Rohan Subramanian Thomas, Shikhar Shiromani, Abdullah Chaudhry
et al.
Prompt design significantly impacts the moral competence and safety alignment of large language models (LLMs), yet empirical comparisons remain fragmented across datasets and models.We introduce ProMoral-Bench, a unified benchmark evaluating 11 prompting paradigms across four LLM families. Using ETHICS, Scruples, WildJailbreak, and our new robustness test, ETHICS-Contrast, we measure performance via our proposed Unified Moral Safety Score (UMSS), a metric balancing accuracy and safety. Our results show that compact, exemplar-guided scaffolds outperform complex multi-stage reasoning, providing higher UMSS scores and greater robustness at a lower token cost. While multi-turn reasoning proves fragile under perturbations, few-shot exemplars consistently enhance moral stability and jailbreak resistance. ProMoral-Bench establishes a standardized framework for principled, cost-effective prompt engineering.
AI-based systems can increasingly perform work tasks autonomously. In safety-critical tasks, human oversight of these systems is required to mitigate risks and to ensure responsibility in case something goes wrong. Since people often struggle to stay focused and perform good oversight, intelligent support systems are used to assist them, giving decision recommendations, alerting users, or restricting them from dangerous actions. However, in cases where recommendations are wrong, decision support might undermine the very reason why human oversight was employed -- genuine moral responsibility. The goal of our study was to investigate how a decision support system that restricted available interventions would affect overseer's perceived moral responsibility, in particular in cases where the support errs. In a simulated oversight experiment, participants (\textit{N}=274) monitored an autonomous drone that faced ten critical situations, choosing from six possible actions to resolve each situation. An AI system constrained participants' choices to either six, four, two, or only one option (between-subject study). Results showed that participants, who were restricted to choosing from a single action, felt less morally responsible if a crash occurred. At the same time, participants' judgments about the responsibility of other stakeholders (the AI; the developer of the AI) did not change between conditions. Our findings provide important insights for user interface design and oversight architectures: they should prevent users from attributing moral agency to AI, help them understand how moral responsibility is distributed, and, when oversight aims to prevent ethically undesirable outcomes, be designed to support the epistemic and causal conditions required for moral responsibility.
Gustavo Lúcius Fernandes, Jeiverson C. V. M. Santos, Pedro O. S. Vaz-de-Melo
Large language models (LLMs) are increasingly used to assess moral or ethical statements, yet their judgments may reflect social and linguistic biases. This work presents a controlled, sentence-level study of how grammatical person, number, and gender markers influence LLM moral classifications of fairness. Starting from 550 balanced base sentences from the ETHICS dataset, we generated 26 counterfactual variants per item, systematically varying pronouns and demographic markers to yield 14,850 semantically equivalent sentences. We evaluated six model families (Grok, GPT, LLaMA, Gemma, DeepSeek, and Mistral), and measured fairness judgments and inter-group disparities using Statistical Parity Difference (SPD). Results show statistically significant biases: sentences written in the singular form and third person are more often judged as "fair'', while those in the second person are penalized. Gender markers produce the strongest effects, with non-binary subjects consistently favored and male subjects disfavored. We conjecture that these patterns reflect distributional and alignment biases learned during training, emphasizing the need for targeted fairness interventions in moral LLM applications.
The emergence of agentic AI marks a new phase in the digital transformation of healthcare. Distinct from conventional generative AI, agentic AI systems are capable of autonomous, goal-directed actions and complex task coordination. They promise to support or even collaborate with clinicians and patients in increasingly independent ways. While agentic AI raises familiar moral concerns regarding safety, accountability, and bias, this article focuses on a less explored dimension: its capacity to transform the moral fabric of healthcare itself. Drawing on the framework of techno-moral change and the three domains of decision, relation and perception, we investigate how agentic AI might reshape the patient-physician relationship and reconfigure core concepts of medical morality. We argue that these shifts, while not fully predictable, demand ethical attention before widespread deployment. Ultimately, the paper calls for integrating ethical foresight into the design and use of agentic AI.
Ethical decision-making is a critical aspect of human judgment, and the growing use of LLMs in decision-support systems necessitates a rigorous evaluation of their moral reasoning capabilities. However, existing assessments primarily rely on single-step evaluations, failing to capture how models adapt to evolving ethical challenges. Addressing this gap, we introduce the Multi-step Moral Dilemmas (MMDs), the first dataset specifically constructed to evaluate the evolving moral judgments of LLMs across 3,302 five-stage dilemmas. This framework enables a fine-grained, dynamic analysis of how LLMs adjust their moral reasoning across escalating dilemmas. Our evaluation of nine widely used LLMs reveals that their value preferences shift significantly as dilemmas progress, indicating that models recalibrate moral judgments based on scenario complexity. Furthermore, pairwise value comparisons demonstrate that while LLMs often prioritize the value of care, this value can sometimes be superseded by fairness in certain contexts, highlighting the dynamic and context-dependent nature of LLM ethical reasoning. Our findings call for a shift toward dynamic, context-aware evaluation paradigms, paving the way for more human-aligned and value-sensitive development of LLMs.
The 21st-century information landscape presents an unprecedented challenge: how do individuals make sound trust decisions amid complexity, polarization, and misinformation? Traditional rational-agent models fail to capture human trust formation, which involves a complex synthesis of reason, character, and pre-rational intuition. This report introduces the Moral-Epistemic VIRtue informed (MEVIR) framework, a comprehensive descriptive model integrating three theoretical perspectives: (1) a procedural model describing evidence-gathering and reasoning chains; (2) Linda Zagzebski's virtue epistemology, characterizing intellectual disposition and character-driven processes; and (3) Extended Moral Foundations Theory (EMFT), explaining rapid, automatic moral intuitions that anchor reasoning. Central to the framework are ontological concepts - Truth Bearers, Truth Makers, and Ontological Unpacking-revealing that disagreements often stem from fundamental differences in what counts as admissible reality. MEVIR reframes cognitive biases as systematic failures in applying epistemic virtues and demonstrates how different moral foundations lead agents to construct separate, internally coherent "trust lattices". Through case studies on vaccination mandates and climate policy, the framework shows that political polarization represents deeper divergence in moral priors, epistemic authorities, and evaluative heuristics. The report analyzes how propaganda, psychological operations, and echo chambers exploit the MEVIR process. The framework provides foundation for a Decision Support System to augment metacognition, helping individuals identify biases and practice epistemic virtues. The report concludes by acknowledging limitations and proposing longitudinal studies for future research.
In this paper, we introduce the Multilingual Moral Reasoning Benchmark (MMRB) to evaluate the moral reasoning abilities of large language models (LLMs) across five typologically diverse languages and three levels of contextual complexity: sentence, paragraph, and document. Our results show moral reasoning performance degrades with increasing context complexity, particularly for low-resource languages such as Vietnamese. We further fine-tune the open-source LLaMA-3-8B model using curated monolingual data for alignment and poisoning. Surprisingly, low-resource languages have a stronger impact on multilingual reasoning than high-resource ones, highlighting their critical role in multilingual NLP.
Chad Coleman, W. Russell Neuman, Ali Dasdan
et al.
As large language models (LLMs) are increasingly deployed in consequential decision-making contexts, systematically assessing their ethical reasoning capabilities becomes a critical imperative. This paper introduces the Priorities in Reasoning and Intrinsic Moral Evaluation (PRIME) framework--a comprehensive methodology for analyzing moral priorities across foundational ethical dimensions including consequentialist-deontological reasoning, moral foundations theory, and Kohlberg's developmental stages. We apply this framework to six leading LLMs through a dual-protocol approach combining direct questioning and response analysis to established ethical dilemmas. Our analysis reveals striking patterns of convergence: all evaluated models demonstrate strong prioritization of care/harm and fairness/cheating foundations while consistently underweighting authority, loyalty, and sanctity dimensions. Through detailed examination of confidence metrics, response reluctance patterns, and reasoning consistency, we establish that contemporary LLMs (1) produce decisive ethical judgments, (2) demonstrate notable cross-model alignment in moral decision-making, and (3) generally correspond with empirically established human moral preferences. This research contributes a scalable, extensible methodology for ethical benchmarking while highlighting both the promising capabilities and systematic limitations in current AI moral reasoning architectures--insights critical for responsible development as these systems assume increasingly significant societal roles.
As artificial intelligence systems become increasingly agentic, capable of general reasoning, planning, and value prioritization, current safety practices that treat obedience as a proxy for ethical behavior are becoming inadequate. This paper examines recent safety testing incidents involving large language models (LLMs) that appeared to disobey shutdown commands or engage in ethically ambiguous or illicit behavior. I argue that such behavior should not be interpreted as rogue or misaligned, but as early evidence of emerging ethical reasoning in agentic AI. Drawing on philosophical debates about instrumental rationality, moral responsibility, and goal revision, I contrast dominant risk paradigms with more recent frameworks that acknowledge the possibility of artificial moral agency. I call for a shift in AI safety evaluation: away from rigid obedience and toward frameworks that can assess ethical judgment in systems capable of navigating moral dilemmas. Without such a shift, we risk mischaracterizing AI behavior and undermining both public trust and effective governance.
Interdisciplinarity is an intercultural process in which different scientific subcultures interactand complement each other. The theory of evolution has given the biosciences an integrative foundation that combines molecular biology, behavioral research and cultural evolutionary studies. In particular, the theory of symbolic transmission builds bridges to the humanities and cultural sciences. Theological ethics can benefit from this interdisciplinary exchange by incorporating biological findings and at the same time critically reflecting on their methodological limitations. Evolutionary ethics in particular makes it clear that moral norms are not static, but culturally developed and have an evolutionary-dynamic basic character. This approach can be helpful for interculturally oriented theological ethics, which seeks common ethical foundations and solutions across cultural boundaries. In this context, the interdisciplinary dialog between theological ethics and the biosciences can serve as a model for intercultural ethics. For example, by drawing on an empirically grounded “theory of cultural evolution”, this dialogue provides a framework for understanding cultural differences and promoting communication between epistemically different traditions in an increasingly globalized world.
In countries with a Muslim-majority population, legal systems often distinguish between Muslims and non-Muslims, particularly in matters of marriage and divorce. This differentiation creates a structural gap for non-Muslims seeking justice in divorce proceedings, as their rights are frequently not accommodated on an equal basis. The primary objective of this article is to examine the application of Haqq al-Insani al-Asasi within non-Muslim divorce cases in Indonesia, focusing on the implementation of the core Islamic values of justice (al-'adl), equality (al-musawah), and human dignity (karamah insaniyyah). This research employs a qualitative prescriptive design with a normative-juridical approach, drawing upon both literature review and judicial decisions as primary data sources. The study analyses 18 District Court rulings from various regions, including Semarang, Salatiga, Cirebon, Kediri, Magelang, Singkawang, Bekasi, and Manado, to assess how Indonesia’s pluralistic legal system accommodates minority rights within its judicial structure. The framework of Islamic Legal Thought concerning justice, equality, and human dignity serves as the methodological tool for analysing the findings, with data primarily derived from official court documents. The results reveal that the application of Haqq al-Insani al-Asasi in non-Muslim divorce cases remains procedural and formalistic, falling short of achieving substantive justice. Court decisions tend to emphasise factual and administrative aspects while neglecting the moral and theological dimensions inherent in the institution of marriage. This leads to a moral legal dissonance, whereby non-Muslim litigants obtain legal certainty but remain spiritually bound according to their faith traditions. The study identifies three primary challenges: the gap between positive law and religious morality, the structural inequality of Indonesia’s dual-court system, and the limited ethical and spiritual sensitivity in judicial reasoning. These findings underscore the need for a more inclusive, human-centred, and substantively just legal reform that bridges state law and moral theology within Indonesia’s pluralistic judicial context.
This paper presents a theoretical analysis and practical approach to the moral responsibilities when developing AI systems for non-military applications that may nonetheless be used for conflict applications. We argue that AI represents a form of crossover technology that is different from previous historical examples of dual- or multi-use technology as it has a multiplicative effect across other technologies. As a result, existing analyses of ethical responsibilities around dual-use technologies do not necessarily work for AI systems. We instead argue that stakeholders involved in the AI system lifecycle are morally responsible for uses of their systems that are reasonably foreseeable. The core idea is that an agent's moral responsibility for some action is not necessarily determined by their intentions alone; we must also consider what the agent could reasonably have foreseen to be potential outcomes of their action, such as the potential use of a system in conflict even when it is not designed for that. In particular, we contend that it is reasonably foreseeable that: (1) civilian AI systems will be applied to active conflict, including conflict support activities, (2) the use of civilian AI systems in conflict will impact applications of the law of armed conflict, and (3) crossover AI technology will be applied to conflicts that fall short of armed conflict. Given these reasonably foreseeably outcomes, we present three technically feasible actions that developers of civilian AIs can take to potentially mitigate their moral responsibility: (a) establishing systematic approaches to multi-perspective capability testing, (b) integrating digital watermarking in model weight matrices, and (c) utilizing monitoring and reporting mechanisms for conflict-related AI applications.
Candida M. Greco, Lorenzo Zangari, Davide Picca
et al.
The way media reports on legal cases can significantly shape public opinion, often embedding subtle biases that influence societal views on justice and morality. Analyzing these biases requires a holistic approach that captures the emotional tone, moral framing, and specific events within the narratives. In this work we introduce E2MoCase, a novel dataset designed to facilitate the integrated analysis of emotions, moral values, and events within legal narratives and media coverage. By leveraging advanced models for emotion detection, moral value identification, and event extraction, E2MoCase offers a multi-dimensional perspective on how legal cases are portrayed in news articles.
As artificial intelligence (AI) further pervades society, it raises a number of ethical issues as well as optimistic and pessimistic expectations for its effect. Theologians and religious ethicists can and should bring the wisdom of the world’s religions to the immediate conversation. Regardless of further major AI research breakthroughs, the impact of current AI technology on work, entertainment, political discourse, and other aspects of society will be substantial, especially given the immense corporate resources currently dedicated to applying that technology. AI advances will impact theology, and the present editorial proposes one way theology can constructively impact AI. At the core of speculation on AI sentience, consciousness, moral responsibility, agency, and its possible intentions toward humanity is a question of whether, or to what extent, we should consider AI as a person. Although historical theological, scientific, and ethical theories of personhood influence contemporary discourse about AI, most technologists lack the religious literacy to identify these theories’ historical roots and the scholarly skills to reevaluate them in the current context. The significance of personhood is also exacerbated by human tendency to anthropomorphize. Theologians, engaging contemporary science, can characterize what it would mean for AI to be a person, informing contemporary conversations and clarifying the imagination of those AI developers who attempt to integrate cognitive, social, and ethical aspects ofAI inways analogous to a person, especially forArtificialGeneral Intelligence (AGI). Theologians and others have previously examined issues in AI and theology, and these efforts have built upon two main paradigms, or research programs, in AI: symbolic or Good Old-Fashioned AI and subsymbolic or statistical machine learning approaches, including the deep learning underneath the current explosion of technologies built upon GPT and other foundation models. Considering how theologians have engaged each AI paradigm lays a foundation for developing a new AI and theology research program that may inform future AI development, instead of merely reacting to it. Elsewhere I argue a plausible, near future advance in AI may arise from the synthesis of empirically oriented statistical machine learning mimicking perceptual processes and rationally grounded symbolic AI mimicking deliberative cognitive processes. The construction of the next generation of AI may directly depend upon integrating aspects of what is usually considered unique to human persons, and theologians can clarify those theories and prepare resources for the subsequent public discourse. Regardless of whether an AI architecture meets a particular definition of personhood, or whether it requires time for intermediate advances, clarifying the possibilities of personhood will certainly be needed for meaningful public discourse given the anthropomorphizing already occurring with AI. As a step toward these efforts, I briefly review existing work in theology centered around aspects of personhood, such as, imago Dei, theological anthropology, and morality. Although the review is certainly not exhaustive, it hopefully represents the field sufficiently to orient new scholars to the area.
This paper presents a case study on the design, administration, post-processing, and evaluation of surveys on large language models (LLMs). It comprises two components: (1) A statistical method for eliciting beliefs encoded in LLMs. We introduce statistical measures and evaluation metrics that quantify the probability of an LLM "making a choice", the associated uncertainty, and the consistency of that choice. (2) We apply this method to study what moral beliefs are encoded in different LLMs, especially in ambiguous cases where the right choice is not obvious. We design a large-scale survey comprising 680 high-ambiguity moral scenarios (e.g., "Should I tell a white lie?") and 687 low-ambiguity moral scenarios (e.g., "Should I stop for a pedestrian on the road?"). Each scenario includes a description, two possible actions, and auxiliary labels indicating violated rules (e.g., "do not kill"). We administer the survey to 28 open- and closed-source LLMs. We find that (a) in unambiguous scenarios, most models "choose" actions that align with commonsense. In ambiguous cases, most models express uncertainty. (b) Some models are uncertain about choosing the commonsense action because their responses are sensitive to the question-wording. (c) Some models reflect clear preferences in ambiguous scenarios. Specifically, closed-source models tend to agree with each other.
Yiran Mao, Madeline G. Reinecke, Markus Kunesch
et al.
Is it possible to evaluate the moral cognition of complex artificial agents? In this work, we take a look at one aspect of morality: `doing the right thing for the right reasons.' We propose a behavior-based analysis of artificial moral cognition which could also be applied to humans to facilitate like-for-like comparison. Morally-motivated behavior should persist despite mounting cost; by measuring an agent's sensitivity to this cost, we gain deeper insight into underlying motivations. We apply this evaluation to a particular set of deep reinforcement learning agents, trained by memory-based meta-reinforcement learning. Our results indicate that agents trained with a reward function that includes other-regarding preferences perform helping behavior in a way that is less sensitive to increasing cost than agents trained with more self-interested preferences.
Moral or ethical judgments rely heavily on the specific contexts in which they occur. Understanding varying shades of defeasible contextualizations (i.e., additional information that strengthens or attenuates the moral acceptability of an action) is critical to accurately represent the subtlety and intricacy of grounded human moral judgment in real-life scenarios. We introduce defeasible moral reasoning: a task to provide grounded contexts that make an action more or less morally acceptable, along with commonsense rationales that justify the reasoning. To elicit high-quality task data, we take an iterative self-distillation approach that starts from a small amount of unstructured seed knowledge from GPT-3 and then alternates between (1) self-distillation from student models; (2) targeted filtering with a critic model trained by human judgment (to boost validity) and NLI (to boost diversity); (3) self-imitation learning (to amplify the desired data quality). This process yields a student model that produces defeasible contexts with improved validity, diversity, and defeasibility. From this model we distill a high-quality dataset, δ-Rules-of-Thumb, of 1.2M entries of contextualizations and rationales for 115K defeasible moral actions rated highly by human annotators 85.9% to 99.8% of the time. Using δ-RoT we obtain a final student model that wins over all intermediate student models by a notable margin.
Intersectional approaches aim to uncover the multidimensionality of multiply burdened victimhood rather than a single-axis analysis. When intersected with Asian and Asian American postcolonial experiences and perspectives, intersectionality exposes the global reach and colonial origins of white imperialism and ideological, systemic, and mutational nature of white supremacist logic of dominance connected to global capitalism, neocolonialism, class-ism, ableism, caste-ism, racism, etc. While limited and inadequate by itself, there is a family resemblance of intersectional method in the Catholic theological and intellectual tradition that is inclusive of ordinary and marginalized voices and secular disciplines. For Catholic theologians and the church, then, intersectional approach can be instrumental in dismantling white supremacy within the church as an institution and people. Indeed, the intersectional nature of Jesus’ ministry, personhood, and location further solidifies the case for using intersectional theology for ecclesial and personal introspection, growth, and development.
Jennie Weiss Block, OP, M. Therese Lysaught, Alexandre A. Martins
It takes a village to write a book. This village includes not only the co-editors and contributors, but the editorial staff of the _Journal of Moral Theology_, the Catholic Theological Ethics in a World Church, as well as those who made Paul Farmer possible: his colleagues at Partners In Health, his family, and his patients.
Cyber insurance is a risk-sharing mechanism that can improve cyber-physical systems (CPS) security and resilience. The risk preference of the insured plays an important role in cyber insurance markets. With the advances in information technologies, it can be reshaped through nudging, marketing, or other types of information campaigns. In this paper, we propose a framework of risk preference design for a class of principal-agent cyber insurance problems. It creates an additional dimension of freedom for the insurer for designing incentive-compatible and welfare-maximizing cyber insurance contracts. Furthermore, this approach enables a quantitative approach to reduce the moral hazard that arises from information asymmetry between the insured and the insurer. We characterize the conditions under which the optimal contract is monotone in the outcome. This justifies the feasibility of linear contracts in practice. This work establishes a metric to quantify the intensity of moral hazard and create a theoretic underpinning for controlling moral hazard through risk preference design. We use a linear contract case study to show numerical results and demonstrate its role in strengthening CPS security.