Gaussian Process Preference Learning (GPPL) is considered to be the state-of-the-art algorithm for learning about a person’s preferences over a continuous domain of tunableparameter values. Unfortunately, the existing literature is vague about the underlying modular structure of the full algorithm. This hinders application of GPPL to embedded environments or to new application domains. We describe the GPPL algorithm as a Forney-style Factor Graph (FFG), which is a formal framework for distributed signal processing (and probabilistic inference) in a graph of sub-components. One of the benefits of the FFG-based specification of GPPL is that the algorithm can easily be changed to the specific circumstances of the application. We demonstrate the FFG-based GPPL algorithm in MATLAB by a perceptual tuning problem for a noise suppression algorithm.