Papers by Joan Nwatu
Bridging the Digital Divide: Performance Variation across Socio-Economic Factors in Vision-Language Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Among the minority groups under-represented in AI, data from low-income households are often overlooked in data collection and model evaluation. |
| Approach: | They evaluate the performance of a vision-language model on a geo-diverse dataset . they highlight insights that can help mitigate these issues and propose actionable steps for economic-level inclusive AI development. |
| Outcome: | The proposed model performs lower for the poorer groups than the wealthier groups across topics and countries. |
Uplifting Lower-Income Data: Strategies for Socioeconomic Perspective Shifts in Large Multi-modal Models (2025.naacl-long)
Copied to clipboard
| Challenge: | Unequal representation of cultures and socioeconomic groups in training data leads to biased Large Multi-modal (LMM) models. |
| Approach: | They propose and evaluate several prompting strategies that use non-English, geographic, and socioeconomic attributes to improve LMM model performance on underrepresented data. |
| Outcome: | The proposed prompts favor retrieving topic appearances from low-income data on lower-income datasets. |