Local Friendliness under Partial Collapse: Extremal Violations and an Information–Friendliness Complementarity

Consciousness-causes-collapse (CCC) models posit that a friend's observation acts as a collapse channel of tunable strength p on her laboratory. Local Friendliness (LF) scenarios turn this into a quantitative question, which we answer with four results. (i) A kinematic limitation: any conscious report that is non-disturbing for both super-observer settings of the deferred protocol, in the joint-instrument sense standard in the LF literature, commutes by compatibility theory with the full logical algebra and is therefore uninformative. Absolute reports of this kind are possible only vacuously, formalizing insights of Żukowski–Markiewicz and Allard Guérin et al. (ii) An exact extremal reduction: when each friend's observation dephases her laboratory with coherence retention ηᵢ and one setting per side is locked to the collapse basis, the maximal LF–CHSH value reduces exactly to a closed-form two-parameter optimization. (iii) Rigorous analytic bounds with unconditional asymptotics: an explicitly solved mirror-symmetric family governed by a cubic, together with a singular-value upper bound, brackets the symmetric value within 0.026. This yields the weak-collapse law S_LF = 2√2 (1 − p²/2) + O(p⁴) for η = √(1 − p²) and a quartic, threshold-free approach to the classical bound. (iv) A tight information–friendliness complementarity: any conscious experience that persistently identifies the friend's branch with bias D forces S_LF ≤ S_LF(√(1 − D²)), where persistence means the information is carried by a subsystem of a dilation of the collapse channel. The bound is saturated by the weak-measurement instrument, and the known qualitative no-go is recovered as the D = 1 endpoint. The predicted S_LF(p) sweep and a collapse-strength-independent record–record correlator are directly testable on existing quantum-processor Local Friendliness platforms with shot budgets of approximately 100,000 measurements.