Working Memory
Definition
Working memory is the cognitive system responsible for temporarily holding and actively manipulating information needed for current tasks. Capacity is famously limited (Miller's classic 7±2, refined to roughly 4±1 chunks in modern research). When working memory load exceeds capacity, attention narrows, decision quality drops, and cognitive performance degrades — a key reason email overload feels overwhelming.
The capacity limit
George Miller’s 1956 paper “The Magical Number Seven, Plus or Minus Two” established the foundational finding: working memory holds roughly 7±2 items. Modern research, including Nelson Cowan’s work, refined the number downward — most adults can actively manipulate 4±1 chunks of unrelated information.
The key word is “manipulate.” Working memory isn’t passive storage (that’s long-term memory). It’s the active workspace where you hold items to reason about, compare, and use. Holding a 10-digit phone number long enough to dial it is working memory. Comparing three job offers in your head is working memory.
How it relates to email overload
An inbox of 50 unread emails is 50 open loops, each demanding a slice of working memory. Even unopened, the inbox creates a background cognitive load: your brain holds “there’s 50 things I haven’t dealt with” as an active concern.
This is why inbox count correlates with subjective overwhelm independent of actual workload. The cognitive cost isn’t the email volume per se; it’s the working memory tax of holding open loops.
Why “trusted system” works
David Allen’s GTD insight that a “trusted external system” frees the mind is precisely about working memory. Items captured in an external system that the brain trusts can be released from working memory; items the brain doesn’t trust the system to surface stay in working memory.
A todo list you don’t review fails this test — the brain doesn’t trust it to surface items, so they stay in working memory. A todo list you do review (the weekly review discipline) passes the test, and working memory clears.
What overloads working memory
Three patterns:
- Too many unresolved decisions. Each pending decision occupies a working memory slot.
- Frequent context switches. Each switch requires loading new context, often before the previous context fully cleared.
- Holding details that should be external. Phone numbers, deadlines, todo items — anything that could be written down but isn’t.
How AI assistants reduce working memory load
AI email assistants reduce working memory load in two ways:
- Triage removes items from active concern. An email handled by the AI (archived, drafted, deferred to task) is one less open loop.
- Daily briefings replace “what’s in my inbox” with “what needs my decision.” The brain holds 3-5 items needing judgment, not 100 items competing for attention.
The cognitive relief is often the first thing users notice when they adopt an AI assistant. The inbox volume hasn’t changed; the working memory load has.
Where alfred_ fits
alfred_ explicitly reduces working memory load by handling 75% of email autonomously and presenting the remaining 25% in a structured Daily Brief. The user holds in working memory only the decisions that need them — typically 3-7 items, comfortably within capacity. The 50+ open email loops that previously occupied background attention are handled or queued.
What working memory isn’t
It isn’t IQ — high IQ correlates with but doesn’t equal working memory capacity. It isn’t long-term memory — that’s a different system with very different capacity. And it isn’t trainable in a strong sense; brain training games don’t reliably expand it. The lever is reducing load, not expanding capacity.