Isha Shamim (Tata Consultancy Services Limited)*; Rekha Singhal (TCS)

Yunsung Chung (Tulane University)*; Janet Wang (Tulane University); Jihun Hamm (Tulane University)

Harshit Verma (BITS Pilani Hyderabad); M Bhargav (BITS Pilani Hyderabad); Ritvik – (BITS Pilani Hyderabad); Chetana Dr Gavankar (BITS Pilani)*; Prajna Devi Upadhyay (BITS Pilani Hyderabad)

Kush Juvekar (Independant)*; Anupam Purwar (Independent)

Co-creating Civic AI: Partnering Academia with Local Communities

I’ll present ongoing work at Tulane’s Center for Community-Engaged AI partnering with local non-profits to build AI tools for transparency and accountability in criminal court and city government. I will first discuss our work with Eye on Surveillance, who have developed a retrieval-augmented generation tool over New Orleans city council transcripts (sawt.us). We have collaborated to use automated LLM evaluations and to engage with community users to improve the trustworthiness of the system. Second, I will discuss our work with Court Watch NOLA to build AI models for monitoring the equity in criminal court, including methods to estimating causal effects from text data. I will conclude with an overview of future directions in how generative AI can be used for a number of civic applications.

Dr. Culotta is an Associate Professor of Computer Science at Tulane University and Director of the Tulane Center for Community-Engaged Artificial Intelligence. He received his PhD in computer science from the University of Massachusetts-Amherst in 2008, and his research focuses on machine learning, natural language processing, and social network analysis, and their implications on society. His NSF-funded research has developed AI methods for several interdisciplinary projects in public health, marketing, political science, and emergency response. He has published over seventy academic articles on AI, serves on the steering committee of the International Conference on Web and Social Media, and has received best paper awards from the ACM Conference on Computer-Supported Cooperative Work and Social Computing and from the AAAI Conference on Artificial Intelligence.

Can synthetic images be better than real images?
A study of utility and privacy of synthetic images in dermatology

Advances in generative models such as GAN, VAE, and more recently, Diffusion models have revolutionized the field of image generation by enabling the generation of photorealistic synthetic images for many potential applications. Along with the advances, the question whether synthetic data can replace real data has become increasingly relevant. Synthetic data has been demonstrated to improve classification and to overcome the data scarcity issues such as data imbalance, robustness, and biases. This is done by generating synthetic data distributions with improved balance and other properties. Furthermore, there is an increasing demand for privacy-preserving synthetic image generation across various domains such as healthcare, finance, and social media. However, achieving the optimal tradeoff between utility and privacy remains a significant technical challenge, and various privacy-preserving techniques are being studied. In this talk, I will first discuss how generative AI can help in skin disease diagnosis problem, and also present benchmarking results on utility-privacy of recent image synthesis methods.

Dr. Jihun Hamm has been an Associate Professor of Computer Science at Tulane University since 2019. He received his PhD degree from the University of Pennsylvania in 2008 supervised by Dr. Daniel Lee. Dr. Hamm’s research interest is in machine learning, from theory to applications. He has worked on the theory and practice of robust and adversarial machine learning, privacy and security and optimization. Dr. Hamm also has worked on medical data analysis. His work in machine learning has been published in top venues such as ICML, NeurIPS, CVPR, JMLR, and IEEE-TPAMI. His work has also been published in medical research venues such as MICCAI, MedIA, and IEEE-TMI. Among other awards, he has earned the Best Paper Award from MedIA, Finalist for MICCAI Young Scientist Publication Impact Award, and Google Faculty Research Award.