This document presents a study on a soft decision low-density parity-check (LDPC) decoder implemented using the Bahl-Cocke-Jelinek-Raviv (BCJR) algorithm, focusing on accurate channel state information (CSI) estimates under uncertainty. The authors propose a Bayesian approach to improve the soft decoding performance by integrating uncertainty in CSI with a new approximation that maintains complexity while enhancing error rates. Experimental results demonstrate that the proposed method outperforms the standard maximum likelihood BCJR equalizer, particularly in high signal-to-noise ratio scenarios.