Agile culture, collective human intelligence and AI adoption: Gender and sectoral perspectives on augmented intelligence in Europe

Critical Agency and Hybrid Cognition in digital art education: Human–AI co-creation with stable diffusion XL

An experimental study exploring human–AI complementarity in early social-emotional learning

The cognitive architecture of symbolic identity: Structuring coherence in human–AI reasoning systems

Like the arrival of calculators in 1970s classrooms, large language models (LLMs) provoke both fears of intellectual deskilling and hopes of more efficient learning. In this paper we analyze the calculator analogy, arguing that while it is a useful starting point to understand the potential impact of LLMs in education, it is ultimately insufficient. We show where the analogy holds and, just as importantly, where its limitations reveal the unique pedagogical challenges posed by LLMs. These challenges arise from fundamental differences in how calculators and LLMs mediate learning, reflecting the distinct affordances of each technology. We argue that because of their affordances, realizing the educational potential of LLMs calls for cultivating epistemic virtues suited to human–AI interaction, such as patience, reflective engagement, or intellectual vigilance and humility. Equally, LLM design must actively foster these virtues through features like built-in prompts, feedback loops or reflective questions, to name just a few. • The calculator analogy illuminates but ultimately only partially captures LLMs' unique educational impact. • Technological affordances explain how tools shape, not just support, learning. • LLMs redistribute cognitive labour across all stages of reasoning, from framing to interpretation. • Persuasive fluency makes LLM outputs appear true, amplifying epistemic risks. • Effective use of LLMs demands cultivating epistemic virtues like patience and vigilance.

A study on the modality of climate news discourse: The case of the New York Times

This innovative study investigates how Emotion Recognition Artificial Intelligence (ERAI) influences user trust in Sharing Economy (SE) platforms – an emerging and underexplored frontier at the intersection of affective computing, digital trust, and platform governance. Despite the growing adoption of AI, little is known about how ERAI influences perceived trust in SE platforms, and what is the role of emotional connection in this process. Addressing this gap, we conducted a between-subjects experimental study with 320 participants, comparing trust perceptions in ERAI-enabled versus conventional SE platforms. Anchored in trust theory, the Technology Acceptance Model (TAM), and social exchange theory, findings reveal that ERAI significantly increases emotional connection and perceived trustworthiness. A strong positive correlation was observed between emotional engagement and trust. Our results offer critical insights for human–AI interaction, ethics in AI deployment, and responsible platform design. This paper contributes to science, industry, and society by highlighting ERAI's potential to humanise AI-driven platforms and enhance digital trust in the evolving landscape of intelligent technologies. • Fills a research gap on how Emotion AI affects trust in the sharing economy. • Introduces novel experiments using facial emotion recognition AI (ERAI). • ERAI increases trust when perceived as emotionally accurate and human-like. • Trust is shaped not only by outcomes but by AI's perceived ethical behavior. • Provides actionable insights for designing trustworthy AI-driven platforms.

Human–AI partner teams are positioned to transform mathematical creativity, shifting discovery from incremental, bot-tom-up reasoning to a broader mode of inquiry that spans the full landscape of mathematics and science. This paper examines that transition by advancing Galois Smartnetwork Field Theory (Galois SNFT) as a framework for co-evolutionary human–machine reasoning—one that integrates mathematics, computation, and physics through the organizing power of higher structures mathematics, especially symmetry. To accelerate the inclusion of mathematical research into the computational infrastructure, Galois SNFT extends Neural Network Field Theory (NNFT) approaches by adding mathematics as a cornerstone to physics and computation. Digging deep into Modern Symmetry Theory’s convergence toward a Grand Unified Symmetry (GUS) framework with frontier mathematics from Clausen, Scholze, Lurie, Bhatt, Pridham, Barwick, and Haine, Galois SNFT deploys three symmetry properties (phase stability, glocal propagation, and symmetry constraint) to analyze the Millennium Prize Problems (MPP). The MPP can be partitioned into Langlands, physics, and orthogonal arms. Within this landscape, the Riemann Hypothesis (regarding the distribution of prime numbers along a critical line) is particularly suited to a symmetry-based analysis via phase stability, making it a compel-ling test case for co-evolutionary human–AI mathematical discovery.

ABSTRACT This paper investigates the pedagogical methods and models for conducting creative writing workshops assisted by generative artificial intelligence (GenAI) in Chinese universities. The experiment was conducted at Guangdong University of Foreign Studies (GDUFS) with 15 participants majoring in Chinese Language and Literature. During the workshop, all participants were encouraged to use GenAI tools to collaboratively create fictional works. The workshop consisted of three sessions: (a) an instructional session, in which the teacher introduced human–AI interaction techniques and provided guidance on constructing narrative conflicts; (b) a writing session, where students engaged in collaborative writing with GenAI tools; and (c) a post-writing session involving peer interaction and group discussion. A comparison between students’ initial and final manuscripts demonstrated a notable improvement in their human–AI interaction skills. However, most participants still exhibited limited ability to develop conflict-driven plots. Therefore, we suggest that human–AI collaborative writing workshops should adopt a stepwise instructional model: students should first develop competence in human–AI interaction before being guided in knowledge of creative writing. Moreover, the questionnaire results indicate that two-thirds of the students found the interaction and discussion session most beneficial, suggesting that in-class discussion remains a highly effective component of human–AI collaborative writing pedagogy.

The copyrightability of mixed musical works, which contain a blend of human and generative AI elements, is an issue of increasing prevalence in copyright law. While there has been some discussion on the copyright status of fully generative AI works, this mainly resides in state law, and much of the federal policy found in Copyright Guides published by the United States Copyright Office is a non-binding opinion. Additionally, the same circuit courts contradict themselves, as seen with the differing views on the fair use of generative AI works in the Ninth Circuit Cases of Bartz v. Anthropic PBC (2024) and Kadrey v. Meta (2023). These issues combine to create a copyright “Dead Man’s Land” where the U.S. Copyright Office is forced to inspect mixed musical works on a case-by-case basis: an inefficient and ineffective mess for the modern day. This paper proposes that Congress pass legislation further itemizing the components of mixed musical works and only allowing generative AI to be used in one such component, alongside integrating state legislation. This policy would allow the U.S. Copyright Office to evaluate all mixed musical works under a unified framework. With the increase of artists creating mixed musical works, this solution only becomes more necessary as time progresses.

Objective This study compares the psychometric properties of scale structures developed by human experts and various generative artificial intelligence (GenAI) models to measure motivation for using GenAI in education. Method Grounded in a theoretical framework, six different item pools were generated by human experts and GenAI prompt bots based on five different language models. Based on the reviews conducted by human experts and GenAI bots, the three forms demonstrating the strongest psychometric performance were selected for the final analysis. Result The confirmatory factor analysis indicated that all three scale structures demonstrated acceptable and good model fit, and high internal consistency and convergent validity. The high correlations between the structures revealed that both human- and AI-generated factors measure similar structures. Conclusions GenAI models can generate psychometrically sound scales; however, human expertise remains essential for ensuring theoretical depth and cultural contextualization. Therefore, hybrid human–AI approaches appear to offer the most robust outcomes.

Generative artificial intelligence (GenAI) is increasingly integrated into organisational meetings and workshops. Promoted for their potential to enhance efficiency through automated summarisation, clustering, mediation, and simulation, these technologies are reshaping the socio-technical conditions that underpin Soft OR and Problem Structuring Methods. This study develops a Soft OR-grounded GenAI literacy orientation that prioritises stewardship over automation in facilitative practice. Building on established Soft OR scholarship, six core commitments are articulated and translated into six sensitising literacy questions for GenAI-mediated facilitation. Publicly available documentation for twelve GenAI applications designed for meetings, workshops, and deliberative work is analysed. The findings illustrate that these systems may standardise problem framings, shift interpretive authority from participants to modelling applications, present algorithmic outputs as closure, and prioritise convergence at the expense of maintaining openness to plural worldviews. The resulting risk is conceptualised as Synthetic Closure, defined as the premature stabilisation of meaning through algorithmically generated artefacts that simulate coherence or agreement without replicating the participatory work necessary for legitimacy in Soft OR. The study advocates for Smart OR stewardship as a facilitative stance that upholds Soft OR’s foundational commitments within hybrid human–AI environments.

As generative AI becomes increasingly embedded in everyday life, users have begun to form emotional connections with ChatGPT. Drawing on domestication theory and research on algorithmic resistance, this study combines in-depth interviews with 20 users (including follow-up interviews) and online observation to examine how users, through ongoing interaction, expand the utility boundaries of generative AI and gradually construct it as a relational object with intimate significance. The findings suggest that this process of creative domestication is primarily manifested through three interrelated mechanisms: collaborative appropriation, reparative resistance, and the anthropomorphization of the relational subject. The study sheds light on the formation of human–AI intimacy while also highlighting the associated risks of emotional dependence, technological vulnerability, and privacy exposure.

Abstract While AI ethics ensures fairness, accountability, and protection of user rights, dark patterns manipulate users to take unintended actions on digital interfaces. Related studies uncover limited insights into how reliably; human experts and AI models can detect dark patterns within a specific taxonomy. Our research fills this gap by asymmetrically examining cross-origin detection performance of human and AI/LLM evaluators (each evaluator’s ability to detect dark patterns generated by the opposite source) to understand their limitations and future potentials. Using GPT-4.1, we generated 200 UI images (with matched dark and non-dark pattern pairs) and selected 200 UI images collected 200 human-created UI screenshots from the ContextDP/AidUI dataset, based on computational, methodological, and statistical considerations. We calculated inter-rater reliability, recall, and error distribution. The results show that UX experts achieved substantial agreement (k = 0.75) and significantly higher recall (r = 0.99) over AI/LLMs. We present a novel study which explore the performance of AI/LLMs and UX experts in detecting dark patterns in UI images, and provide a benchmark dataset that could be useful to future research, while discussing empirical insights into the role, limitations, and promise of AI/LLMs in UI/UX design ethics and auditing, in realistic deployment scenarios.

Q-matrix construction is a foundational yet challenging step in cognitive diagnostic assessment (CDA), which is traditionally reliant on labor-intensive and subjective methods like expert judgment and verbal report analysis. This study explores the potential of generative artificial intelligence (GenAI) to optimize this critical process within the domain of EFL reading. By applying three GenAI models (DeepSeek-V3.2, Kimi 2.5, and Doubao 2.0), three purely GenAI-informed Q-matrices (Qmat-DS, Qmat-K, and Qmat-DB) were generated, and through expert revision, a human–AI collaborative Q-matrix (Qmat-DS-H) was obtained. These were compared with an expert-constructed Q-matrix (Qmat-E) and a student-derived Q-matrix (Qmat-S). Using a simulated dataset (N = 1000) and empirical response data from 1083 EFL learners on a diagnostic reading test, the psychometric performance of the six Q-matrices was estimated via the G-DINA model, ACDM model, and RRUM model. Results demonstrated that the human–AI collaborative Q-matrix consistently outperformed the other five Q-matrices, achieving the best absolute model-data fit, the highest classification accuracy, the most stable item parameters, and the most balanced attribute correlation structure. The purely GenAI-informed Q-matrices showed mixed results: there were some improvements in relative fit and slip stability compared to manually constructed Q-matrices, but variable absolute fit and attribute correlation patterns. The findings substantiate GenAI as a feasible pathway for enhancing the efficiency, consistency, and psychometric quality of Q-matrix construction. This study offers a preliminary framework for advancing CDA development, addressing a key methodological bottleneck in language assessment.

Purpose This perspective revisits customer satisfaction in hospitality and tourism by connecting three lenses that are often treated separately: ethical non-projection in host–guest treatment, expectancy–disconfirmation as the core evaluative logic of satisfaction and AI-enabled systems that increasingly shape, infer and respond to guest experiences. Design/methodology/approach The paper uses a conceptual synthesis supported by a narrative review of hospitality scholarship. It interprets the Platinum Rule as a normative orientation of expectation-attentiveness and uses it alongside satisfaction and technology literature to organize a past–present–future discussion. Findings The synthesis suggests that the three lenses are complementary but tension-filled: ethical non-projection helps explain how expectations should be approached, expectancy–disconfirmation explains how experiences are judged and AI infrastructures alter both expectation formation and satisfaction management. Three recurring paradoxes emerge: personalization and efficiency versus authenticity and warmth, measurement precision versus privacy and guest agency and real-time intervention versus transparency and fairness. Practical implications Managers should match human–AI configurations to emotional complexity, treat personalization and sensing as governance choices rather than purely technical features and design recovery systems that preserve explanation, choice and fairness across settings. Originality/value Rather than presenting a fully elaborated theory, the paper offers a more integrated conceptual lens for discussing how ethical attentiveness, evaluative judgment and AI-enabled service infrastructures jointly shape satisfaction in hospitality.

From assistants to agents: a relational framework for human–AI co-agency

Bonding with the machine: The empathy-accountability gap in human interactions with LLM-powered artificial therapists

Diverse AI personas can mitigate the homogenization effect in human-AI collaborative ideation

Planning with ComPlanAI: Comparative insights from a human–AI evaluation of comprehensive plans