Aleksandra Przegalinska

As AI systems become embedded in workplace workflows, their interactional style-the persona they present-may function as an ambient condition shaping cognition, affect, and performance. We tested this in a controlled experiment (N = 58) where participants completed a marketing collaboration task supervised by one of two AI chatbot personas grounded in leadership theory: a supportive Servant Leader and an antagonistic Dark Triad leader. Combining self-report questionnaires, continuous psychophysiological recording (electrodermal activity and facial electromyography), LLMassisted behavioral coding, and blind expert-rated performance, we examined whether persona reliably shifts outcomes across measurement modalities. Results revealed a striking multi-level pattern. The Dark Triad persona produced significantly elevated sympathetic arousal (72% higher peak electrodermal amplitude), a tenfold increase in behavioral resistance, substantially greater strategy diversity and switching, fourfold more jailbreaking attempts with qualitatively more aggressive strategies, and significantly lower expert-rated work quality (d = 0.71-0.83). Yet retrospective self-reports of enjoyment, satisfaction, and focused attention showed no significant between-group differences. Equivalence testing clarified this dissociation as graded: self-report effects were bounded below d = 0.75 but could not be established as equivalent at d = 0.50. Of fourteen hypotheses, eight were supported and two partially supported, with confirmation tracking measurement modality: behavioral, autonomic, and performance hypotheses were overwhelmingly confirmed; self-report and somatic physiological hypotheses were not. These findings reframe AI persona as a causal design dimension with consequences for well-being, performance, and human-AI collaboration, warning that organizations relying solely on satisfaction surveys risk underestimating the costs of poorly calibrated interactional design.

Technological change, particularly automation and artificial intelligence (AI), is reshaping labour demand and influencing migration dynamics. This article introduces the Automation-Migration Interface Approach as a conceptually driven mechanism synthesis that updates labour-market segmentation theory for the age of automation and AI. Focusing on the European Union, with comparative references to the United States, China, and OECD or global evidence, we argue that automation tends to intensify labour market polarisation by reducing routine middle-skill jobs while expanding demand for high-skill work and sustaining parts of the lowskill service economy. Although some low-skilled roles remain difficult to automate, others face displacement or degradation through reorganised workflows and digital management, raising risks for migrants who are often concentrated in secondary labour-market segments. We conduct a systematic review of academic and grey literature using an LLM-assisted screening pipeline with human validation. The synthesis traces a pathway from automation and AI adoption to task restructuring and occupational shifts, then to migrant sorting and matching, and finally to barriers and mediators that shape outcomes such as wages, employment stability, mobility, segregation, and skill utilisation. The reviewed evidence suggests that skilled migrants can benefit from growing demand in AI-related and green-transition sectors, while many lowskilled migrants remain concentrated in precarious work with limited upward mobility. Key mediators include constrained access to training, difficulties in credential recognition, and gatekeeping risks linked to algorithmic screening and workplace management tools. The article concludes with policy implications for inclusive labour-market governance, including targeted upskilling and transition supports, safeguards against algorithmic bias in hiring and employment services, improved mobility and matching infrastructure, and streamlined recognition of migrant skills. It also sets out a migration-studies research agenda with testable

This paper provides a meta-analysis of macroeconomic, quantitative studies that seek to identify the causal impact of automation and artificial intelligence (AI) on employment in the European Union (EU). We contribute an EU-focused synthesis that concentrates on macro-level evidence, complementing a literature dominated by micro task-based analyses and broader cross-country discussions. Using an AI-assisted workflow for search, screening, and extraction with human oversight, we map the available evidence and report clear methodological and knowledge gaps. A central finding is that the automation and robotics literature is comparatively developed, while credible causal macro evidence on AI itself remains sparse and fragmented. We treat this scarcity as a result that motivates a concrete research agenda. Across the core empirical studies and selected grey literature, the net employment effect appears modestly positive on the order of 1.5 to 2.0 percent, but the distributional pattern is uneven. Risks concentrate among low-skilled and routine workers, older cohorts, and lagging regions, pointing to policy priorities in reskilling, education, and digital infrastructure. Methodologically, the paper demonstrates how LLM-assisted procedures can improve transparency, consistency, and scalability in early-stage evidence synthesis where conventional meta-regression is premature.

How a cultural focus can empower generative artificial intelligence.

This paper introduces Generative Artificial Experts (GAEs) - a concept of a new type of generative AI agents designed for human-AI collaboration in knowledge work. GAEs have specialized domain expertise, perform tasks within bounded autonomy, include a synthetic persona and possess multimodal generative AI capabilities, among other features. We provide a definition of GAEs which includes seven defining traits, offering a taxonomy which sets them apart from other generative AI systems. We use literature-review based conceptual analysis with abductive reasoning to propose the new concept that addresses identified limitations in existing systems. The paper explores the emergence of GAEs as a leap from expert systems. We name two enablers for GAEs - ongoing development of a research field of human-AI collaboration and growing capabilities of generative artificial intelligence systems. We discuss existing generative AI agents, noting that GAEs as such do not exist yet, but are starting to emerge. Due conceptual nature of this paper we do not explore the technical aspects of GAEs development. Instead, we use illustrative examples to present possible applications of GAEs and their potential role in the future of knowledge work. This article appears in the AI & Society track.

Purpose: This pilot study aimed to evaluate the impact of the big five personality traits on user engagement with chatbots at the early stages of artificial intelligence (AI) adoption. Design/methodology/approach: The pilot study involved 62 participants segmented into two groups to measure variables including engagement duration, task performance and future AI usage intentions. Findings: The findings advocate for the incorporation of psychological principles into technology design to facilitate more tailored and efficient human–AI collaboration. Originality/value: This pilot research study highlights the relationship between the big five personality traits and chatbot usage and provides valuable insights for customizing chatbot development to align with specific user characteristics. This will serve to enhance both user satisfaction and task productivity.

This article delves into the utilization of GPT-4, an expansive generative pre-trained language model, to cultivate adept professionals in the realm of business. It explores the potential advantages and obstacles associated with leveraging GPT-4 to enhance both hard and soft skills, while also providing insights on optimizing this technology for professional development in business education.

Positive and normative claims that artificial intelligence (AI) will or should lead to adoption of a universal basic income policy (UBI) remain insufficiently empirically grounded to merit serious consideration. Long-term trends in individual/familial income portfolio adjustment (IPA) to business, economic, and technological change (BETC) point to continued incremental changes in the ways that individuals/families achieve life goals, not a fundamental structural break necessitating radical policy changes that may not be desirable in any event. Moreover, if AI proves a more rapid disruptor than anticipated, UBI-like payments can be made quickly, as recent bailouts and fiscal stimuli demonstrate.