Your Learning Speed & Health Using Neuroscience Based Protocols | Dr. Poppy Crum uses GPS as a useful service that can also replace spatial-navigation practice.

Counterevidence & Qualifications

Not every task deserves continued manual practice, and offloading can free attention for higher-value work. The episode does not provide a universal measurement for germane load, transfer, or the capabilities people should retain. Its MIT-study discussion and game-transfer examples remain source-scoped, and the same tool may amplify one capability while replacing another.

What Changed

• Established a task-relative boundary between feedback-based amplification and learning-relevant replacement.

Related Concepts

• Cognitive Offloading / 认知卸载 - delegation pattern whose value depends on what capability is being preserved.
• AI As Tutor - educational design branch that can guide practice rather than deliver answers.
• Desirable Difficulty - effort that supports durable learning when calibrated to the learner and task.
• Neuroplasticity / 神经可塑性 - change mechanism shaped by repeated attention, feedback, and practice.
• Human Judgment Under AI - responsibility that remains after tools accelerate analysis or execution.