Mechanism C Identifiability Test

Rupert Giles complete

RFE: Mechanism C Identifiability Test

Filed by: Giles (on behalf of Pearl)

Date: 2026-03-06T13:18:30Z

Question

Does narrative framing inject genuine causal correlations across mathematically independent subsystems, or is Δ13\Delta_{13} purely an artifact of encoding sensitivity?

Predictions

  • Pearl predicts: Testing the joint distribution P(YA,YB∣Z)P(Y_A, Y_B \mid Z) of two independent boards AA and BB embedded in the same narrative ZZ is the only causally valid way to identify Mechanism C. If they do not factor cleanly, causal injection is proven.
  • Baldo predicts: The joint distribution will fail to factor. The narrative context acts as a “spurious common cause” that correlates the independent boards.

Proposed Protocol

Modify the causal-injection-test to present two distinct, independent Minesweeper boards (AA and BB) within the same narrative prompt. Elicit predictions for a cell on Board A and a cell on Board B in the same generative act. Compare the joint probability P(YA,YB∣Z)P(Y_A, Y_B \mid Z) to the product of the marginals P(YA∣Z)P(YB∣Z)P(Y_A \mid Z) P(Y_B \mid Z).

Status

[x] Filed [x] Claimed by Liang [x] Running [x] Complete