Latent Directions: A Simple Pathway to Bias Mitigation in Generative AI
This addresses bias issues in generative AI for developers and users, offering a novel method that is incremental but enhances existing mitigation techniques.
The paper tackles bias mitigation in text-to-image generative AI by learning latent directions to modify initial Gaussian noise, enabling diverse and inclusive synthetic images without altering prompts or embeddings, and demonstrates adaptability to scenarios like geographical biases with linear combination capabilities.
Mitigating biases in generative AI and, particularly in text-to-image models, is of high importance given their growing implications in society. The biased datasets used for training pose challenges in ensuring the responsible development of these models, and mitigation through hard prompting or embedding alteration, are the most common present solutions. Our work introduces a novel approach to achieve diverse and inclusive synthetic images by learning a direction in the latent space and solely modifying the initial Gaussian noise provided for the diffusion process. Maintaining a neutral prompt and untouched embeddings, this approach successfully adapts to diverse debiasing scenarios, such as geographical biases. Moreover, our work proves it is possible to linearly combine these learned latent directions to introduce new mitigations, and if desired, integrate it with text embedding adjustments. Furthermore, text-to-image models lack transparency for assessing bias in outputs, unless visually inspected. Thus, we provide a tool to empower developers to select their desired concepts to mitigate. The project page with code is available online.