|
on Artificial Intelligence |
| By: | Alexander Erlei; Lukas Meub |
| Abstract: | As AI agents increasingly act on behalf of human stakeholders in economic settings, understanding their behavior in complex market environments becomes critical. This article examines how Large Language Models coordinate on markets that are characterized by information asymmetries and in which providers of services have incentives to exploit that asymmetry for their own economic gain. To that end, we conduct simulations with GPT-5.1 agents in credence goods markets, manipulating the institutional framework (free market, verifiability, liability), LLM agent's social preferences (default, self-interested, inequity-averse, efficiency-loving), and reputation mechanisms across one-shot and repeated 16-round interactions. In one-shot settings, LLM agents largely fail to establish cooperation, with markets breaking down except under liability rules or when experts have efficiency-loving preferences. Repeated interactions solve consumer participation through competitive price reduction, but expert fraud remains entrenched absent explicit other-regarding preferences. LLM consumers focus narrowly on price levels rather than understanding strategic incentives embedded in markups, making them vulnerable to exploitation. Compared to human experiments, LLM markets exhibit substantially higher consumer participation but much greater market concentration, lower prices, and more polarized fraud patterns. The effect of institutions like verifiability and reputation is also much more ambiguous. Surplus shifts dramatically toward consumers under social-preference objectives. These findings suggest that institutional design for AI agent markets requires fundamentally different approaches than those effective for human actors, with social preference alignment emerging as the primary determinant of market efficiency. |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2603.08853 |
| By: | Koch, Alexander (Aarhus University); Kragl, Jenny (EBS Business School); Ming, Sijuan (Aarhus University); Nafziger, Julia (Aarhus University) |
| Abstract: | Persistent gender gaps in self-promotion contribute to unequal labor market outcomes. In this study, we investigate how AI-assisted writing tools shape self-promotion, and, as a secondary outcome, confidence and how these effects interact with gender. For this purpose, we conducted an online experiment in China in which participants wrote self-promotion texts, provided a numerical self-promotion score and stated their confidence about how they will perform in an upcoming math and logic test. We find suggestive evidence that AI assistance reduces numerical self-evaluations. Neither gender nor the interaction between gender and AI assistance is significantly related to self-promotion or confidence. We conduct a text analysis to investigate the mechanisms behind these results. |
| Keywords: | self-promotion, confidence, AI assistance, gender gaps |
| JEL: | C90 D03 D83 J16 M12 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18441 |
| By: | Mathieu Chevrier (Université Côte d'Azur, CNRS, GREDEG, France); Sébastien Massoni (Université de Lorraine, Université de Strasbourg, CNRS, BETA, Nancy, France) |
| Abstract: | Confidence often accompanies advice, but its usefulness depends on what confidence actually reveals. This paper distinguishes between two dimensions of confidence quality: discrimination, that is, whether confidence tracks correctness at the decision level, and calibration, that is, whether average confidence matches average accuracy. In a controlled advice-taking experiment comparing human and algorithmic advisors, discrimination is the main driver of both advice adoption and post-advice accuracy, whereas calibration plays a more limited role. Source matters only in a specific case: when discrimination is high, participants are more likely to follow overconfident algorithmic advice than equally overconfident human advice. Advice taking also varies with participants’ own metacognitive characteristics. Higher discrimination ability is associated with more conservative advice taking, while better-calibrated participants rely more on stated confidence, benefiting when advisor confidence has high discrimination and performing worse when it is miscalibrated. |
| Keywords: | Algorithm; Advice; Overconfidence; Discrimination; Laboratory experiment |
| JEL: | C92 D91 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:gre:wpaper:2026-09 |
| By: | Ufuk Akcigit; Craig A. Chikis; Emin Dinlersoz; Nathan Goldschlag |
| Abstract: | We construct a novel dataset linking academic publication records to U.S. Census employer–employee data to track 42, 000 AI researchers over two decades. We document systematic changes in the allocation of AI talent. Industry increasingly attracts younger and foreign-born researchers, while gender representation improves more in academia. The top 1% of publishing industry scientists now earn $1.5 million more annually than comparable academics, a fivefold increase since 2001. Rising wage premia coincide with greater sorting into large incumbent firms. Researchers who move to industry publish less but patent more, consistent with a shift from open science toward proprietary innovation. |
| JEL: | I23 J45 L33 O31 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:34964 |
| By: | Huan Cai; Ziqing Lu; Catherine Xu; Weiyu Xu; Jie Zheng |
| Abstract: | With recent development of artificial intelligence, it is more common to adopt AI agents in economic activities. This paper explores the economic actions of agents, including human agents and AI agents, in an economic game of trading products/services, and the equilibria in this economy involving multiple agents. We derive a range of equilibrium results and their corresponding conditions using a Markov chain stationary distribution based model. One distinct feature of our model is that we consider the long-term utility generated by economic activities instead of their short-term benefits. For the model consisting of two agents, we fully characterize all the possible economic equilibria and conditions. Interestingly, we show that unless each agent can at least double (not merely increase) its marginal utility by purchasing the other agent's products/services, purchasing the other agent's products/services will not happen in any economic equilibrium. We further extend our results to three and more agents, where we characterize more economic equilibria. We find that in some equilibria, the ``more powerful'' AI agents contribute zero utility to ``less capable'' agents. |
| Date: | 2026–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2603.00858 |
| By: | Lodefalk, Magnus (Örebro University School of Business); Löthman, Lydia (Örebro University School of Business); Koch, Michael (Aarhus University); Engberg, Erik (Örebro University School of Business) |
| Abstract: | We show that the age composition of employment within Swedish employers shifts after the arrival of generative AI, with no corresponding reduction in aggregate labour demand. Using 4.6 million job advertisements from Sweden’s largest recruitment platform, we find that the broad decline in postings since 2022 aligns with monetary tightening rather than AI, exploiting Sweden’s seven-month gap between the Riksbank’s first rate hike and the launch of ChatGPT as a timing test. We then use full-population employer– employee register data and an employer-level difference-in-differences design to estimate how AI exposure affects employment composition across six age groups. An event study documents an accelerating decline in employment of 22–25-year-olds in high-AI-exposure occupations, reaching 5.5 per cent by early 2025 relative to less exposed occupations within the same employers, while employment of workers over 50 rose by 1.3 per cent. The widening age gradient suggests that generative AI reshapes hiring composition rather than aggregate demand, with the adjustment burden falling disproportionately on entry-level workers. |
| Keywords: | Generative artificial intelligence; Job postings; Labour demand; Employment composition; Monetary policy |
| JEL: | J23 J24 O33 |
| Date: | 2026–03–16 |
| URL: | https://d.repec.org/n?u=RePEc:hhs:oruesi:2026_002 |
| By: | Gschwendt, Christian (University of Bern); Viarengo, Martina (Graduate Institute of International and Development Studies, Geneva); Zollner, Thea S. (University of Bern) |
| Abstract: | The economic impact of technological change will critically depend on how future workers invest in their human capital. Yet, little is known about how future workers themselves evaluate and choose their educational and occupational paths in light of emerging technologies. This paper examines how adolescents currently at the school-to-work transition stage value working with generative artificial intelligence (GenAI) in their future occupations, and how automation risk and opportunities for continuing education shape these preferences. We field a discrete-choice experiment among a nationally representative sample of over 7, 000 Swiss adolescents aged around 15. We find that adolescents generally exhibit an aversion to collaborating with GenAI at work, with females consistently more averse than males. However, preferences are nuanced: adolescents welcome greater GenAI collaboration, provided that GenAI usage levels remain moderate and that it is not accompanied by increases in job automation risk. Finally, our findings suggest that AI-related educational opportunities in occupations improve attitudes towards working with GenAI across genders. |
| Keywords: | occupational choice, gender gaps, GenAI, choice experiment, continuing education, automation risk |
| JEL: | I24 J24 O33 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18456 |
| By: | Etheridge, Ben |
| Abstract: | This paper investigates Artiï¬ cial Intelligence (AI) adoption and its labour market conse-quences among UK small and medium enterprises, using novel data from the British Cham-bers of Commerce Business Outlook Survey, collected in early 2026. AI adoption is increas-ingly widespread, with over half of responding ï¬ rms currently using AI, up from around a third in 2025. Most users rely on generic tools such as ChatGPT or Copilot, but around one in ten ï¬ rms have adopted bespoke AI implementations. We ï¬ nd that bespoke adoption in par-ticular is associated with a coherent bundle of workforce adjustment. Approximately one-ï¬ fth of bespoke users report stafï¬ ng reductions attributable to AI, and bespoke adopters are roughly three times more likely to have restructured job roles. Restructuring is in turn strongly associated with headcount reductions and shifts in skills requirements. Surprisingly, ï¬ rms investing in AI-related training are signiï¬ cantly more likely to anticipate headcount reductions than those not investing in training. We also ï¬ nd that current AI users are substan-tially more optimistic about future productivity gains than non-users. Our ï¬ ndings provide a novel ï¬ rm-level picture of how SMEs are reorganising work, adjusting workforces, and in-vesting in skills in response to AI. |
| Date: | 2026–03–18 |
| URL: | https://d.repec.org/n?u=RePEc:ese:iserwp:2026-01 |
| By: | Lünich, Marco; Keller, Birte; Marcinkowski, Frank |
| Abstract: | This brief report presents the key findings of a segmentation study conducted in June 2025 as part of the Opinion Monitor Artificial Intelligence 3.0 (MeMo:KI 3.0) project. It analyzes the frequency of artificial intelligence (AI) use among 1, 987 working people in Germany and its association with occupational classes, sociodemographic characteristics, and work-related attitudes. The results show clear differences between occupational classes based on the Oesch classification. Higher-skilled groups with technical or sociocultural work logic report regular AI use significantly more often. Lower usage rates, on the other hand, are found in occupational classes with lower formal qualifications or more standardized job profiles, such as skilled workers, skilled workers in the service sector, or employees in the commercial sector. Overall, it appears that a significant proportion of the working population has rarely or never worked with AI. More frequent AI use is associated with higher subjective AI competence and more positive affective attitudes toward AI in the workplace. Frequent users also rate the expected impact of AI on working conditions much more positively and report fewer negative affective and behavioral reactions to the introduction of AI in the workplace. Overall, the findings point to a digital divide in AI use along lines of age, education, and professional position. |
| Date: | 2026–03–20 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:7fkvq_v1 |
| By: | Christian Gschwendt; Claudio Schilter |
| Abstract: | Generative AI (GenAI) adoption is spreading rapidly and reshaping work, yet its implications for firms' training decisions remain largely unexplored. This paper examines how automation in the post-GenAI era affects firms' entry-level training positions using a vignette experiment with recruiters at over 2, 800 Swiss firms, covering more than 100 distinct occupations. Firms plan to reduce training positions in response to automation prospects, with larger reductions the greater the expected automated task share and the earlier the expected implementation. Effects are markedly stronger in routine-intensive and AI-exposed occupations, as well as among large firms. Our experiment allows us to disentangle an "erosion of the training pipeline, " where firms reduce training even though demand for trained specialists remains, from an overall decline in occupational labor demand. We find that pipeline erosion accounts for less than one third of the average reduction in training, but substantially more when automation is particularly intensive - measured by a high share of tasks being automated - and in routine-intensive and AI-exposed occupations. Overall, the results suggest that GenAI adoption is likely to reallocate firms' human capital investment with potential downstream implications for early career formation, and to reinforce labor market de-routinization trends. |
| Keywords: | Automation, firm training, technological change, generative AI, artificial intelligence, entry-level employment |
| JEL: | J24 M53 O33 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:iso:educat:0252 |
| By: | Ajay K. Agrawal; John McHale; Alexander Oettl |
| Abstract: | We explore the impact of artificial intelligence (AI) on the knowledge production function. We characterize AI as a tool, not for full automation but rather for augmentation through enhanced search over combinatorial spaces. This leads to increased scientific productivity. We decompose knowledge production into a multi-stage process to shed light on the "jagged frontier" of AI in science, revealing differential returns to different tools across domains (e.g., data-rich biology vs. anomaly-sparse physics) and workflow stages (e.g., strong design aids like AlphaFold vs. subtler question generation tools). We treat human judgment as indispensable for tasks involving abductive inference, contextual nuance, and trade-offs, particularly in data-sparse environments. Drawing on a task-based model that distinguishes "ordinary" from AI-expert scientists, we describe how exogenous improvements in AI yield nonlinear productivity gains amplified by the share of scientists that are AI-experts to underscore the role of AI complements like skills training and organizational design. |
| JEL: | I23 O14 O31 O33 O41 |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:34953 |
| By: | Fuchs, Anna; Haensch, Anna-Carolina; Weber, Wiebke |
| Abstract: | Designing survey questions is easy; however designing good survey questions is a complex task. Large language models (LLMs) have the potential to support this task by automating parts of the item-generation process, but their suitability for survey research has not yet been systematically evaluated. Published research in this area remains sparse, and little is known about the quality and characteristics of survey items generated by LLMs or the factors influencing their performance. This work provides the first in-depth analysis of LLM-based survey item generation and systematically evaluates how different design choices affect item quality. Five LLMs, namely GPT-4o, GPT-4o-mini, GPT-oss-20B, LLaMA 3.1 8B, and LLaMA 3.1 70B, were used to generate survey items on four substantive domains: work, living conditions, national politics, and recent politics. We additionally evaluate three prompting strategies: zero-shot, role, and chain-of-thought prompting. To assess the quality of the generated survey items, we use the Survey Quality Predictor (SQP), a tool for estimating the quality of attitudinal survey items based on codings of their formal and linguistic characteristics. To code these characteristics, we used an LLM-assisted procedure. The findings show striking differences in survey item characteristics across the different models and prompting techniques. Both the choice of model and the prompting technique employed influence the quality of LLM-generated survey items. Closed-source GPT models generally produce more consistent items than open-source LLaMA models. Overall, chain-of-thought prompting achieved the best results. GPT-4o, GPT-4o-mini, and LLaMA 3.1 70B achieved similar item quality, while the LLaMA model showed greater variability. |
| Date: | 2026–03–12 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:fzn7t_v1 |
| By: | Chen Zhu; Xiaolu Wang |
| Abstract: | Large language models (LLMs) have enabled agent-based systems that aim to automate scientific research workflows. Most existing approaches focus on fully autonomous discovery, where AI systems generate research ideas, conduct analyses, and produce manuscripts with minimal human involvement. However, empirical research in economics and the social sciences poses additional constraints: research questions must be grounded in available datasets, identification strategies require careful design, and human judgment remains essential for evaluating economic significance. We introduce HLER (Human-in-the-Loop Economic Research), a multi-agent architecture that supports empirical research automation while preserving critical human oversight. The system orchestrates specialized agents for data auditing, data profiling, hypothesis generation, econometric analysis, manuscript drafting, and automated review. A key design principle is dataset-aware hypothesis generation, where candidate research questions are constrained by dataset structure, variable availability, and distributional diagnostics, reducing infeasible or hallucinated hypotheses. HLER further implements a two-loop architecture: a question quality loop that screens and selects feasible hypotheses, and a research revision loop where automated review triggers re-analysis and manuscript revision. Human decision gates are embedded at key stages, allowing researchers to guide the automated pipeline. Experiments on three empirical datasets show that dataset-aware hypothesis generation produces feasible research questions in 87% of cases (versus 41% under unconstrained generation), while complete empirical manuscripts can be produced at an average API cost of $0.8-$1.5 per run. These results suggest that Human-AI collaborative pipelines may provide a practical path toward scalable empirical research. |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2603.07444 |