We present a model selection framework for the extraction of the CKM matrix element $|V_{cb}|$ from exclusive $B \to D^* l \nu$ decays. By framing the truncation of the Boyd-Grinstein-Lebed (BGL) parameterization as a model selection task, we apply the Akaike Information Criterion (AIC) to choose the optimal truncation order. We demonstrate the performance of our approach through a comprehensive toy study, comparing it to the Nested Hypothesis Test (NHT) method used in previous analyses. Our results show that the AIC-based approach produces unbiased estimates of $|V_{cb}|$, albeit with some issues of undercoverage. We further investigate the impact of unitarity constraints and explore model averaging using the Global AIC (gAIC) approach, which produced unbiased results with correct coverage properties. Our findings suggest that model selection techniques based on information criteria and model averaging offer a promising path towards more reliable $|V_{cb}|$ determinations.