Enhanced prompts for ethical representation: Bias mitigation in text-to-image generation models
Abstract
Text-to-Image (T2I) generation model is opening up new applications that assist designers, artists, and end users of productivity software in general by producing highly photo-realistic graphical content based on a brief description. However, as these models are trained on a vast quantity of online data, there is a risk that the creation process might include undesirable biases. We show that three widely used text-to-image systems—DALL-E 3 XL v2, ByteDance&s;s Hyper Stable Diffusion, and Stable Diffusion—have a bias. After assessing these models, we identified a number of important features: (1) The kind of bias that can exist in T2I models has been highlighted. (2) After presenting several approaches for evaluating bias, it was shown that the CLIP score model outperformed the others. (3) We have demonstrated how various bias reduction strategies affect the outcomes. (4) We have demonstrated the beneficial effects of moral interventions on prompts that yield meaningful outcomes and fair visuals. Finally, our work shows that prompt enhancement has a considerable impact on the range of text-to-picture generations in the presence of moral interventions. Highlighting the significant advancements made by DALLE-3-xl-v2, Hyper Stable Diffusion, and Stable Diffusion after text augmentation. We identify future research directions that it is possible to create a machine learning model that can automatically augment prompts by determining which prompts are likely to provide biased…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.