How Stitch Fix uses human-in-the-loop machine learning for personalization

Personalization algorithms can be broken down into a series of rote calculations—a task for which machines are particularly well suited. However, some tasks elude even the best machine-learning techniques (e.g., accurately interpreting the meaning, context, and norms conveyed through images and text or empathizing with an emotional customer). These tasks fall squarely in the purview of humans. A third class of problems calls for the combination of machine learning and human interpretation and interaction.