LLMs are increasingly fine-tuned using RLHF datasets to align them with human preferences and values. However, very limited research has investigated which specific human values are operationalized through these datasets. In this paper, we introduce Value Imprint, a framework for auditing and classifying the human values embedded within RLHF datasets. To investigate the viability of this framework, we conducted three case study experiments by auditing the Anthropic/hh-rlhf, OpenAI WebGPT Comparisons, and Alpaca GPT-4-LLM datasets to examine the human values embedded within them. Our analysis involved a two-phase process. During the first phase, we developed a taxonomy of human values through an integrated review of prior works from philosophy, axiology, and ethics. Then, we applied this taxonomy to annotate 6,501 RLHF preferences. During the second phase, we employed the labels generated from the annotation as ground truth data for training a transformer-based machine learning model to audit and classify the three RLHF datasets. Through this approach, we discovered that information-utility values, including Wisdom/Knowledge and Information Seeking, were the most dominant human values within all three RLHF datasets. In contrast, prosocial and democratic values, including Well-being, Justice, and Human/Animal Rights, were the least represented human values. These findings have significant implications for developing language models that align with societal values and norms. We contribute our datasets to support further research in this area. https://github.com/hv-rsrch/valueimprint

Every year, supermarkets discard tons of food that are completely fit for consumption. This happens because of inefficient labeling on packages and confusion between different terminologies like best by, use by, sell by, etc., used on the food packages. However, we now have the technology to eliminate human error and put this food to better use. A group of children in the ACM Interaction Design and Children Research & Design Challenge (IDC ’23) booklet came up with the idea of a mobile application that could distribute such food products to those in need. Our team expanded on this idea and other related and relevant ideas from the booklet and designed an application that uses Artificial Intelligence (AI) to sort and package safe for consumption good food that would otherwise go to waste. Our solution aims to provide weekly food packages for individuals or families in need, thereby ensuring that no one goes hungry and also mitigating unnecessary food wastage. It is a positive outcome for every stakeholder involved: supermarkets reduce unnecessary food wastage and underserved communities receive food, thus contributing to a more sustainable future.

In the realm of dementia care-giving within India, the caregivers often resemble hidden patients, bearing the weight of an often-overlooked role. However, there remains a dearth of comprehensive data regarding the emotional and physical challenges caregivers face in the Indian subcontinent. We present a qualitative study that delves into the realities of being a dementia caregiver in India, investigating caregivers’ experiences and examining the impact of several challenges on their daily lives. Through in-depth interviews, we engaged with four primary caregivers of individuals with dementia, delving into how care-giving shapes their social interactions, mental equilibrium, and emotional states. The findings highlight the importance of having an emotional outlet, a support system, and accessible resources for enhancing caregivers’ quality of life. Drawing from these insights, we propose a set of design implications that can guide future endeavors focused on enhancing the overall well-being of dementia caregivers in India. This research effectively bridges a substantial knowledge gap, extending support and understanding to this vital cohort of individuals who stand at the forefront of dementia care.