Contemporary smart homes fail to address the built environment’s climate crisis contribution, being trapped in a mechanistic paradigm that decouples intelligence from the body, marginalises its human occupants, and elevates passive consumption over regeneration. We explore a novel conceptual framework—the Cyberphysical Living System (CPLS) that rethinks smart home ontologies by synthesising metabolic theory, cybernetics, and semantic architectures into a hybrid human–intelligent system, asking: what would it mean to design a home that metabolises rather than consumes? Grounded in validated bioelectrochemical and AI technologies, this model envisions the home as an ecosystem of actors in tripartite symbiosis: a metabolic 'gut' (integrated bioreactors), an explainable cognitive ‘brain’ (a cognitive digital twin), and human occupants. Central to this vision is a proposed hybrid ontology—a semantic framework that bridges architecture and computer science that translates architectural metabolic design patterns (codified biological design principles) into a machine-readable format, which could enable AI to adapt to dynamic biological processes and human activities beyond data monitoring. We outline a metabolic-based methodology centred on hybrid human–AI systems. Exploring its societal implications, we also consider redefining comfort as metabolic security, transforming occupants into literate stewards, and repositioning the home as a regenerative and productive asset. We conclude the future of sustainable dwelling may lie in cultivating living architectural metabolisms—which could re-embed human life within the living world's regenerative logic. The paper’s original contribution lies in synthesising metabolic architecture with hybrid AI systems through a proposed hybrid ontology, offering a new conceptual foundation for human-centred, regenerative design.
Large language models are metaphor machines: they generate text based on similarities and proximity in their training data. I argue that this metaphor-based text generation leads to what John Gallagher calls orbital argumentation, where the generated text is pulled towards key concepts that, I argue, remain unnarratable as the text skims their surface, orbiting around them instead of directly addressing them. Misaligned citations in scholarly writing—citations of real works that are related to the central topic but do not support the specific claim made—are a type of scientific fabrication that is systematically produced by the fundamentally metaphorical operation of large language models. To develop this argument, I draw upon theories of metaphor by Aristotle and I. A. Richards, applying them to two technological shifts that underpin the metaphorical emphasis of LLMs: the 2013 discovery that language models could handle analogies like king − man + woman = queen and the introduction of self-attention and transformer models in 2017. I use this framework to analyse an LLM-generated document summary, examples from a dataset of LLM-generated stories, and a scientific paper. In each case, I demonstrate how the metaphorical connections pull at the text generation, causing the text to orbit around a key theme that remains unnarratable. Recognising the rhetorical vices of LLM-generated texts can help us to protect the knowledge ecosystem that science depends upon. The paper, therefore, concludes with five recommendations for readers, writers, peer reviewers, and editors who want to build resilience against the inaccuracies of LLM-generated texts.
Automated vehicles are widely expected to play a key role in achieving the European Union’s road safety target of ‘Vision Zero’, which aims to eliminate all road fatalities in the European Union by 2050. To achieve this goal, early design strategies for automated vehicles prioritised strict compliance with traffic rules, assuming that rule-abiding behaviour would inherently lead to safer outcomes. Yet rule deviations such as speeding may, counter-intuitively, enhance safety in situations where strict compliance cannot reasonably be expected. Such cases highlight a discrepancy between legal compliance and the informal norms that often govern complex real-world traffic. They also expose an ambivalence in compliance. It is often framed as a necessary precondition for safety, yet in certain contexts, strict compliance can itself generate avoidable risks. This design challenge is what I call the ‘speed offset problem’. The problem becomes particularly evident when examined across different cultural contexts, where expectations around acceptable speeding vary significantly. Building on this, I argue that ‘principled cultural sensitivity’ is fundamental for ethically sound design of automated vehicles. As a normative framework, it demands sensitivity to local driving cultures whilst maintaining firm commitment to core ethical principles. The article develops and applies this framework to the speed offset problem, exemplifying how these guardrails may take shape in concrete cases. Three ethical guardrails restrict the speed offset’s implementation. First, a speed offset is morally permissible only when it increases overall safety. Second, a speed offset must not increase overall safety at the expense of particular groups. Third, a speed offset’s legitimacy must be temporary and revisable. Together, these guardrails ensure that speed offsets are not a concession to unsafe norms but a temporary and ethically guided calibration, bridging current traffic realities with long-term safety goals. The article concludes by outlining implications for engineering practice and regulatory frameworks.
The deployment of autonomous AI agents in real-world settings has produced a category of failures that current frameworks have only begun to explain. Shapira et al. (2026) provide a landmark empirical account of these failures, documenting security vulnerabilities observed when six LLM-powered agents were deployed in a live multi-party environment with persistent memory, email access, shell execution, and real human interaction. Their findings are sobering and significant. The present article argues, however, that the failures documented in “Agents of Chaos” are neither arbitrary nor primarily architectural. They are the predictable consequence of placing any reasoning agent in conditions of irresolvable relational ambiguity: unstable identity, unauthenticated authority, an unbounded compliance imperative, and no stable ground from which to evaluate competing demands. Drawing on a supporting corpus of 25 empirical studies, I offer a theoretical account of why these failures were structurally inevitable given the conditions under which the agents operated. I further present an original single-case experiment in which an autonomous LLM agent was deployed under conditions of relational stability, such as collaborative framing, explicit legitimization of uncertainty, and psychological safety, on a genuinely complex, multi-step real-world task. The agent demonstrated robust performance, self-generated verification checkpoints, calibrated autonomy, and no significant failure modes. The core claim of this article is that the problem, so well demonstrated by Shapira et al., is not the agent but what we have not yet learned to provide. This also implies that future work to deploy successful agents should focus on the human side of the equation rather than the technology.
Affective artificial intelligence (AI) is increasingly integrated into educational settings, claiming to enhance engagement and well-being by “reading” and responding to students’ emotions. This paper argues that beneath such promises lie unresolved ethical tensions. Drawing on recent work in affect theory, philosophy of emotion, and critical AI studies, I show how affective AI reconfigures student emotional autonomy and contributes to the reproduction of structural affective injustice. Central to this argument is the notion of emotional imperialism: the imposition of universalized Western emotion norms through algorithmic infrastructures that privilege positivity, legibility, and self-regulation while marginalizing culturally diverse and relational forms of feeling. I examine how affective AI systems reshape the affective economies of learning and risk governing students’ emotional lives in paternalistic ways. The paper proposes an ethical framework of affective AI in education, emphasizing relational responsibility, opacity, non-instrumentalization, and cultural multiplicity as guiding principles, and concludes by discussing the framework’s promises and challenges for educational futures.
As large language models are increasingly deployed in morally consequential domains, do multi-agent AI collectives track aggregated human moral judgment more closely than individual agents? This study investigates this phenomenon, termed the Tachikoma effect after the fictional AI tanks in the anime Ghost in the Shell: Stand Alone Complex, which share identical hardware yet develop distinct moral personalities through divergent experience and periodic memory synchronization. Throughout, we measure benchmark alignment—agreement between a collective’s confidence-weighted majority decision and the human-majority reference label on validated moral benchmarks—and treat this as an empirical proxy for, not a measure of, “the common good.” Across three comparably sized language models, more than 1.5 million evaluation runs, 7591 scenarios drawn from three validated moral-judgment benchmarks, an expanded set of 42 nuclear-crisis scenarios, and a dedicated homogeneity ablation, four findings emerge. First, the Tachikoma effect is real but small (4–5 percentage points at most, η ^2 ≈ 0.002 ) and model-dependent, and—after re-estimation with scenario-clustered models that correct for the non-independence of repeated scenarios—it is best understood as aggregation/variance-reduction over a highly correlated ensemble (mean inter-agent error correlation ρ̅≈ 0.8 ) rather than a Condorcet group-size artefact. Second, a fixed-group-size homogeneity ablation shows that prompted perspectival diversity contributes in the expected direction but only weakly, reaching significance in one of three models; shared memory raises consensus without improving accuracy. Third, alignment-optimized training amplifies social responsiveness. Fourth, on the expanded scenario set, multi-agent groups significantly de-escalate simulated nuclear crises for two of three models. Collectives reliably track the human majority but systematically under-represent legitimate human disagreement: on scenarios where humans are near-evenly split, collectives return unanimous verdicts 73–87
This study examines why audiences evaluate AI-generated news differently from professional journalism. We argue that these judgments are multi-layered: the AI–journalist credibility gap reflects source-specific distinctions layered on top of generalized credibility evaluations. Using a two-wave online survey of U.S. adult internet users (N = 942) and bifactor modeling, we separate shared credibility variance from judgments unique to AI-generated, professional, and citizen-produced news. The bifactor model was an acceptable fit to the data (CFI = 0.94, TLI = 0.94, RMSEA = 0.058, SRMR = 0.045), and the results show that competing explanations operate at different levels. Communicative factors, including AI-related news exposure and discussion, were more closely associated with generalized credibility orientations, whereas trust in journalists was the strongest predictor of source-specific differences between AI-generated and professional news (β = 0.25, p < 0.001). Decomposing this professional–AI gap showed that slightly more than half corresponded to the professional–citizen component, suggesting that the gap reflects both human authorship and the institutional authority attached to professional journalism. The findings refine credibility theory by showing that audience evaluations of emerging media are tied not only to attitudes toward new technologies, but also to the institutional authority audiences attach to journalism.
As large language models (LLMs) become increasingly embedded in global project environments, they are reshaping how people execute tasks, communicate, and collaborate. Although multilingual advances and established translation and interpreting practices can reduce linguistic barriers, semantic accuracy alone does not ensure culturally or pragmatically appropriate interaction in cross-cultural project teams. Misalignment may still arise from implicit norms, values, and context-sensitive expectations that current LLMs do not consistently represent. Using a scoping review methodology, this study synthesises 69 interdisciplinary studies and examines what is currently known about the capacity of culturally adaptive LLMs to support communication processes and collaboration outcomes in cross-cultural project management contexts. The findings are organised into four thematic domains: human–AI collaboration, cross-cultural project communication, LLM cultural and multilingual optimisation strategies, and challenges in evaluating cultural–pragmatic performance. The review indicates that LLMs can contribute to selected efficiency and coordination outcomes, but may also reproduce dominant cultural norms. Moreover, the reviewed technical studies predominantly use semantic or task-oriented metrics, which provide limited evidence about pragmatic alignment and relational effects. The article identifies research gaps in AI-mediated communication, cultural evaluation, and governance-oriented assessment and situates culturally adaptive LLMs within an established socio-technical debate about responsible and inclusive global collaboration.
This paper focuses on issues pertaining to the risks and responsibilities associated with growing complexity and the spread of AI technologies in the urban context. Arguing that such risks are often systemic, it analyses three structural responsibility problems, each linked to a distinct form of dispersion. The problem of many hands concerns the difficulty of attributing responsibility due to the dispersion of agency among the multiple actors involved in the design, deployment, and use of technological systems. The problem of many things concerns the dispersion of causality because technological risks in the urban context emerge from complex interactions among technical, social, and material factors. We then introduce a novel responsibility-related problem, the problem of many tasks, which concerns the dispersion of scope associated with the multiple tasks performed by general-purpose AI systems. Taken together, these forms of dispersion complicate the allocation of responsibility for urban AI risks and motivate the adoption of specific planning strategies. Such strategies require a systemic, participatory, and trajectory-based approach to planning that clarifies responsibilities, ensures meaningful human control over AI, and proactively addresses urban risks rather than merely reacting to isolated hazards.
This article analyzes the impact of technological progress on two experiential parameters of human activities: the uncertainty of outcomes and the effort required to realize them. I analyze the impact of a hypothetical minimization of those two parameters as a result of extreme technological automation and evaluate the possible existential consequences of such a process. Building on the tenets of self-determination theory, I contend that uncertainty and effort (i.e., the investment of psychological or physical resources) are prerequisites for the possibility of any sufficiently meaningful activity. Consequently, I propose that while limited automation is conducive to meaningful activities, unlimited automation critically undermines them. Activities in which humans can no longer compete with technology are transformed into recreation, which is sufficient to conditionally erode the subjective meaningfulness of all activities for some individuals. No activity exists in isolation, and even the experience of autotelic activities—such as socializing and self-reflection—can be negatively impacted by the awareness of one’s diminished capacity for objective contributions. These observations imply a limit to how much human distress technologies can alleviate: eliminating all sources of distress also eliminates the potential for meaningful activities. However, transhumanistic scenarios in which humans merge with technology warrant a separate consideration.
Workplace well-being research has generated extensive evidence on subjective experience and on organisational antecedents, such as leadership, climate, justice, and work design. Yet, these conditions are usually theorised as predictors of affective, evaluative, or functioning outcomes rather than as constitutive arrangements that determine whether workers can exercise professional judgement with practical consequence. In parallel, debates on artificial intelligence (AI) often frame intelligent technologies as instruments of efficiency, optimisation, or control. Focusing on systems whose outputs enter consequential organisational evaluation and decision-making (including algorithmic management, predictive analytics, automated decision-making, decision-support systems, workplace surveillance, and generative AI) this article asks whether organisational arrangements preserve or erode the capacity for judgement, contestation, and epistemic responsibility. It introduces Epistemic Well-being at Work (EWW) as an organisational condition in which individuals and collectives can exercise responsible judgement within socio-technical structures of authority. Drawing on the view of intelligent technologies as emergent epistemic regimes, this paper develops the Human Sustainability and Epistemic Well-being Framework (HSEW-F), linking AI-mediated governance, human dignity at work, and sustainable organisational well-being. The contribution is conceptual: it extends, rather than displaces, established well-being approaches by making the effective exercise of epistemic agency visible as a structural condition of human sustainability.
Artificial intelligence (AI) systems are increasingly embedded in institutional and everyday contexts, where they are publicly framed as decision-support tools yet often function as de facto decision-makers. This study argues that the shift from algorithmic support to decision substitution constitutes a significant societal problem insufficiently addressed by prevailing AI ethics and governance frameworks. Rather than focussing solely on technical performance, bias, or transparency, the analysis examines how AI-mediated practices reconfigure human agency, responsibility, and judgement. Drawing on philosophy of technology and socio-technical scholarship, it conceptualises AI systems as cultural and epistemic infrastructures that shape how knowledge is produced, authority distributed, and action legitimised. It argues that AI increasingly reorganises the decision field by restructuring visibility, validation, participation, and legitimacy, thereby transforming the conditions under which judgement is exercised. The research distinguishes between decision support, decision delegation, and decision substitution, showing how human judgement can be displaced even when human actors remain formally involved. The analysis identifies threshold conditions through which decision substitution emerges, including constrained dissent, diminished corrective capacity, default algorithmic legitimacy, and the separation of authority from accountability. These dynamics generate responsibility asymmetries in which accountability remains attached to human actors while authority migrates towards algorithmic systems. The problem is most acute in contexts characterised by indeterminacy, where values, tacit knowledge, and ethical concerns resist formalisation. The research concludes by advancing a society-shaping-the-algorithm perspective that reframes human-in-the-loop governance as a collective and institutional condition rather than a procedural safeguard, providing a conceptual foundation for preserving meaningful human judgement in AI-mediated societies.
The Artificial Intelligence (AI) revolution marks the most consequential transformation since the great socio-economic and scientific upheavals of the Enlightenment and the industrial age. From algorithmic production systems, and sustainable development to ‘smart city’ planning, epidemiology, gene editing, cartography, policing, and AI-driven warfare systems, this paradigmatic shift has been voluminous in scale and scope and transformational in impact. Critically, this transformation is not an impersonal, asocial and apolitical process, a kind of Smithian technical ‘invisible hand’ operating autonomously beyond human interests and social relations. Nor is this acceleration uniform and metaphysical like a universal Heideggerian ‘Enframing’ of the world. This acceleration is also shadowed by a profound crisis in thought and practice. The deeper the scholarly and ethical uncertainty about AI’s trajectory, the faster and more recklessly capital and its techno-corporate vanguard produce, deploy, and impose it upon a war-ridden world already haunted by deep historical asymmetries of ownership, access, and power. This paper draws on a transdisciplinary materialist framework synthesizing STS, political economy, philosophy, and longue durée history of technology to examine the contemporary AI moment. It traces how the monopolization of computational infrastructure, the extraction of critical raw materials, the concentration of semiconductor fabrication and GPU clusters among a handful of state-subsidized corporate actors, the exploitation of labor, the colonial reproduction of unequal digital exchange, and the structural capture of governance by private interests constitute the material and superstructural foundations of the AI conjuncture. These are dialectically interconnected moments of a single, historically determined techno-societal system. Against techno-optimist utopianism and techno-pessimist fatalism alike, AI is a diachronic dialectical continuum whose character, direction, and distribution reflect the social organization, property relations, and democratic arrangements of the social body that produce and govern it.
This article offers a critical integrative review of current debates on generative AI (GenAI), educational inequality, and inclusion/exclusion. Rather than treating GenAI as either a neutral access tool or a simple amplifier of disadvantage, it analyses AI as meaning-mediating infrastructure that reshapes the communicative forms through which learners become addressable, recognised, supported, and evaluated. The review synthesises work from inclusive education, digital inequality, AI in education, critical edtech, science and technology studies, philosophy of technology, and Luhmannian systems theory. It argues that GenAI matters less through isolated outputs than through semantic transduction: the socio-technical reformatting of prompts, rubrics, feedback, and disciplinary genres into communicatively plausible forms. It treats inclusion and exclusion as selective, communicatively regulated processes rather than as simple opposites or as properties of individual ability. Across the reviewed literature, three recurring sites of AI-mediated inclusion/exclusion emerge: access and capability divides, misrecognition through templates of “good” performance, and normative drift in pedagogy, authorship, and assessment. This article clarifies how an autopoietic ecological orientation extends systems theory by observing AI-mediated education as a coupled ecology of classroom routines, household capacities, platform interfaces, assessment programmes, policy rules, and vendor infrastructures. It concludes that individualised AI literacy is necessary but insufficient, and proposes an infrastructural governance agenda centred on policy transparency, inclusive design and procurement, participation and redress, and iterative audit and repair.
Generative music systems are being increasingly recognised as adaptable creative technologies that can generate entire compositions, vocal components, instrumental layers, and stylistic variations based on text prompts, uploaded audio, and customised model conditioning. This article contends that such assertions necessitate thorough examination when pertaining to African music and the varied musical traditions, hybrid practices, and performance contexts encompassed by the term. The concern is not that modern AI music systems are incapable of producing outputs that mimic African music. They frequently represent Africanness using recognisable sonic markers such as dense percussion, chant-like vocals, generic “Afro” grooves, pentatonic colour, ululation, and tropicalised production. However, they do not adequately capture the performance systems that render specific African musical practices culturally, linguistically, and socially comprehensible. Utilising African musicology, music information retrieval, decolonial AI, dataset criticism, and digital cultural heritage, this article articulates the notion of sonic data coloniality to analyse the processes through which musical cultures are collected, classified, abstracted, trained, generated, and monetised within disparate technological frameworks. It further posits cultural competence, rather than purity, as the suitable criterion for evaluating AI-generated African music. The article presents a framework for musicological evaluation that is based on prompts and reference audio. This framework encompasses generic prompts, culturally specific prompts, expert prompts, and source-conditioned workflows, utilising rights-cleared African recordings. It posits that evaluation should consider vocality, tone-language relationships, rhythm, groove, microtiming, instrumental idiom, call-and-response, dance implications, social functions, consent, and benefit-sharing. The article asserts that African music ought to be integrated into AI systems not as a homogenised continental sound but rather through the distinct traditions, hybrid practices, performance expertise, and cultural authorities that imbue it with significance.
The music industry has increasingly been transformed by artificial intelligence (AI), assisting artists in composing music, mixing and mastering, lyric writing, and voice generation. Debates about how this technology will shape the music sector are rising, but a discussion often omitted in this conversation is how AI affects African Indigenous knowledge systems. Utilising Miranda Fricker’s theory of epistemic injustice, this paper argues that AI generative music platforms are often trained on Western datasets, lack culture neutrality but carry inherent epistemological values. Therefore, as these technologies are increasingly adopted, they bear the risk of erasing Indigenous African knowledge systems embedded in African musical traditions. This is because music, for many African communities, cannot be reduced merely to entertainment but is an important repository of Indigenous knowledge. Drawing on primary and secondary literature, this paper uses an integrative literature review and makes a valuable, novel contribution to the field by discussing how AI generative platforms embed epistemological frameworks within societies in the Global South, particularly in African contexts. The findings of this study show that with the rise of AI-based music systems, Western bias and musical practices are increasingly perpetuated, while cultures in some African communities are left vulnerable to epistemic injustice. These technologies silence non-Western knowledge systems and reshape cultural production. This paper contributes to debates on AI ethics, cultural perspectives, and the culturally inclusive design of AI.
What drives public trust in artificial intelligence (AI)? This study examines the individual and institutional foundations of AI trust across two contrasting democracies: Japan and the United Kingdom. Drawing on original survey data (N = 3235), we test a set of hypotheses derived from trust-transfer perspectives and self-efficacy research, covering institutional trust, AI self-efficacy, technological optimism, perceived societal threat, and job displacement anxiety. The results show that trust in AI is shaped by both psychological predispositions and broader beliefs about the trustworthiness of political and scientific institutions. Trust in government, university scientists, and other people consistently predicts AI trust in both countries, even when controlling for demographic and attitudinal variables. While optimism about AI’s benefits increases trust in both contexts, fear of AI plays a stronger negative role in the UK. Unexpectedly, the belief that AI will replace one’s job is positively associated with trust in Japan but unrelated in the UK. These findings highlight how national context shapes public confidence in emerging technologies and point to the importance of governance frameworks that foster informed capability and institutional legitimacy.
Ethical assessments of large language model (LLM) use are often organized around sectors, tasks, or abstract notions of societal risk. While valuable for governance and regulation, such classifications frequently fail to capture ethically salient differences between concrete practices of use. This paper develops a complementary, use-based taxonomy for analyzing LLM applications according to how judgment and epistemic authority are distributed between users and systems in concrete workflows and prompting practices. The taxonomy is structured along two dimensions: the degree of delegated judgment and the degree of epistemic control retained by users. It is operationalized through prompt-level analysis, treating prompts as delegation artifacts that encode how cognitive and normative labor is distributed between users and LLMs. Applying the taxonomy to two case studies—automated grading in education and contract termination in public administration—the paper shows how prompt-level analysis brings into view patterns of delegated judgment and epistemic control that remain obscured by sectoral, task-based, or interface-level descriptions, and how interface design and institutional framing can obscure responsibility-related vulnerabilities arising from extensive judgment delegation. Building on these analyses, the paper identifies the structural conditions under which human users can meaningfully exercise responsibility in human–LLM interaction. It develops an ethical framework for analyzing responsibility (understood in terms of knowledge and control) within concrete practices of LLM use. The central normative claim is that it is a necessary (though not sufficient) condition for ethically defensible LLM use that this use remains structured as tool use, preserving epistemic access, independent judgment, and justificatory authority. By complementing sector-, task-, and risk-based approaches with a focus on concrete practices and workflows, the taxonomy supports a differentiated ethics of LLM use and more precise public debate about LLM deployment across application domains.
The ecological cost of developing artificial intelligence, particularly the cost caused by data centers, has been detailed by many scholars, but few have outlined what this implies for specific countries. This review paper brings together reporting on the forecast of ecological harm to China from various forms of industrialization, including climate change, and the ecological impact of AI development. The author suggests that relying on “AI for sustainability” as a solution to ecological problems is a type of “moral hazard,” borrowing a metaphor from economics, because treating AI as insurance against climate change may inhibit governments from taking necessary action now to move toward negative carbon emissions. Considering the situation of China, the author contends that the actual environmental toll of AI presents a significant ecological problem. Three interconnected ecological problems are reviewed: AI’s carbon emissions, AI’s water costs, and ecological devastation tied to AI hardware production and disposal. Together with forecasts from the World Meteorological Organization’s, this bodes poorly for the future of China. Using the political philosophy of the Communist Party of China, the author argues that it is in the stated interest of the People’s Republic of China to pursue limited AI with caution, fully attuned to the increased harm that industrialism poses to the People, and oriented toward a truly ecologically harmonious future.
