Introduces a 21-page paper focused exclusively on the Copycat Workspace,
proposing three graph-metric substitutions for hardcoded constants and
unifying them under a common information-theoretic framework:
- Clustering coefficient as support vector (replaces 0.6^(1/n^3) factor)
- Betweenness centrality as salience weights (replaces fixed 0.2/0.8 ratios)
- Subgraph density as length factors (replaces step function {5,20,60,90})
Central contribution: all three metrics are ratios actual/maximum in [0,1],
interpretable as probabilities, yielding Shannon self-information in bits.
This common unit enables principled arithmetic via workspace entropy
H(W) = sum(I_C) + sum(I_B) + sum(I_rho), with the conjecture T ∝ H(W).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
b5323b5a74