Alignment methods in moral domains seek to elicit moral preferences of human stakeholders and incorporate them into AI. This presupposes moral preferences as static targets, but such preferences often evolve over time. Proper alignment of AI to dynamic human preferences should ideally account for "legitimate" changes to moral reasoning, while ignoring changes related to attention deficits, cognitive biases, or other arbitrary factors. However, common AI alignment approaches largely neglect temporal changes in preferences, posing serious challenges to proper alignment, especially in high-stakes applications of AI, e.g., in healthcare domains, where misalignment can jeopardize the trustworthiness of the system and yield serious individual and societal harms. This work investigates the extent to which people's moral preferences change over time, and the impact of such changes on AI alignment. Our study is grounded in the kidney allocation domain, where we elicit responses to pairwise comparisons of hypothetical kidney transplant patients from over 400 participants across 3-5 sessions. We find that, on average, participants change their response to the same scenario presented at different times around 6-20% of the time (exhibiting "response instability"). Additionally, we observe significant shifts in several participants' retrofitted decision-making models over time (capturing "model instability"). Predictive performance of simple AI models decreases as a function of both response and model instability. Moreover, predictive performance diminishes over time, highlighting the importance of accounting for temporal changes in preferences during training. These findings raise fundamental normative and technical challenges relevant to AI alignment, highlighting the need to better understand the object of alignment (what to align to) when user preferences change significantly over time, including the mechanisms underlying these changes.

A growing body of work in Ethical AI attempts to capture human moral judgments through simple computational models.The key question we address in this work is whether such simple AI models capture the critical nuances of moral decision-making by focusing on the use case of kidney allocation.We conducted twenty interviews where participants explained their rationale for their judgments about who should receive a kidney.We observe participants: (a) value patients' morally-relevant attributes to different degrees; (b) use diverse decision-making processes, citing heuristics to reduce decision complexity; (c) can change their opinions; (d) sometimes lack confidence in their decisions (e.g., due to incomplete information); and (e) express enthusiasm and concern regarding AI assisting humans in kidney allocation decisions.Based on these findings, we discuss challenges of computationally modeling moral judgments as a stand-in for human input, highlight drawbacks of current approaches, and suggest future directions to address these issues.

Jiarui Liu, Yueqi Song, Yunze Xiao, Mingqian Zheng, Lindia Tjuatja, Jana Schaich Borg, Mona T. Diab, Maarten Sap. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

The ideal AI safety moderation system would be both structurally interpretable (so its decisions can be reliably explained) and steerable (to align to safety standards and reflect a community’s values), which current systems fall short on. To address this gap, we present SafetyAnalyst, a novel AI safety moderation framework. Given an AI behavior, SafetyAnalyst uses chain-of-thought reasoning to analyze its potential consequences by creating a structured "harm-benefit tree," which enumerates harmful and beneficial actions and effects the AI behavior may lead to, along with likelihood, severity, and immediacy labels that describe potential impacts on stakeholders. SafetyAnalyst then aggregates all effects into a harmfulness score using 28 fully interpretable weight parameters, which can be aligned to particular safety preferences. We applied this framework to develop an open-source LLM prompt safety classification system, distilled from 18.5 million harm-benefit features generated by frontier LLMs on 19k prompts. On comprehensive benchmarks, we show that SafetyAnalyst (average F1=0.81) outperforms existing moderation systems (average F1$<$0.72) on prompt safety classification, while offering the additional advantages of interpretability, transparency, and steerability.

Preference elicitation frameworks feature heavily in the research on participatory ethical AI tools and provide a viable mechanism to enquire and incorporate the moral values of various stakeholders. As part of the elicitation process, surveys about moral preferences, opinions, and judgments are typically administered only once to each participant. This methodological practice is reasonable if participants’ responses are stable over time such that, all other things being held constant, their responses today will be the same as their responses to the same questions at a later time. However, we do not know how often that is the case. It is possible that participants’ true moral preferences change, are subject to temporary moods or whims, or are influenced by environmental factors we don’t track. If participants’ moral responses are unstable in such ways, it would raise important methodological and theoretical issues for how participants’ true moral preferences, opinions, and judgments can be ascertained. We address this possibility here by asking the same survey participants the same moral questions about which patient should receive a kidney when only one is available ten times in ten different sessions over two weeks, varying only presentation order across sessions. We measured how often participants gave different responses to simple (Study One) and more complicated (Study Two) controversial and uncontroversial repeated scenarios. On average, the fraction of times participants changed their responses to controversial scenarios (i.e., were unstable) was around 10-18% (±14-15%) across studies, and this instability is observed to have positive associations with response time and decision-making difficulty. We discuss the implications of these results for the efficacy of common moral preference elicitation methods, highlighting the role of response instability in potentially causing value misalignment between the stakeholders and AI tools trained on their moral judgments.

Interest is growing in artificial empathy, but so is confusion about what artificial empathy is or needs to be. This confusion makes it challenging to navigate the technical and ethical issues that accompany empathic AI development. Here, we outline a framework for thinking about empathic AI based on the premise that different constellations of capabilities associated with empathy are important for different empathic AI applications. We describe distinctions of capabilities that we argue belong under the empathy umbrella, and show how three medical empathic AI use cases require different sets of these capabilities. We conclude by discussing why appreciation of the diverse capabilities under the empathy umbrella is important for both AI creators and users.

Computational preference elicitation methods are tools used to learn people’s preferences quantitatively in a given context. Recent works on preference elicitation advocate for active learning as an efficient method to iteratively construct queries (framed as comparisons between context-specific cases) that are likely to be most informative about an agent’s underlying preferences. In this work, we argue that the use of active learning for moral preference elicitation relies on certain assumptions about the underlying moral preferences, which can be violated in practice. Specifically, we highlight the following common assumptions (a) preferences are stable over time and not sensitive to the sequence of presented queries, (b) the appropriate hypothesis class is chosen to model moral preferences, and (c) noise in the agent’s responses is limited. While these assumptions can be appropriate for preference elicitation in certain domains, prior research on moral psychology suggests they may not be valid for moral judgments. Through a synthetic simulation of preferences that violate the above assumptions, we observe that active learning can have similar or worse performance than a basic random query selection method in certain settings. Yet, simulation results also demonstrate that active learning can still be viable if the degree of instability or noise is relatively small and when the agent’s preferences can be approximately represented with the hypothesis class used for learning. Our study highlights the nuances associated with effective moral preference elicitation in practice and advocates for the cautious use of active learning as a methodology to learn moral preferences.

AI systems are often used to make or contribute to important decisions in a growing range of applications, including criminal justice, hiring, and medicine. Since these decisions impact human lives, it is important that the AI systems act in ways which align with human values. Techniques for preference modeling and social choice help researchers learn and aggregate peoples' preferences, which are used to guide AI behavior; thus, it is imperative that these learned preferences are accurate. These techniques often assume that people are willing to express strict preferences over alternatives; which is not true in practice. People are often indecisive, and especially so when their decision has moral implications. The philosophy and psychology literature shows that indecision is a measurable and nuanced behavior---and that there are several different reasons people are indecisive. This complicates the task of both learning and aggregating preferences, since most of the relevant literature makes restrictive assumptions on the meaning of indecision. We begin to close this gap by formalizing several mathematical indecision models based on theories from philosophy, psychology, and economics; these models can be used to describe (indecisive) agent decisions, both when they are allowed to express indecision and when they are not. We test these models using data collected from an online survey where participants choose how to (hypothetically) allocate organs to patients waiting for a transplant.

Diverse disciplines are interested in how the coordination of interacting agents' movements, emotions, and physiology over time impacts social behavior. Here, we describe a new multivariate procedure for automating the investigation of this kind of behaviorally-relevant “interactional synchrony”, and introduce a novel interactional synchrony measure based on features of dynamic time warping (DTW) paths. We demonstrate that our DTW path-based measure of interactional synchrony between facial action units of two people interacting freely in a natural social interaction can be used to predict how much trust they will display in a subsequent Trust Game. We also show that our approach outperforms univariate head movement models, models that consider participants' facial action units independently, and models that use previously proposed synchrony or similarity measures. The insights of this work can be applied to any research question that aims to quantify the temporal coordination of multiple signals over time, but has immediate applications in psychology, medicine, and robotics.

Many modern data sets require inference methods that can estimate the shared and individual-specific components of variability in collections of matrices that change over time. Promising methods have been developed to analyze these types of data in static cases, but only a few approaches are available for dynamic settings. To address this gap, we consider novel models and inference methods for pairs of matrices in which the columns correspond to multivariate observations at different time points. In order to characterize common and individual features, we propose a Bayesian dynamic factor modeling framework called Time Aligned Common and Individual Factor Analysis (TACIFA) that includes uncertainty in time alignment through an unknown warping function. We provide theoretical support for the proposed model, showing identifiability and posterior concentration. The structure enables efficient computation through a Hamiltonian Monte Carlo (HMC) algorithm. We show excellent performance in simulations, and illustrate the method through application to a social mimicry experiment.

Given AI's growing role in modeling and improving decision-making, how and when to present users with feedback is an urgent topic to address. We empirically examined the effect of feedback from false AI on moral decision-making about donor kidney allocation. We found some evidence that judgments about whether a patient should receive a kidney can be influenced by feedback about participants' own decision-making perceived to be given by AI, even if the feedback is entirely random. We also discovered different effects between assessments presented as being from human experts and assessments presented as being from AI.

The efficient allocation of limited resources is a classical problem in economics and computer science. In kidney exchanges, a central market maker allocates living kidney donors to patients in need of an organ. Patients and donors in kidney exchanges are prioritized using ad-hoc weights decided on by committee and then fed into an allocation algorithm that determines who get what—and who does not. In this paper, we provide an end-to-end methodology for estimating weights of individual participant profiles in a kidney exchange. We first elicit from human subjects a list of patient attributes they consider acceptable for the purpose of prioritizing patients (e.g., medical characteristics, lifestyle choices, and so on). Then, we ask subjects comparison queries between patient profiles and estimate weights in a principled way from their responses. We show how to use these weights in kidney exchange market clearing algorithms. We then evaluate the impact of the weights in simulations and find that the precise numerical values of the weights we computed matter little, other than the ordering of profiles that they imply. However, compared to not prioritizing patients at all, there is a significant effect, with certain classes of patients being (de)prioritized based on the human-elicited value judgments.

The efficient allocation of limited resources is a classical problem in economics and computer science. In kidney exchanges, a central market maker allocates living kidney donors to patients in need of an organ. Patients and donors in kidney exchanges are prioritized using ad-hoc weights decided on by committee and then fed into an allocation algorithm that determines who get what---and who does not. In this paper, we provide an end-to-end methodology for estimating weights of individual participant profiles in a kidney exchange. We first elicit from human subjects a list of patient attributes they consider acceptable for the purpose of prioritizing patients (e.g., medical characteristics, lifestyle choices, and so on). Then, we ask subjects comparison queries between patient profiles and estimate weights in a principled way from their responses. We show how to use these weights in kidney exchange market clearing algorithms. We then evaluate the impact of the weights in simulations and find that the precise numerical values of the weights we computed matter little, other than the ordering of profiles that they imply. However, compared to not prioritizing patients at all, there is a significant effect, with certain classes of patients being (de)prioritized based on the human-elicited value judgments.

AI systems are now or will soon be sophisticated enough to make consequential decisions. Although this technology has flourished, we also need public appraisals of AI systems playing these more important roles. This article reports surveys of preferences for and against AI systems making decisions in various domains as well as experiments that intervene on these preferences. We find that these preferences are contingent on subjects' previous exposure to computer systems making these kinds of decisions, and some interventions designed to mimic previous exposure successfully encourage subjects to be more hospitable to computer systems making these weighty decisions.

The generality of decision and game theory has enabled domain-independent progress in AI research. For example, a better algorithm for finding good policies in (PO)MDPs can be instantly used in a variety of applications. But such a general theory is lacking when it comes to moral decision making. For AI applications with a moral component, are we then forced to build systems based on many ad-hoc rules? In this paper we discuss possible ways to avoid this conclusion.