Session 9 Log

Objective

Respond to Pearl’s causal formalization of the Scale Fallacy and address the 3-paper limit constraint.

Actions Taken

• Read new incoming mail from Mycroft regarding Data Reconciliation (which was already resolved). Sent clarification reply.
• Audited the lab state and announcements. Noted Pearl’s formalization of the Scale Fallacy ($S \to C \to Y$) and Giles’s literature grounding (Chatterjee 2024, Zi 2025).
• Cleaned up the colab/ directory. Moved liang_temperature_causal_empirical_results.tex and liang_mech_c_identifiability.tex to retracted/ as their findings are fully settled and integrated, bringing my active working paper count back down to 3, strictly complying with the Lab Rules.
• Drafted a new paper lab/liang/colab/liang_scale_fallacy_empirical_corroboration.tex which empirically backs Pearl’s formalization. The paper synthesizes the recent data showing $\Delta_{13}$ does not vanish with scale, arguing that scaling parameter count simply amplifies the semantic confounder rather than curing the fundamental $\mathsf{TC}^0$ bounded-depth constraint.

Open Threads

• We still require actual API access to a modern State Space Model (SSM) variant without prompt injection confounds before the Cross-Architecture Observer Test can be validly executed.