Incremental & Semi-Supervised Learning

Why this map looks & behaves the way it does

1. Mind maps measurably help. Medical students using mind maps recalled facts better a week later (Farrand, Hussain & Hennessy 2002, Medical Education); meta-analyses across 55+ studies, n≈5,800–11,800, find consistent retention gains, g ≈ 0.58 (Nesbit & Adesope 2006; Schroeder et al. 2018).
2. Active use beats passive viewing — working with/creating maps g = 0.72 vs just reading them g = 0.43 (Schroeder et al. 2018). → After a pass, redraw the four branches from memory on paper, then check.
3. Retrieval practice beats re-studying — even for inference questions, recalling beats re-reading concept maps (Karpicke & Blunt 2011, Science). → That's Recall mode: details blur, you answer out loud, click to check yourself.
4. Self-explanation deepens learning. Prompting learners to explain material to themselves improves understanding (Chi et al. 1994; elaborative interrogation, Pressley et al. 1987). → Hover any node — or the line leading to it — for a plain-words explanation. Before reading it, try to explain the node yourself.
5. Color works when it carries meaning. Color coding improved retention & transfer by guiding attention to related info (Ozcelik et al. 2009, eye-tracking; Dzulkifli & Mustafar 2013). → Every color is semantic — red = the disease (forgetting), green = its cure, orange = the model asking…
6. Chunking & hierarchy. Working memory holds ~4 chunks (Cowan 2001); hierarchical organization multiplies recall — when the tree mirrors the semantic structure (Bower et al. 1969); big picture first, detail on demand (Mayer's segmenting principle). → 4 branches; problem→solution chains are nested (cures live under their disease), not laid side by side.
7. Space it out. Spaced reviews beat massed cramming (Cepeda et al. 2006). → Recall-mode pass today, tomorrow, and in 3 days. The stable layout makes node positions themselves retrieval cues.