DebiasPI: Inference-time Debiasing by Prompt Iteration of a Text-to-Image Generative Model

• URL: http://arxiv.org/abs/2501.18642v1
• Date: Tue, 28 Jan 2025 23:17:20 GMT
• Title: DebiasPI: Inference-time Debiasing by Prompt Iteration of a Text-to-Image Generative Model
• Authors: Sarah Bonna, Yu-Cheng Huang, Ekaterina Novozhilova, Sejin Paik, Zhengyang Shan, Michelle Yilin Feng, Ge Gao, Yonish Tayal, Rushil Kulkarni, Jialin Yu, Nupur Divekar, Deepti Ghadiyaram, Derry Wijaya, Margrit Betke,
• Abstract summary: We propose an inference-time process called DebiasPI for Debiasing-by-Prompt-Iteration. DebiasPI enables the user to control the distributions of individuals' demographic attributes in image generation.
• Score: 20.915693552625502
• License:
• Abstract: Ethical intervention prompting has emerged as a tool to counter demographic biases of text-to-image generative AI models. Existing solutions either require to retrain the model or struggle to generate images that reflect desired distributions on gender and race. We propose an inference-time process called DebiasPI for Debiasing-by-Prompt-Iteration that provides prompt intervention by enabling the user to control the distributions of individuals' demographic attributes in image generation. DebiasPI keeps track of which attributes have been generated either by probing the internal state of the model or by using external attribute classifiers. Its control loop guides the text-to-image model to select not yet sufficiently represented attributes, With DebiasPI, we were able to create images with equal representations of race and gender that visualize challenging concepts of news headlines. We also experimented with the attributes age, body type, profession, and skin tone, and measured how attributes change when our intervention prompt targets the distribution of an unrelated attribute type. We found, for example, if the text-to-image model is asked to balance racial representation, gender representation improves but the skin tone becomes less diverse. Attempts to cover a wide range of skin colors with various intervention prompts showed that the model struggles to generate the palest skin tones. We conducted various ablation studies, in which we removed DebiasPI's attribute control, that reveal the model's propensity to generate young, male characters. It sometimes visualized career success by generating two-panel images with a pre-success dark-skinned person becoming light-skinned with success, or switching gender from pre-success female to post-success male, thus further motivating ethical intervention prompting with DebiasPI.

Related papers

• MoESD: Mixture of Experts Stable Diffusion to Mitigate Gender Bias [23.10522891268232]
  We introduce a Mixture-of-Experts approach to mitigate gender bias in text-to-image models. We show that our approach successfully mitigates gender bias while maintaining image quality.
  arXiv  Detail & Related papers  (2024-06-25T14:59:31Z)
• New Job, New Gender? Measuring the Social Bias in Image Generation Models [85.26441602999014]
  Image generation models are susceptible to generating content that perpetuates social stereotypes and biases. We propose BiasPainter, a framework that can accurately, automatically and comprehensively trigger social bias in image generation models. BiasPainter can achieve 90.8% accuracy on automatic bias detection, which is significantly higher than the results reported in previous work.
  arXiv  Detail & Related papers  (2024-01-01T14:06:55Z)
• Improving Fairness using Vision-Language Driven Image Augmentation [60.428157003498995]
  Fairness is crucial when training a deep-learning discriminative model, especially in the facial domain. Models tend to correlate specific characteristics (such as age and skin color) with unrelated attributes (downstream tasks) This paper proposes a method to mitigate these correlations to improve fairness.
  arXiv  Detail & Related papers  (2023-11-02T19:51:10Z)
• Mitigating stereotypical biases in text to image generative systems [10.068823600548157]
  We do this by finetuning text-to-image models on synthetic data that varies in perceived skin tones and genders constructed from diverse text prompts. Our diversity finetuned (DFT) model improves the group fairness metric by 150% for perceived skin tone and 97.7% for perceived gender.
  arXiv  Detail & Related papers  (2023-10-10T18:01:52Z)
• FACET: Fairness in Computer Vision Evaluation Benchmark [21.862644380063756]
  Computer vision models have known performance disparities across attributes such as gender and skin tone. We present a new benchmark named FACET (FAirness in Computer Vision EvaluaTion) FACET is a large, publicly available evaluation set of 32k images for some of the most common vision tasks.
  arXiv  Detail & Related papers  (2023-08-31T17:59:48Z)
• Word-Level Explanations for Analyzing Bias in Text-to-Image Models [72.71184730702086]
  Text-to-image (T2I) models can generate images that underrepresent minorities based on race and sex. This paper investigates which word in the input prompt is responsible for bias in generated images.
  arXiv  Detail & Related papers  (2023-06-03T21:39:07Z)
• Uncovering Bias in Face Generation Models [0.0]
  Recent advancements in GANs and diffusion models have enabled the creation of high-resolution, hyper-realistic images. These models may misrepresent certain social groups and present bias. This work is a novel analysis covering and embedding spaces for fine-grained understanding of bias over three approaches.
  arXiv  Detail & Related papers  (2023-02-22T18:57:35Z)
• Auditing Gender Presentation Differences in Text-to-Image Models [54.16959473093973]
  We study how gender is presented differently in text-to-image models. By probing gender indicators in the input text, we quantify the frequency differences of presentation-centric attributes. We propose an automatic method to estimate such differences.
  arXiv  Detail & Related papers  (2023-02-07T18:52:22Z)
• How well can Text-to-Image Generative Models understand Ethical Natural Language Interventions? [67.97752431429865]
  We study the effect on the diversity of the generated images when adding ethical intervention. Preliminary studies indicate that a large change in the model predictions is triggered by certain phrases such as 'irrespective of gender'
  arXiv  Detail & Related papers  (2022-10-27T07:32:39Z)
• DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Models [73.12069620086311]
  We investigate the visual reasoning capabilities and social biases of text-to-image models. First, we measure three visual reasoning skills: object recognition, object counting, and spatial relation understanding. Second, we assess the gender and skin tone biases by measuring the gender/skin tone distribution of generated images.
  arXiv  Detail & Related papers  (2022-02-08T18:36:52Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.