Lesson 2 — Spreading activation

Duration: ~45 min · Mode: shared read-only project (classroom) · Instrument: the Graph tab's pulse animation and link structure

Learning objectives

1. Explain spreading activation: recall propagates along associative links.
2. Predict which "neighbor" concepts a retrieval will warm up.
3. Connect priming effects in humans to graph-based retrieval in machines.

The science in one paragraph

Hearing doctor makes you faster to recognize nurse — that's priming, and Collins & Loftus (1975) explained it with a semantic network where activating one node sends activation flowing outward along links, weaker with distance. Anderson (1983) built the same mechanism into ACT-R. Memory Layer's graph is a real semantic network: when a memory is used, a boost also flows to memories linked to it — halving per hop and split across a hub's links, so tight clusters warm together but a heavily connected hub cannot inflate everything it touches. Retrieval itself is graph-aware: results can include memories reached through links, not just direct text matches.

Prep (2 min)

Same setup as lesson 1. The graph already carries some heat from the previous lesson — useful, mention it.

Guided activity

1. See the network (5 min). On the projector, click any memory node. Watch the pulse particles travel outward along its links — that animation is the spreading-activation mechanism drawn literally. The links exist because curation recorded real relations (same topic, same evidence, supersedes, summarizes).

2. Predict, then fire (15 min). Pick a well-connected node (the consolidation memory works well). Students write down which two neighbors they predict will warm up if the class queries about it. Then everyone asks variations of "What does the value gate decide?" / "When does consolidation trigger?". Refresh: check predictions against the graph. The neighbors warmed without being cited — activation reached them through links alone.

3. Hops and halving (8 min). Use the graph's depth stepper to expand one and then two hops from the hot node. Board math: a boost of 1.0 becomes 0.5 one hop out, 0.25 two hops out — and is also divided across the hub's links. Ask: why divide? (Otherwise a hub with fifty links would smear activation everywhere and destroy the ranking signal.)

4. Human priming tie-back (10 min). Classic word-pair demos (bread → butter; ocean → ?). The class experiences priming, then maps it onto what they just watched: cue = query, node = memory, link = association, warmth = readiness to recall.

5. Where it pays off (5 min). Show one query whose answer includes a memory that doesn't share the question's words. Graph-aware retrieval found it through the network — the machine equivalent of "that reminds me of…".

Exit ticket

"Give one example from your life where remembering A made you remember B, and name the link between them."

References

• Collins, A. M. & Loftus, E. F. (1975). A spreading-activation theory of semantic processing. Psychological Review, 82(6).
• Anderson, J. R. (1983). A spreading activation theory of memory. Journal of Verbal Learning and Verbal Behavior, 22(3).
• Meyer, D. E. & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words. Journal of Experimental Psychology, 90(2) — the priming effect.