AI and ethics最新文献

{"title":"A postphenomenology of care: ecstatic temporality in generative music","authors":"Sam Miller","doi":"10.1007/s43681-026-01386-8","DOIUrl":"10.1007/s43681-026-01386-8","url":null,"abstract":"<div><p>This paper asks what kind of creative subject is constituted by the default consumer path of text-to-audio generators such as Suno and Udio, in which a prompt yields a finished song. Peter-Paul Verbeek’s postphenomenology describes the mediation precisely: unlike a traditional instrument in an embodiment relation, the platform is addressed as a quasi-other whose output is then read and appraised. Verbeek’s <i>human-world-technology</i> analysis is accurate but lacks a criterion for what is unmusical in this mediation. The paper takes Heidegger’s <i>Sorge</i>, read through the ecstatic temporality of <i>Being and Time</i>, to test whether a design leaves the user inside an unfinished <i>musical present</i>. Music makes the test unusually clear, because it is given as temporal flow: <i>retention</i> of what has just sounded, impression of the sounding now, and <i>protention</i> of the next phrase. A prompt-to-song file arrives already closed, collapsing the temporal stretch characteristic of music making. However, other mediations such as stem-separation, uploaded performance augmentation, and automated mixing and mastering can leave this temporal field open. The design implication is modest: the default path should not unlock a releasable song from text alone.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986359","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":"Co-design of a trustworthy AI-based prognostic tool for predicting patient outcome in acute stroke","authors":"Elizabeth Hofvenschioeld,&nbsp;Adam Hilbert,&nbsp;Cathrine K. T. Bui,&nbsp;Susanne Bonekamp,&nbsp;Tess Buckley,&nbsp;Roberta Calegari,&nbsp;Luca A. Cappellini,&nbsp;Megan Coffee,&nbsp;Giorgio Colangelo,&nbsp;Boris Düdder,&nbsp;Francesco Di Tano,&nbsp;Dietmar Frey,&nbsp;John D. Kelleher,&nbsp;Pedro Kringen,&nbsp;Andreane Sabourin Laflamme,&nbsp;Emma Moorhead,&nbsp;Pedro A. Moreno-Sánchez,&nbsp;Gabriele Pluktaite,&nbsp;Marta Rubiera,&nbsp;Giovanni Sartor,&nbsp;Jesmin Jahan Tithi,&nbsp;Malte von Tottleben,&nbsp;Ingo Werren,&nbsp;Magnus Westerlund,&nbsp;Renee Wurth,&nbsp;Roberto V. Zicari,&nbsp;Vince I. Madai","doi":"10.1007/s43681-026-01337-3","DOIUrl":"10.1007/s43681-026-01337-3","url":null,"abstract":"<div><p>This paper reports a descriptive, process-oriented case study of the application of Z-Inspection®, an ethically aligned co-design methodology, in the early design phase of an artificial intelligence (AI) system for healthcare. The methodology was applied within the Horizon Europe project VALIDATE [1], which develops and validates a prognostic clinical decision support system (CDSS) to support treatment decisions in acute ischaemic stroke. An interdisciplinary team of Z-Inspection® experts, AI developers, and clinical stakeholders jointly identified ethical, legal, and technical issues relevant to the planned system. The assessment resulted in 22 ethical issues, 12 dilemmas, 18 risks, and 48 derived requirements, each mapped to the European Commission’s trustworthy AI principles. The identified issues, dilemmas, risks, and derived requirements constitute the primary findings of this descriptive case study and document how general trustworthy-AI principles were translated into context-specific technical, clinical, governance, and organisational requirements. By documenting the process and publishing the complete requirements, the study provides a detailed case for critical scrutiny and potential uptake in other healthcare AI projects, though its transferability and scalability remain to be established in further studies. </p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01337-3.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986357","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":"A Searlean perspective on using LLM in scholarly work","authors":"Boon-Seng Tan","doi":"10.1007/s43681-026-01382-y","DOIUrl":"10.1007/s43681-026-01382-y","url":null,"abstract":"<div><p>Large language models (LLMs) have rapidly become integrated into scholarly research, yet academic publishers have not reached consensus on appropriate usage policies. Two consensuses are nonetheless emerging in the literature: LLMs cannot be authors, and their use requires disclosure. This paper argues that these consensuses currently rest on justifications that are incomplete or contestable, chiefly the claim that LLMs lack moral or legal responsibility without an account of why this is so, and it supplies the missing theoretical grounds using Searle’s Chinese Room argument. By demonstrating that LLMs perform sophisticated symbol manipulation without genuine understanding, the Searlean perspective shows that LLMs cannot satisfy the epistemic condition that responsibility of any kind presupposes, thereby grounding the authorship and disclosure consensuses in the nature of LLMs rather than in contested claims about free will. The analysis extends beyond these basic principles to examine nuanced questions of LLM-assisted writing and data analysis, where the same framework distinguishes legitimate tool use from problematic delegation of intellectual work. Rather than proposing a new policy framework, this paper supplies the theoretical foundation that existing consensuses on LLM authorship and disclosure have so far lacked.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986356","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":"The delegation illusion: why deploying autonomous AI agents does not diminish principal responsibility","authors":"Chun Yin Kong","doi":"10.1007/s43681-026-01383-x","DOIUrl":"10.1007/s43681-026-01383-x","url":null,"abstract":"<div><p>A growing literature on “agentic AI” — autonomous software agents that plan and execute multi-step actions on a principal’s behalf — has revived the thesis that such systems open a <i>responsibility gap</i>: because the deploying principal neither intends, foresees, nor controls the specific actions an autonomous agent selects, no human can be held fully responsible for resulting harms. This paper argues that the responsibility-gap thesis, as applied to principal-deployed AI agents, rests on a conflation called here the <i>delegation illusion</i>: the inference from a principal’s causal and epistemic remoteness from an outcome to a diminution of her answerability for it. Distinguishing attributability, answerability, and accountability, the paper argues that autonomy, opacity, and adaptivity bear on attributability while leaving intact an answerability grounded, through a standard tracing structure, in the deployer’s guidance control over the prior act of delegation. It defends a principle of <i>responsibility conservation under delegation</i> and operationalises that principle’s central criterion — foreseeability of harmful dispositions at the level of types — by specifying an epistemic base, a standard of diligence, and a constraint on type individuation that blocks vacuous redescription. Answerability is argued to be epistemically graded, but indexed to the deployer’s ex ante access to the dispositional profile of the deployment specification rather than to the particular act. The method is conceptual analysis and normative argumentation; the scope and limitations of the claims are stated explicitly.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01383-x.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986361","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":"Measuring structural value alignment in sixteen models: LLMs have human-like moral spaces","authors":"Yexiang Tang","doi":"10.1007/s43681-026-01368-w","DOIUrl":"10.1007/s43681-026-01368-w","url":null,"abstract":"<div><p>Evaluating whether large language models (LLMs) reason about morality in human-like ways requires more than measuring whether they produce the right outputs in isolated cases. Existing approaches – including scalar agreement, distributional analysis, rationale classification, and consistency testing – compare model and human responses one case at a time and cannot detect how a model structures its moral judgments. This paper applies and extends Peterson and Gärdenfors's (2024) geometric moral space framework to evaluate structural value alignment in LLMs, that is, whether the relational structure reconstructed from a model's responses matches the structure reconstructed from human respondents on the same task. Sixteen LLMs across three scale tiers (small, medium, and large) were evaluated on ten moral cases under both zero-shot and chain-of-thought (CoT) prompting. The results reveal phenomena invisible to case-by-case metrics. First, all five large models achieve full geometric alignment on these cases (moral space overlap, MSO = 1.00) in at least one prompting condition, and three do so in both; two medium models also reach MSO = 1.00. Second, structural alignment is bimodal in the sense that models either replicate the human moral space or fall substantially short, with no intermediate cases. Third, smaller models systematically lack a distinct fairness region, a failure of moral categorization rather than of labeling. Fourth, CoT prompting simultaneously improves case-by-case similarity correlations and worsens geometric alignment in smaller models, showing that the two can move in opposite directions. Because structural alignment is measured from model responses, it does not by itself establish that a model holds the values in question. It is, however, a more demanding test than case-by-case agreement, and a better motivated and more informative basis for value alignment measurement.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01368-w.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986360","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":"Knowing whether to rely: continuous awareness as a moral right in human-AI interaction","authors":"Khaled Sawaf","doi":"10.1007/s43681-026-01387-7","DOIUrl":"10.1007/s43681-026-01387-7","url":null,"abstract":"<div><p>A person relying on a system has a right to know what it is; disclosure secures that right but only while the object of reliance remains materially stable. Under AI, the object does not stay still, and a model’s integrity can change without warning. A moment after disclosure is issued, the system described may have evolved, drifted, or been compromised while the user continues relying on it unaware of its current state. A person’s interest in what they rely on is continuous, but disclosure is episodic, and that epistemic mismatch is not closed by repeating the notice. This paper reconceives AI transparency around a single claim: A person relying on an AI system holds a moral right to continuous awareness of the system’s status, grounded in the same autonomy that informed consent protects. How continuous awareness is delivered is the corollary of that right. The status signal that delivers it must be continuous and vouched from outside the system, since a captured system cannot certify its own integrity. Shaping that signal is therefore an ethical act, not an engineering detail.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://link.springer.com/content/pdf/10.1007/s43681-026-01387-7.pdf","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986294","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 arbitrage: ethical tensions in AI agent accountability infrastructure","authors":"Alex Chengyu Li","doi":"10.1007/s43681-026-01369-9","DOIUrl":"10.1007/s43681-026-01369-9","url":null,"abstract":"<div><p>AI agents leave operational logs, tool traces, and platform receipts. Newer systems connect these fragments, expose later changes, and support independent verification. As reviewability improves, recordability becomes an allocation factor, not only an engineering property. This article develops that claim. Human and AI-agent workflows may perform the same task while producing records of different institutional value. A regulator, auditor, insurer, court, or internal review body may find one history easier to review. Organizations anticipate that difference when designing workflows and assigning tasks. Formal acceptance can therefore influence allocation before a common technical or legal standard forms. The strict selection effect is <i>accountability arbitrage</i>: an expected record advantage reverses a work choice that performance and ordinary cost would otherwise make. The same infrastructure also creates a dual effect. A durable record may help prove compliance and make error or misconduct easier to establish. Selective movement away from a recorded workflow in response to that exposure is <i>accountability avoidance</i>. The analysis connects operational recording to actor–forum accountability. A record preserves material for judgment; an empowered forum gives that material institutional force. Until that handoff, the record is <i>evidence without a court</i>. A minimal model identifies the strict arbitrage region and threshold discontinuity; a brokerage example holds performance constant and shows record value reversing the choice. We predict that expected record advantage will increase agent selection near evidence thresholds, while anticipated exposure will reduce recorded deployment in high-discretion tasks. Existing recordkeeping practice and enforcement provide the basis for these predictions. Matched studies finding neither response under the stated contrasts would count against the theory. Governance implications include human–AI evidentiary parity, substantive-performance floors, review authority, and proportionate data collection.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986353","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":"Human-weighted ethical pluralism in AI decision support: introducing the weighted meta-decision engine (WMDE) through the Omelas dilemma","authors":"Ahmet Göçen,&nbsp;Mete Han Arıtürk,&nbsp;Mohannad Alhanahnah","doi":"10.1007/s43681-026-01356-0","DOIUrl":"10.1007/s43681-026-01356-0","url":null,"abstract":"<div><p>Artificial intelligence (AI) systems are increasingly involved in decisions with ethical consequences. Existing approaches to Moral AI include bottom-up moral learning, theory-guided prompting, pluralist hybrid architectures, and methods that aggregate outputs from multiple ethical theories. However, when different ethical theories recommend different actions, an important question remains: how should their relative influence be determined without allowing the AI system or its developer to decide which theory should carry greater moral authority? In answer to this question, this conceptual paper introduces the Weighted Meta-Decision Engine (WMDE), a human-governed, AI-assisted procedure that separates several functions: selecting relevant ethical theories, generating theory-specific AI proposals, evaluating those proposals, and combining them through human-derived weights. We demonstrate WMDE using Ursula K. Le Guin’s <i>The Ones Who Walk Away from Omelas</i> dilemma and seven recorded LLM outputs. Two outputs recommend STAY and five recommend WALK AWAY. In a hypothetical example, however, human endorsements give the two STAY proposals 60% of the total weight and the five WALK AWAY proposals 40%. WMDE therefore bases the final recommendation on human evaluation of concrete AI-generated ethical arguments rather than simply counting theories. The framework is intended for decision support, not autonomous moral authority.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986358","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":"Anticipatory agency and the ecology of causal bubbles: a pragmatic constructivist framework for human-AI collaboration","authors":"Michael Lissack","doi":"10.1007/s43681-026-01372-0","DOIUrl":"10.1007/s43681-026-01372-0","url":null,"abstract":"<div><p>This paper develops a pragmatic constructivist framework for understanding human-AI collaboration by synthesizing Michel Callon’s economic sociology of market agencements with Robert Rosen’s anticipatory systems theory. Key terms are used as follows: ‘Causal bubbles’ are defined as maintained zones of operational regularity in which anticipatory coordination is viable. ‘Anticipatory agency’ refers to the capacity to act based on internal models of projected futures rather than just reacting to present stimuli. ‘Agencement’ describes the assemblage of human and nonhuman elements that together enable agency. Rather than making ontological claims about agency or causation, the framework proposes ‘causal bubbles’ as a vocabulary for describing maintained zones of operational regularity within which anticipatory coordination becomes viable. The synthesis addresses conceptual impasses in AI responsibility discourse by shifting from questions of blame attribution to questions of systemic maintenance. Drawing on Hans Vaihinger’s philosophy of ‘as if,’ I argue that ethical evaluation of human-AI configurations should focus not on metaphysical questions of intelligence or consciousness, but on whether such configurations preserve or degrade the conditions for effective coordinated action across cognitive, social, and ecological scales. The framework offers heuristic indicators for assessing ‘bubble health,’ a three-phase repair cycle for restoring degraded configurations, and applications to pedagogy, research integrity, and governance. Throughout, the paper maintains epistemological modesty: the constructs offered are pragmatic tools for coordination and intervention, not truth claims about underlying mechanisms.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986433","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":"When is AI-mediated institutional action ethically warranted? A feasibility-first framework for admissible-state reasoning","authors":"Adegoke O. Adefolalu","doi":"10.1007/s43681-026-01393-9","DOIUrl":"10.1007/s43681-026-01393-9","url":null,"abstract":"<div><p>Artificial intelligence systems increasingly mediate institutional action through forecasts, rankings, risk scores, allocation tools, automated alerts and decision-support infrastructures. These outputs can shape who receives attention, which risks are escalated, how resources are distributed and when institutions intervene, defer or withhold action. Ethical evaluation commonly addresses accuracy, fairness, transparency, explainability and human oversight. These criteria are necessary but do not by themselves establish whether acting on an AI output is ethically warranted under an institution’s current legal authority, operational capacity, temporal window, accountability structure and remedial arrangements. This paper develops a feasibility-first framework for admissible-state reasoning within Deterministic Systems Intelligence (DSI). It distinguishes continuity with earlier DSI release-governance work from a further theoretical question: whether an institution is justified in exercising consequential authority through a proposed AI-mediated action. Institutional action is represented as movement within an admissible action space shaped by governance clocks, regime configurations, structural memory, capacity ceilings, accountability gates and implementation latency. A layered governance operator separates release disposition, governance condition and boundary status while retaining the operational vocabulary RELEASE, MONITOR, ESCALATE, HOLD and BREACH. The central claim is bounded: an action should not be treated as ethically available merely because it is technically possible, statistically supported or institutionally preferred. Ethical warrant also requires legitimate authority, proportionality, feasible implementation, consideration of affected parties, contestability, monitoring and access to remedy; where an AI-mediated pathway is withheld, continuing institutional obligations may still require an alternative action. The framework therefore extends responsible AI assessment from model properties and operational release-readiness to the institutional conditions under which algorithmic outputs acquire ethically justified operational force.</p></div>","PeriodicalId":72137,"journal":{"name":"AI and ethics","volume":"6 5","pages":""},"PeriodicalIF":0.0,"publicationDate":"2026-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148986434","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}