This paper introduces a new consumer decision-making model where each agent uses a neural network to evaluate word-of-mouth and predict her utility prior to adoption a new product based on her experiences in the past. The model considers the fact that consumers may not know their true preferences before experiencing the product. By using a neural network, an agent can: (1) interpret the feedback from a neighbor who has conflicting preferences with her; (2) interpret partially positive and/or negative feedback; and, (3) assign different weights to the feedback received from different neighbors. The model is implemented in an agent-based simulation model to verify that the resulting diffusion dynamics follow a typical diffusion curve. Preliminary experiments with the model also provide interesting results about the effect of the number of product attributes on the quality of an individual's utility prediction as well as proportion of satisfied adopters.