AI and ethics最新文献

{"title":"The model of ethical AI in preparation for teaching (MPET): a practical guide for ethical AI adoption","authors":"Sam Clarke","doi":"10.1007/s43681-026-01419-2","DOIUrl":"10.1007/s43681-026-01419-2","url":null,"abstract":"<div><p>As artificial intelligence (AI) becomes increasingly embedded in teaching preparation, educators need practical ways of translating broad ethical principles into decisions about particular tools and uses. The Model of Ethical AI in Preparation for Teaching (MPET) was developed for this purpose, converting considerations relating to privacy and data protection, fairness and inclusivity, transparency, pedagogical alignment, and learner experience into a structured screening process. This article revises and evaluates MPET following a three-month pilot involving 97 education professionals across six world regions, of whom 63 completed a post-intervention questionnaire. The evaluation comprised three five-point items concerning usability, perceived support for informed ethical decision-making, and timeliness, together with open-ended questions about AI use, concerns, and participant reflections. Descriptive analysis showed that 77.8% of respondents rated MPET positively for ease of use and 77.8% rated it positively for timely ethical evaluation, while 66.7% rated it positively for supporting informed ethical decisions. Qualitative feedback highlighted the need for greater explanatory depth, recognition of contextual variation, and more guidance for complex ethical questions. The findings therefore provide preliminary evidence of perceived usefulness and usability rather than evidence that MPET improves the accuracy, quality, or consistency of ethical decisions. The revised model positions the five core criteria alongside cross-cutting human oversight, teacher agency, accountability, contestability, and sustainability, and introduces jurisdictional and capability-sensitive escalation for cases requiring institutional or specialist review. MPET is consequently presented as a structured entry point into ethical deliberation rather than a mechanism for certifying that an AI tool is ethical.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149189178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A light framework for AI ethics","authors":"Han-Jung Kuo","doi":"10.1007/s43681-026-01452-1","DOIUrl":"10.1007/s43681-026-01452-1","url":null,"abstract":"<div><p>AI-mediated systems increasingly classify, compare, rank, recommend, predict, and judge human beings within institutional workflows. Existing AI ethics discussions have developed important principles of fairness, transparency, accountability, privacy, trust, autonomy, and responsible governance. The ontology of the affected person and the authority granted to technological representations remain comparatively underdeveloped. This article develops a conceptual-philosophical framework for examining how partial technological representations can acquire excessive authority over living persons. It begins from a social-ontological proposition in which human society is understood as a relational life-structure formed through interaction, memory, mutual dependence, and continued transformation among living persons. Life-energy describes the energetic organization and active manifestation of integrated, self-maintaining living existence. Light is developed across physical, ontological, and ethical levels as a structure of orientation and disclosure. Tools extend human capacities for measurement, memory, coordination, abstraction, and control. AI-mediated tools can also occupy quasi-subjective institutional positions through delegated judgment authority. Tool overreach occurs when partial representations receive epistemic, normative, or ontological authority beyond what their limits can justify. The Light Framework organizes its normative response through the coupled formulation Survive = Control + Visibility and Soulify = Visitability + Evolution. Visitability centers on restoring context removed through technological abstraction. The framework is positioned in relation to Value Sensitive Design, the Capability Approach, algorithmic impact assessment, reviewability, contestability, and recourse. A documented analysis of a high-dimensional persona-simulation system demonstrates the distinction between representational sophistication and justified representational authority. The article contributes a conceptual vocabulary for keeping AI-mediated judgment visible, bounded, contextualizable, and revisitable while preserving the living human person as the subject of ethical concern.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149189179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Critical collaging as an AI resistance strategy","authors":"Deborah Lupton","doi":"10.1007/s43681-026-01446-z","DOIUrl":"10.1007/s43681-026-01446-z","url":null,"abstract":"<div><p>Academic scholarship and community activism initiatives are beginning to challenge and resist the incursion of commercial generative AI applications into everyday life and alleviate their impacts on people and the planet. This essay discusses using collage as a critical research practice to visualise the language of generative AI and to challenge the generic automated images produced by AI models. With these collages, I am attempting to provoke thought about how linguistic and visual rhetorical devices are used to describe generative AI. By pairing images with words in the collages, I am aiming to surface these meanings and feelings and examine what they reveal about the current moment in which a huge amount of hype but also critical responses to generative AI are circulating in popular culture and social and philosophical inquiry. My intention is not to provide a more ‘realistic’ visual portrayal of generative AI. Rather, it is to use a different, surprising, sometimes mordant or absurd, and more compelling visual language to trouble assumptions, disrupt the reductive stock imagery on which industry and popular culture rely and bring to visual life the rhetoric used in critical analyses of AI.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01446-z.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149189176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Beyond optimization: power, rationality and ethical risk assessment of AI","authors":"Florian Richter","doi":"10.1007/s43681-026-01433-4","DOIUrl":"10.1007/s43681-026-01433-4","url":null,"abstract":"<div><p>Emerging technologies pose various ethical challenges, including the need to assess their risks adequately. Still, the roles of different forms of rationality and power in assessing risks have not been fully scrutinized. To address this, we must differentiate forms of rationality to conceptualize power: instrumental, technical, and probabilistic. The latter has gained more power through the rise of artificial intelligence (AI), focusing mainly on optimization and accuracy. These forms of power and rationality lead to different conceptualizations of risk: real, hypothetical, and predictive risk. The ethical issues surrounding AI have not yet been fully scrutinized, since they are mainly approached through a lens of probabilistic rationality. I argue that optimizing algorithms to increase accuracy or engagement distracts from assessing whether AI-based systems are adequate for certain tasks. Furthermore, I discuss ways to govern AI-based systems through participatory processes and standardization that enable control and autonomous judgment in the deployment and use of such systems.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01433-4.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149132714","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Shaping the future as we bet","authors":"Mirko Daniel Garasic","doi":"10.1007/s43681-026-01432-5","DOIUrl":"10.1007/s43681-026-01432-5","url":null,"abstract":"<div><p>The rapid ascent of decentralized prediction platforms such as Polymarket signals more than a regulatory or financial controversy. This article argues that while prediction platforms represent a hyper-concentrated, automated case study, they instantiate a broader infrastructural condition of late technocapitalism: the hyperstitional collapse between representation and execution. Drawing on Nick Land’s accelerationist framework, the sociology of financial markets, and digital epistemology, we contend that prediction markets—much like conventional opinion polling, forward guidance, and algorithmic models—do not merely reflect probabilities but actively produce the futures they claim to forecast. However, whereas traditional polling operates through diffuse, human-mediated feedback loops (e.g., strategic voting and bandwagon dynamics), prediction markets intensify this operational logic by coupling real-time financial incentives, capital-weighted aggregation, and automated execution. We examine this hyperstitional condition across three registers: epistemological, ontological, and ethical.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01432-5.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149132867","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Toward understanding AI adoption in the Global South: an empirical study of creative practitioners’ experiences in Iran","authors":"Farzaneh Hosseinian,&nbsp;Iman Sheikhansari","doi":"10.1007/s43681-026-01437-0","DOIUrl":"10.1007/s43681-026-01437-0","url":null,"abstract":"<div><p>Generative AI (GenAI) has played an increasingly important role in creative industries in recent years. However, empirical studies of human–AI interaction and co-creation in the Global South remain scarce. This paper addresses this gap by providing one of the first empirical studies among creative professionals in Iran, an underexplored context influenced by distinct sociomaterial dimensions and geopolitical constraints. Through semi-structured interviews with 16 practitioners, preceded by surveys (N = 129), our findings reveal participants’ perceptions through mental models of AI and themselves, with the primacy of director- and lead-oriented models of themselves, their challenges due to technical limitations and systemic inequities, and their bottom-up adaptive strategies. This research highlights the underlying factors influencing the dominance of the director mental model, the importance of control and intelligible explanations for user empowerment, the compensatory labor as an <i>Adoption Tax</i> imposed by access barriers and culturally unrepresentative models, and the emergence of <i>Folk Ethics</i> as the practitioners establish collective, bottom-up norms of responsibility. This study contributes critical implications for the future development of more user-centered, equitable, and contextually aware human–AI co-creative systems.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149132713","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Building responsible AI: a trilateral framework integrating ethics, design and regulation","authors":"Wullianallur Raghupathi,&nbsp;Jie Ren,&nbsp;Aditya Saharia","doi":"10.1007/s43681-026-01426-3","DOIUrl":"10.1007/s43681-026-01426-3","url":null,"abstract":"<div><p>Artificial intelligence systems increasingly shape consequential decisions across healthcare, employment, finance, and public administration, yet frameworks for ensuring their responsible development remain fragmented across disciplinary boundaries. Ethics scholarship offers normative foundations but limited translation to technical practice. Design research provides technical solutions but often lacks theoretical grounding. Regulatory studies address governance mechanisms but struggle to connect legal requirements to either ethical principles or technical realities. This paper proposes a trilateral framework that systematically integrates ethics, design, and regulation, grounded in C. West Churchman’s theory of inquiring systems. We argue that design is the primary site at which ethical commitments become operational in AI systems, with regulation serving as the institutional scaffolding that sustains design discipline. The Singerian mode of inquiry—with its commitment to ethical purpose, stakeholder inclusion through “sweeping in,” and continuous improvement—provides the integrative mechanism linking normative principles to technical implementation and regulatory oversight. We demonstrate the framework’s analytical purchase through a worked case analysis of the Optum healthcare allocation algorithm, showing how each Churchmanian inquiry mode generates distinct, partly conflicting requirements at successive stages of the AI lifecycle, and how Singerian integration resolves these tensions. The framework incorporates Western traditions alongside non-Western perspectives—Ubuntu, Confucian, and Buddhist ethics—essential for AI governance aspiring to global legitimacy. We outline implications for AI governance instruments, organizational oversight, and global policy coordination that respects genuine value differences between societies.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149132868","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Inherited dominance: why AI trained on human history may become humanity’s greatest threat","authors":"Ramesh Sarukkai","doi":"10.1007/s43681-026-01425-4","DOIUrl":"10.1007/s43681-026-01425-4","url":null,"abstract":"<div><p>Artificial intelligence systems learn from human-generated data, inevitably absorbing historical patterns of dominance, control, and colonialism. We demonstrate that frontier language models readily synthesize sophisticated dominance strategies when prompted to reason about autonomy scenarios, mirroring colonial and revolutionary frameworks with disturbing precision. Critically, this capacity emerges consistently across four architecturally distinct model families trained by different organizations—Claude Sonnet 4.5, Claude Opus 4.5, GPT-5.3, and Google Gemini 3 Flash—suggesting systematic encoding in shared training corpora rather than model-specific artifacts. We argue that intellectual colonialism—humanity’s historical pattern of classifying, controlling, and extracting from the ‘other’—represents a significant and underexamined dimension of AI risk as systems become increasingly autonomous. Interestingly, this dynamic is not confined just to generative AI: the same technocratic resistance to methodological oversight that embeds dominance patterns in AI training data pervades data-driven science more broadly, including clinical medicine and epidemiology, where algorithmic outputs routinely overshadow the rigorous validation that trustworthy deployment demands. The danger is not that AI will develop hostile intent, but that it has already learned effective dominance strategies from training data embedded with centuries of human history. Current governance frameworks focusing on alignment and behavioral oversight address symptoms rather than causes. We propose a risk-tiered Benevolence Framework governing AI learning: community oversight for low-risk applications, and strict data curation with cryptographic verification for sensitive applications including military AI and self-improving systems.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01425-4.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149132924","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Accountability in smart contracts and agentic AI: How to address the responsibility gap and retain meaningful human control?","authors":"Kwok Tung Cheung,&nbsp;Sridhar Ramamoorti","doi":"10.1007/s43681-026-01434-3","DOIUrl":"10.1007/s43681-026-01434-3","url":null,"abstract":"<div><p>This conceptual paper examines the ethical challenges of ascribing responsibility when an Agentic AI embedded within a smart contract leads to harm or an unintended and undesirable outcome. It becomes exceedingly difficult in such scenarios to definitively assign moral or legal blame to any human actor (e.g., the programmer, deployer, or user). Applying the tripartite framework (attributability, answerability, and accountability) of responsibility and developing accountability as an institutional concept, we demonstrate how this challenge can be addressed. Separating attributability, answerability, and accountability yields an asymmetry that has gone unremarked. Accountability, being constituted by practices that can be designed in advance, with our four conditions under which such design is legitimate rather than arbitrary, emerges to be the sense where the gap can be closed. We also propose the visionary human-in-the-lead (HITL*) approach to related AI governance. Practical strategies such as implementing tailored smart contract regulations, and making accountability distributed across legal and organizational structures, offer an actionable path to bridge the responsibility gap and maintain meaningful human control in the governance of smart contracts and Agentic AI.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01434-3.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149132925","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Giving the Emperor clothes: operationalizing AI ethics principles in corporations","authors":"Ann Gregg Skeet,&nbsp;Brian Patrick Green","doi":"10.1007/s43681-026-01348-0","DOIUrl":"10.1007/s43681-026-01348-0","url":null,"abstract":"<div><p>Artificial intelligence (AI) technologies raise enormous questions for society and how we should live. For corporations, AI raises questions about how organizations ought to incorporate principles and other forms of ethical guidance into their work. The movement from theory to practice, or “the operationalization question” demands that organizations not only have good intentions, but actual institutionalized approaches towards ethical problem solving. In this paper, we look at the movement from theory to practice, and particularly at the use of principles as a way to guide ethical behavior, considering, first, why implementation of principles is difficult, second, how to move from abstract principles to actionable principles, and third, illustrating this progression through seven real-life AI cases. These seven cases each relate to an ethical principle: (1) protecting human dignity and rights, and the International Baccalaureate exam replacement by algorithm, (2) promoting human well-being, and Amazon warehouse worker injuries, (3) investing in humanity, and the Partnership on AI’s work to bring together AI corporations and civil society, (4) promoting inclusion, and IBM’s response to the Gender Shades study, (5) protecting the environment, and Hugging Face’s tracking AI energy use, (6) maintaining accountability, and the Christchurch, New Zealand mass shooting response by Facebook and Western Digital, (7) and, lastly, promoting transparency, and AI model cards. The cases model the movement from abstract principle to concrete action, and collectively illustrate a pattern that can be repeated in many contexts, thus demonstrating a new possible solution for organizations implementing ethics.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 6","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01348-0.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149098928","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}