Search

Human annotation plays a core role in machine learning -- annotations for supervised models, safety guardrails for generative models, and human feedback for reinforcement learning, to cite a few avenues. However, the fact that many of these human annotations are inherently subjective is often overlooked. Recent work has demonstrated that ignoring rater subjectivity (typically resulting in rater disagreement) is problematic within specific tasks and for specific subgroups. Generalizable methods to harness rater disagreement and thus understand the socio-cultural leanings of subjective tasks remain elusive. In this paper, we propose GRASP, a comprehensive disagreement analysis framework to measure group association in perspectives among different rater sub-groups, and demonstrate its utility in assessing the extent of systematic disagreements in two datasets: (1) safety annotations of human-chatbot conversations, and (2) offensiveness annotations of social media posts, both annotated by diverse rater pools across different socio-demographic axes. Our framework (based on disagreement metrics) reveals specific rater groups that have significantly different perspectives than others on certain tasks, and helps identify demographic axes that are crucial to consider in specific task contexts., Comment: Presented as a long paper at NAACL 2024 main conference

Conversational AI systems exhibit a level of human-like behavior that promises to have profound impacts on many aspects of daily life -- how people access information, create content, and seek social support. Yet these models have also shown a propensity for biases, offensive language, and conveying false information. Consequently, understanding and moderating safety risks in these models is a critical technical and social challenge. Perception of safety is intrinsically subjective, where many factors -- often intersecting -- could determine why one person may consider a conversation with a chatbot safe and another person could consider the same conversation unsafe. In this work, we focus on demographic factors that could influence such diverse perceptions. To this end, we contribute an analysis using Bayesian multilevel modeling to explore the connection between rater demographics and how raters report safety of conversational AI systems. We study a sample of 252 human raters stratified by gender, age group, race/ethnicity group, and locale. This rater pool provided safety labels for 1,340 human-chatbot conversations. Our results show that intersectional effects involving demographic characteristics such as race/ethnicity, gender, and age, as well as content characteristics, such as degree of harm, all play significant roles in determining the safety of conversational AI systems. For example, race/ethnicity and gender show strong intersectional effects, particularly among South Asian and East Asian women. We also find that conversational degree of harm impacts raters of all race/ethnicity groups, but that Indigenous and South Asian raters are particularly sensitive to this harm. Finally, we observe the effect of education is uniquely intersectional for Indigenous raters, highlighting the utility of multilevel frameworks for uncovering underrepresented social perspectives.

Machine learning approaches often require training and evaluation datasets with a clear separation between positive and negative examples. This risks simplifying and even obscuring the inherent subjectivity present in many tasks. Preserving such variance in content and diversity in datasets is often expensive and laborious. This is especially troubling when building safety datasets for conversational AI systems, as safety is both socially and culturally situated. To demonstrate this crucial aspect of conversational AI safety, and to facilitate in-depth model performance analyses, we introduce the DICES (Diversity In Conversational AI Evaluation for Safety) dataset that contains fine-grained demographic information about raters, high replication of ratings per item to ensure statistical power for analyses, and encodes rater votes as distributions across different demographics to allow for in-depth explorations of different aggregation strategies. In short, the DICES dataset enables the observation and measurement of variance, ambiguity, and diversity in the context of conversational AI safety. We also illustrate how the dataset offers a basis for establishing metrics to show how raters' ratings can intersects with demographic categories such as racial/ethnic groups, age groups, and genders. The goal of DICES is to be used as a shared resource and benchmark that respects diverse perspectives during safety evaluation of conversational AI systems.

In this paper, we present findings from an semi-experimental exploration of rater diversity and its influence on safety annotations of conversations generated by humans talking to a generative AI-chat bot. We find significant differences in judgments produced by raters from different geographic regions and annotation platforms, and correlate these perspectives with demographic sub-groups. Our work helps define best practices in model development -- specifically human evaluation of generative models -- on the backdrop of growing work on sociotechnical AI evaluations., Comment: 5 pages

Expanding upon recent work, a new class of $A$-functions is introduced that can be viewed as an appropriate generalization of the class of regular $A$-functions, the class of structured $A$-functions, and the class of perfect $A$-functions. Algebraic results are discussed and analysis is performed to deduce new results concerning rational arithmetical functions.

Accurate estimation of healthcare costs is crucial for healthcare systems to plan and effectively negotiate with insurance companies regarding the coverage of patient-care costs. Greater accuracy in estimating healthcare costs would provide mutual benefit for both health systems and the insurers that support these systems by better aligning payment models with patient-care costs. This study presents the results of a generalizable machine learning approach to predicting medical events built from 40 years of data from >860,000 patients pertaining to >6,700 prescription medications, courtesy of Marshfield Clinic in Wisconsin. It was found that models built using this approach performed well when compared to similar studies predicting physician prescriptions of individual medications. In addition to providing a comprehensive predictive model for all drugs in a large healthcare system, the approach taken in this research benefits from potential applicability to a wide variety of other medical events., Comment: 4 pages, 3 figures

We provide a new proof that the upper nilradical of a compact ring coincides with the sum of its left nil ideals using the properties of orthogonal idempotents in compact rings., Comment: An error in the proof of the main theorem has been corrected

One well-known necessary and sufficient condition for equality in the Frobenius rank inequality due to Tian and Styan is that the matrix equation B = BCX + YAB be solvable for X and Y. We develop an algorithm to construct the matrices X and Y using a quotient space formulation of the Frobenius rank inequality, and provide several necessary and sufficient conditions for solvability., Comment: 7 pages

Incremental computation has recently been studied using the concepts of change structures and derivatives of programs, where the derivative of a function allows updating the output of the function based on a change to its input. We generalise change structures to change actions, and study their algebraic properties. We develop change actions for common structures in computer science, including directed-complete partial orders and Boolean algebras. We then show how to compute derivatives of fixpoints. This allows us to perform incremental evaluation and maintenance of recursively defined functions with particular application to generalised Datalog programs. Moreover, unlike previous results, our techniques are modular in that they are easy to apply both to variants of Datalog and to other programming languages., Comment: 34 pages, 2 figures

Engaging residential communities with each other and with management remains a challenge. Housing providers deploy a variety of engagement strategies, some of which are supported by digital technologies. Their individual success is varied and integrated, multipronged approaches are seen to be more successful. As part of those, it is important to address people's perceptions of community and places, as well as any practical issues that they face. We present the design and evaluation of Photo Screen, a situated, public photo taking and viewing screen which was deployed in the context of a new flagship housing estate as part of a range of community engagement measures. In a new context, we confirm the high levels of engagement that can be achieved with this simple mechanism. We propose that photo 'tagging' might offer a second-stage engagement mechanism and enable meaningful dialogue between residents and management. Finally, we discuss how this playful activity allowed residents to positively shape the perception of their community., Comment: 4 pages, plus references