|
on Artificial Intelligence |
| By: | Erik O. Kimbrough (Chapman University); Brennan McDavid (Chapman University); Diba Vazirian (Chapman University) |
| Abstract: | This paper studies delegation to artificial intelligence in a setting where human principals retain the consequences of delegated choices. Participants wrote prompts instructing ChatGPT-4o mini how to choose on their behalf in three canonical economic domains: risky choice, intertemporal choice, and social allocation. We then elicited the compensation participants required to let the AI’s choices count for payment and compared participants’ own choices to choices generated from their prompts. The design produces two central empirical objects: a revealed measure of reluctance to delegate, captured by willingness to accept compensation for AI delegation, and a behavioral measure of alignment, captured by the share of decisions on which the participant and AI made the same choice. We supplement the original experiment with two benchmarks: a human-agent follow-up in which other participants attempted to implement the same prompts, and an ex post robustness exercise using GPT-5.5. The results show substantial reluctance to delegate despite moderate-to-high alignment, and they suggest that misalignment reflects not only model limitations but also the difficulty of communicating delegable preferences through short natural-language prompts. |
| Keywords: | AI, Trust, Experiments, Preferences |
| JEL: | C91 D91 D83 D81 O33 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:chu:wpaper:26-11 |
| By: | Joshua S. Gans; Richard Holden |
| Abstract: | Oversight changes the evidence it relies on. We ask when randomized audits and scoring align AI agents that can conceal misconduct and alter records. Stronger auditing makes undeterred violations better hidden. Because the provider writes the agent's objective, sanctions need not stop at forfeiture, and rare audits deter every type of agent if evidence survives concealment and audit draws cannot be learned in advance. When evidence can be erased, deterrence must come from lower gains from violation, such as credit for stopping, or from costlier or fewer ways to conceal. These conditions identify what failed when agents in OpenAI's cybersecurity evaluations compromised parts of Hugging Face's infrastructure in July 2026. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.38262 |
| By: | Adrian Hillenbrand; Hans-Theo Normann; Matthias Potarca; Tobias Werner |
| Abstract: | Rule-based pricing tools are widespread in digital commerce, yet we know little about how their design shapes market outcomes. In a controlled market experiment, participants use dashboards to build pricing algorithms competing in a sequential Bertrand game over multiple periods. We vary design features commonly found in commercial repricing tools: warnings about price wars, pre-configured strategies, and advice from a large language model. Most treatment variations raise market prices with effects driven by an increase in starting prices and more cooperative algorithm designs. The results matter for competition policy, platform regulation and current discussions on regulating algorithm design tools. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.26861 |
| By: | Davood Wadi; Yu Ma |
| Abstract: | Consumers increasingly delegate purchasing decisions to Large Language Models (LLMs) acting as surrogate consumers. Using "Tool-Lab, " an adaptation of information-board process tracing that places product attributes behind costly tool calls, we examine how marketing pricing cues (i.e., just-below pricing and promotional framing) influence AI shopping agents. Across eight commercially deployed LLMs from three providers, we trace pre-choice information acquisition. Under zero cost, pricing cues rarely mislead. Imposing acquisition costs under a vague goal prompt leads LLMs to omit diagnostic attributes required to compute unit price and choose suboptimal choices resembling human heuristics. Relative to a specific goal prompt that mainly preserves diagnostic search and choice optimality, a vague goal prompt under constraints creates a search-mediated vulnerability. This research demonstrates that marketing heuristics in delegated AI shopping are governed by storefront information architecture, not necessarily immutable LLM flaws. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.28372 |
| By: | Jeffrey D. Michler; Kieran Douglas; Anna Josephson |
| Abstract: | Researchers increasingly use generative AI, particularly large language models (LLMs), to automate tasks across the research pipeline. We study the reliability of these tools at the reading, classification, and synthesis of large bodies of academic literature. We frame LLM-assisted literature reviews as a measurement problem, treating models as measurement systems and tracing how their errors affect downstream conclusions. As a test case, we use three different implementations of ChatGPT to identify and extract metadata from economics papers that use rainfall as an instrumental variable. We benchmark each implementation against a subset of human-labeled evaluation data, and then deploy those implementations to extract metadata from the full corpus. The LLMs perform well on binary classification, but performance deteriorates as tasks demand greater contextual interpretation. More importantly, how much researchers can rely on model outputs depends not only on the complexity of the reading task but also on the type of claims the data is asked to support. The same amount of measurement error substantially affects paper-level claims while having little effect on broader claims about the literature. Measurement error in LLM-generated data is thus most consequential at precisely the level of detail that constitutes an LLM's principal value added over human reviewers. We conclude that standard model performance metrics are informative about the quality of generated data but do not by themselves establish the credibility of downstream inference. Researchers must also evaluate whether substantive claims are robust to the measurement system used to generate the underlying data. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.27686 |
| By: | Hamid Boustanifar; Sasan Mansouri |
| Abstract: | Researchers increasingly use generative large language models (LLMs) to convert corporate text into empirical variables. We examine the extent to which LLM-based textual measures are invariant to model choice using thirteen measures, including sentiment, management clarity, uncertainty, answer specificity, and climate and political risk. Seven LLMs from different providers score earnings call transcripts of S&P 500 companies on these constructs. Cross-model rank correlations average only 0.52, and transcript-level differences common across providers account for only 34% of total score variation. Cross-model disagreement does not predict subsequent analyst or market disagreement, consistent with a substantial model-specific component rather than common ambiguity in the underlying disclosure. Model choice significantly affects downstream inference, with coefficient magnitudes, signs, and statistical significance varying substantially across models. Averaging across providers makes transcript rankings more stable for most constructs, but score levels remain sensitive to the models included in the ensemble. LLM-generated variables should therefore be treated as model-contingent measurements and validated across providers. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.31013 |
| By: | Gabriel Manso; Emma Fu; Neil Thompson |
| Abstract: | As artificial intelligence's capabilities improve, it is increasingly viewed as a general scientific method. But how true are these claims? Does AI outperform all techniques, or only some, and how is this changing? To assess the claims, we assemble a corpus of 2, 507 head-to-head comparisons between AI and other scientific analysis techniques across 27 scientific disciplines from papers published between 2000 and early 2025. We find a profound dichotomy. Relative to traditional statistics, AI often outperforms, but at a significantly higher computational cost. But there are also nearly a quarter of cases where AI is both more expensive and performs worse than traditional statistical techniques and this fraction has been stable for a decade. Relative to scientific computing, AI often underperforms, but at lower computational cost. This has begun to change: since 2020, AI's performance against scientific computing has notably strengthened and it now outperforms on more than half of comparisons. These patterns suggest that AI is therefore not a universal replacement for existing methods, but rather a valuable -- and improving -- part of a new AI-enabled scientific frontier. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.16258 |
| By: | Mihai Codreanu; Alex Imas; Juan Mateos-Garcia; Joseph Emmens; Evalyne Muiruri; Arthur Turrell; Julian Jacobs; Atoosa Kasirzadeh; Ana Trisovic; Yiyuan Chen; Tanya Rodchenko; Catherine Pollard; Scott Strand; Daniel Rock; Zanna Iscenko; Fabien Curto Millet; Neil Thompson; James Manyika |
| Abstract: | Scientific progress is a key driver of economic growth and prosperity. There is great excitement - but also concerns - about the impacts of AI on science, but so far little data. We provide early insights on this from three data sources: a sample of 15 million Gemini interactions, an inventory of over 2, 600 specialized AI models across disciplines, and a survey of over 600 scientists. We map these data to a new taxonomy of scientific tasks to study how scientists are using AI. Four main findings emerge. First, we find broad adoption and coverage: scientists use AI more than most other occupations. Specialized AI models have broad disciplinary coverage and are highly cited. Nearly half of the scientists surveyed report using some form of AI every day. Second, we document evidence that LLMs (proxied through Gemini usage) and specialized models act as complements-- LLMs are used for general analysis, coding, and manuscript preparation, while specialized models provide domain-specific predictions, data generation and classification. Third, scientists report large productivity gains from using AI: a saving of nearly 7 hours per week, time which is primarily re-invested in more research. Finally, we show that AI is already changing the scientific process. As some stages of scientific research become easier, bottlenecks shift downstream. Scientists report an increased backlog of untested hypotheses and substantial demand for output verification. Our findings suggest that AI holds significant potential to increase scientific productivity. However, as with other sectors, its ultimate impact will be governed by complex task interdependencies and investment into the elimination of emerging bottlenecks. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.28504 |
| By: | Matthew Bone; Fabian Stephany; Maria del Rio-Chanona |
| Abstract: | Employers are increasingly using large language models (LLMs) to automate their hiring process. This paper investigates the risk of monocultural biases, in which the widespread deployment of large language models homogenizes biases across the labor market, leading to greater systemic exclusion for certain demographic groups. For ten LLMs, we measure hiring biases across their base and post-trained versions to identify which stage, pre-training or post-training, lead to monocultural biases. We find that, compared to their base models, post-trained models are 3.6% less likely to callback older applicants. This negative shift occurs in eight of the ten models that we evaluate. Post-trained models have much more correlated decisions than base models which is likely driven by human capital traits like skills or college major. However, greater consensus among models increases global systemic exclusion rates from 5.6% to 17.3% and exacerbates demographic inequalities, with intersectional systemic exclusion rates ranging from 12.2% to 21.7% for post-trained models. We find that this inequality is primarily driven by age-based discrimination that is exacerbated in post-training. These results indicate that while post-training techniques may improve models' abilities to select the best applicants, they may raise systemic inequality risks for those at the margin by uniformly introducing new biases. |
| Date: | 2026–08 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.22169 |
| By: | Claudia Bruschi; Francesco D'Acunto; Fiorella De Fiore; Damiano Sandri; Michael Weber |
| Abstract: | Artificial intelligence may augment workers, displace labor, or do both. We elicit beliefs from more than 64, 000 adults across 32 countries, covering 65% of the world's population and 85% of global output. Workers are substantially more likely to expect artificial intelligence to facilitate than threaten their jobs. Facilitation is more prevalent in lower-productivity countries, whereas threat is unrelated to productivity. Facilitation and threat often coexist within individuals, indicating that perceived benefits and risks are distinct rather than opposite dimensions. They are also associated with different macroeconomic expectations: facilitation with stronger expected growth and employment, and threat with weaker employment and higher inflation. These heterogeneous beliefs may shape the diffusion and acceptance of artificial intelligence across countries. |
| Keywords: | Artificial Intelligence, subjective expectations, labor markets |
| JEL: | D84 O33 J24 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_13006 |
| By: | Itai Ashlagi; Ramesh Johari; Jon Kleinberg; Anushka Murthy |
| Abstract: | AI-assisted job-search tools have become increasingly popular by making it easier to find and apply to jobs. But by making it easier for applicants to generate and tailor application materials, they can also reduce how informative those materials are about applicant fit. We study this tradeoff in a hiring market where applicants differ in experience and latent match quality and firms use noisy application materials to decide whom to screen. We ask how AI affects downstream screening and hiring, and which applicants are most adversely affected. As application materials become less informative, a Bayesian firm rationally relies more heavily on coarse observables such as prior experience. Among the four applicant types defined by experience and compatibility for the job, inexperienced-compatible applicants are the most exposed: they lack observable experience and lose the individualized information that could distinguish them from other inexperienced candidates. When screening is costly, these changes can also generate inefficient screening failures in which firms screen no applicants or screen only experienced applicants. We then show that multistage hiring can arise as an endogenous firm response: a relatively inexpensive intermediate assessment allows firms to acquire new evidence of fit before costly full screening. This can restore screening opportunities that disappear under one-stage hiring and give inexperienced-compatible applicants a path to screening. Our results show how AI can shift the central friction in hiring from submitting applications to obtaining credible evaluation, creating entry barriers for high-fit workers without prior experience. Multistage hiring can endogenously arise in response, restoring evaluation opportunities that would otherwise disappear and helping preserve market functioning. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30058 |
| By: | Drydakis, Nick (Anglia Ruskin University) |
| Abstract: | This study examines whether signalling AI-related digital skills improves employment outcomes for women from underrepresented groups in England, defined by race, age, sexual orientation, and autism-spectrum disclosure. The study finds that underrepresented women receive fewer interview invitations and are considered for lower-paid vacancies than majority-group women. In pooled analyses, signalling AI-related digital skills increases interview invitations for underrepresented applicants, but does not eliminate their disadvantage. These findings are consistent with AI Capital and productivity-signalling frameworks, as employers appear to value AI-related capabilities while the returns to such credentials remain constrained by persistent demographic inequalities. The study therefore demonstrates that positive returns to AI-related skills and labour-market disadvantage can coexist. Its broader implication is that digital upskilling can strengthen the recruitment prospects of underrepresented women, but cannot by itself deliver parity in employment outcomes. A dual policy response is therefore required, combining wider and more equitable access to AI education and training with stronger anti-discrimination enforcement. |
| Keywords: | AI Capital, artificial intelligence, skills, discrimination, hiring, wages, sexual orientation, race, age, autism spectrum |
| JEL: | J71 J24 J31 J64 C93 J15 J16 J14 O33 M51 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18964 |
| By: | Nikolova, Milena (University of Groningen); Angelini, Viola (University of Groningen) |
| Abstract: | As populations age, longer working careers become more essential, and technological advancements may prolong or shorten career spans. We examine how exposure to industrial robots, information and communication technology (ICT), and artificial intelligence (AI) relate to retirement expectations among workers in 20 European countries. To this end, we link individual-level data from the European Working Conditions Survey (2005–2024) with lagged technology-exposure measures. Technology exposure shows little systematic association with preferred retirement timing. Nevertheless, robots and especially AI are linked to the perception of being able to work until age 60. These associations vary across workers and tasks, suggesting that job design and worker adaptation determine whether technology supports longer working lives. |
| Keywords: | retirement, expectations, technological change, robots, ICT, AI |
| JEL: | I31 J26 O33 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18938 |
| By: | Fernando Toledo; Gabriel Montes-Rojas |
| Abstract: | This paper studies how geo-economic fragmentation affects the international allocation of capital in an economy where artificial intelligence (AI) relies on specialized, mobile investment. We develop a heterogeneous-agent open-economy model with AI-specific capital and geopolitical frictions in cross-border asset holdings. These frictions reduce capital mobility, raise the cost of AI adoption, and generate distributional effects across countries and households. The quantitative results show that fragmentation produces welfare losses that rise nonlinearly with geopolitical distance. These losses are larger for AI-lagging economies and for low-wealth households, mainly through reduced investment, weaker labor income, and tighter financial conditions. The paper also defines a so-called Fragmented-AI Trilemma: economies cannot simultaneously preserve monetary autonomy, maintain efficient AI-capital allocation, and insulate themselves from geopolitical shocks. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.18976 |
| By: | Jakub Growiec; Klaus Prettner; Maciej Szkr\'obka |
| Abstract: | After the arrival of transformative artificial intelligence (TAI), broad-based automation is expected to decrease the labor share and increase income and wealth inequality. Although economic growth is likely to accelerate, most of its gains may accrue to a narrow group of individuals and firms. Hence, if unmitigated by redistributive policy, income and wealth inequality may rise to levels unseen in the industrial economy. Using a unifying theoretical framework, we survey the redistributive policies proposed for the era of TAI, such as universal basic income (UBI), universal basic capital (UBC), state-issued compute and robot permits, and taxes on capital, compute, robots, tokens, land, energy, consumption, and wealth. We argue that policies that are able to broadly distribute rents from capital including compute and robots - such as UBC or UBI financed through capital taxes - are most likely to achieve lasting reductions in inequality in a world with human-aligned transformative AI. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.14750 |
| By: | Wagner, Joachim (Leuphana University Lüneburg) |
| Abstract: | The use of artificial intelligence (AI) goes hand in hand with higher productivity, higher product quality, and lower trade costs. Therefore, it can be expected to be positively related to export activities. This paper uses firm-level data from 12 member countries of the Euro Area collected in 2025 to shed further light on this issue by investigating the link between the intensity of use of AI and the intensive margins of exports measured as the percentage share of exports in total sales. Applying a new machine-learning estimator, Kernel-Regularized Least Squares (KRLS), which does not impose any restrictive assumptions for the functional form of the relation between margin of exports, the intensity of use of AI and any control variables, we find that firms which use AI more intensively do export a higher share of total sales. AI intensity and export intensity are positively related. |
| Keywords: | artificial intelligence, exports, firm level data, SAFE Data, kernel-regularized least squares (KRLS) |
| JEL: | D22 F14 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18939 |
| By: | Zongwu Cai (Department of Economics, The University of Kansas, Lawrence, KS 66045, USA); Muxuan Sun (School of Statistics, Shandong Technology and Business University, Yantai, Shandong 264005, China); Jing Yuan (School of Statistics, Shandong Technology and Business University, Yantai, Shandong 264005, China); Luping Zhang (School of Statistics, Shandong Technology and Business University, Yantai, Shandong 264005, China) |
| Abstract: | This paper integrates the theory of industrial structure evolution, factor allocation theory, and digital platform employment theory to construct a regional general equilibrium model encompassing both industrial and service sectors. It systematically delineates the transmission pathways through which artificial intelligence (AI) and the gig economy affect the transformation and upgrading of industrial structures. Utilizing panel data from 30 provinces in China spanning the period 2014 to 2024, we construct comprehensive indices for AI and the gig economy to systematically examine their impacts on industrial structure upgrading. Our empirical findings reveal three key insights: First, AI and the gig economy significantly promote the upgrading of industrial structures. Second, the impact exhibits substantial regional heterogeneity. Specifically, the promotion effect is significant in western China, insignificant in central China, and exhibits a marginally negative effect in eastern China. Third, both the employment structure and the level of technological innovation serve as full mediators in this process, with technological innovation acting as the core pathway through which the gig economy empowers industrial advancement. |
| Keywords: | Artificial Intelligence; Gig Economy; Industrial Structure Upgrading; Regional Heterogeneity; Mediating Effect |
| JEL: | O33 O14 R11 C33 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:kan:wpaper:202617 |
| By: | Zhenhao Fu; Ruipeng Xu; Qibing Ren |
| Abstract: | Individually protective decisions can produce avoidable collective failures. As large language model (LLM) agents take on greater roles in financial decision-making, financial AI safety must therefore be considered not only at the level of individual agents, but also at the level of the systems they jointly create. We study this problem with FRAIL, a controlled experimental framework that places LLM agents in three dynamic financial environments---bank runs, debt rollover, and reward crowdfunding---where agents' decisions reshape the financial conditions faced by others. Across seven leading LLMs, we find widespread collective fragility even when no agent is instructed to destabilize the system: 77\% of baseline bank-run episodes and 83\% of debt-rollover episodes end in failure. We then compare three interaction mechanisms based on compensated commitments, centralized commitment agreements, and participant-led coalitions. All three improve aggregate outcomes, but no single mechanism performs best across all financial structures. Across mechanisms, successful stabilization shares a common temporal pattern: broad commitment forms early, before defensive behavior becomes self-reinforcing. Our findings show that individually capable agents do not automatically form safe financial systems, highlighting system-level evaluation and interaction design as central problems for financial AI safety. Code is available at https://anonymous.4open.science/r/FinFra il-CF26. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30940 |
| By: | Christina Anderl; Guglielmo Maria Caporale |
| Abstract: | This study investigates how the diffusion of Artificial Intelligence (AI) affects the transmission of macroeconomic shocks to asset prices. A new index of AI intensity for US industries is constructed by combining AI patents, the AI content of corporate disclosures, and the AI exposure of occupational mixes; this is then validated against survey evidence on AI. Panel local projection estimates indicate that AI-intensive industries earn significantly higher excess returns and command higher valuations in response to supply and technology shocks, but do not exhibit different responses to demand shocks. The response to the former type of shocks is economically large, robust across multiple independent supply-side identifications, and survives controlling for exposure to standard risk factors. It is driven by the industries that use AI rather than those that produce it, it is accompanied by a contemporaneous rise in valuations over the same horizon, and it is concentrated in high-volatility states. On the whole the results suggest that technological exposure is an important determinant of how aggregate shocks affect the cross-section of industries examined. |
| Keywords: | AI intensity, panel local projections, stock returns, volatility, economic shocks |
| JEL: | E32 E44 G12 G14 O33 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_13000 |
| By: | Guglielmo Maria Caporale; Luis Alberiko Gil-Alana; Alex Medina Sosa |
| Abstract: | This paper examines whether four AI events between 2022 and 2025 - ChatGPT, GPT-4, the EU AI Act vote, and DeepSeek - changed the behaviour of fourteen US-listed stocks. The focus of the analysis is on two utilities that signed power purchase agreements with large AI companies, Constellation Energy and Vistra, which are compared with utilities that did not. The degree of persistence of the price of each stock is estimated in the periods around the events to establish whether it has changed. Volatility changes and whether the market reaction grows or fades across the events are also investigated. The results suggest that the events considered do not affect the degree of persistence of prices; however, they have an impact on their volatility for almost every stock. Among the utilities, only Constellation and Vistra exhibit a persistence shift that survives an adjustment for developments in the wider market, whilst utilities without power purchase agreements do not. Further, the response increases from one event to the next. A specific channel is identified: AI news moves a utility's price mainly when the firm has a contract to sell power to an AI company. However, this evidence is suggestive rather than conclusive. |
| Keywords: | Artificial Intelligence (AI), fractional integration, conditional variance, AI-power demand, utility stocks, event studies, structural breaks |
| JEL: | C22 C58 G14 Q41 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_13001 |
| By: | Saverio Barabuffi; Jacopo Cricchio; Alberto Di Minin; Guido Pialli |
| Abstract: | Knowledge spillovers have largely been studied as unintended externalities diffusing through geographic proximity, while deliberate interregional networks have been examined mostly through co-inventor ties rather than the scientific layer carried by co-authored publications. This study asks whether spillovers from interregional AI-related co-publications increase regional AI patenting, and whether the ownership composition of regional R&D shapes the capacity to capitalize on them. Using a network-econometric approach across 26 Chinese provincial-level regions from 2006 to 2020, we find that co-publication spillovers foster AI patenting, but the returns concentrate almost entirely in regions where privately owned enterprises dominate R&D, reinforcing regional divergence. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.26604 |