|
on Artificial Intelligence |
| By: | David Almog; Romain Gauriot; Lionel Page; Daniel Martin |
| Abstract: | Powered by the increasing predictive capabilities of machine learning algorithms, artificial intelligence (AI) systems have begun to be used to overrule human mistakes in many settings. We provide the first field evidence this AI oversight carries psychological costs that can impact human decision-making. We investigate one of the highest visibility settings in which AI oversight has occurred: the Hawk-Eye review of umpires in top tennis tournaments. We find that umpires lowered their overall mistake rate after the introduction of Hawk-Eye review, in line with rational inattention given psychological costs of being overruled by AI. We also find that umpires increased the rate at which they called balls in, which produced a shift from making Type II errors (calling a ball out when in) to Type I errors (calling a ball in when out). We structurally estimate the psychological costs of being overruled by AI using a model of rational inattentive umpires, and our results suggest that because of these costs, umpires cared twice as much about Type II errors under AI oversight. |
| Date: | 2024–01 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2401.16754&r=ain |
| By: | Greiner, Ben; Grünwald, Philipp; Lindner, Thomas; Lintner, Georg; Wiernsperger, Martin |
| Abstract: | Managerial decision-makers are increasingly supported by advanced data analytics and other AI-based technologies, but are often found to be hesitant to follow the algorithmic advice. We examine how compensation contract design and framing of an AI algorithm influence decision-makers’ reliance on algorithmic advice and performance in a price estimation task. Based on a large sample of almost 1, 500 participants, we find that compared to a fixed compensation, both compensation contracts based on individual performance and tournament contracts lead to an increase in effort duration and to more reliance on algorithmic advice. We further find that using an AI algorithm that is framed as incorporating also human expertise has positive effects on advice utilization, especially for decision-makers with fixed pay contracts. By showing how widely used control practices such as incentives and task framing influence the interaction of human decision-makers with AI algorithms, our findings have direct implications for managerial practice. |
| Keywords: | artificial intelligence; algorithmic advice; human-augmented algorithmic advice; trust; financial incentives; decision-making |
| Date: | 2024–01–31 |
| URL: | http://d.repec.org/n?u=RePEc:wiw:wus055:60237853&r=ain |
| By: | Alexander Erlei; Lukas Meub |
| Abstract: | In credence goods markets such as health care or repair services, consumers rely on experts with superior information to adequately diagnose and treat them. Experts, however, are constrained in their diagnostic abilities, which hurts market efficiency and consumer welfare. Technological breakthroughs that substitute or complement expert judgments have the potential to alleviate consumer mistreatment. This article studies how competitive experts adopt novel diagnostic technologies when skills are heterogeneously distributed and obfuscated to consumers. We differentiate between novel technologies that increase expert abilities, and algorithmic decision aids that complement expert judgments, but do not affect an expert's personal diagnostic precision. We show that high-ability experts may be incentivized to forego the decision aid in order to escape a pooling equilibrium by differentiating themselves from low-ability experts. Results from an online experiment support our hypothesis, showing that high-ability experts are significantly less likely than low-ability experts to invest into an algorithmic decision aid. Furthermore, we document pervasive under-investments, and no effect on expert honesty. |
| Date: | 2024–01 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2401.17929&r=ain |
| By: | Christoph Engel (Max Planck Institute for Research on Collective Goods); Richard H. McAdams (University of Chicago Law School) |
| Abstract: | We report on our test of the Large Language Model (LLM) ChatGPT (GPT) as a tool for generating evidence of the ordinary meaning of statutory terms. We explain why the most useful evidence for interpretation involves a distribution of replies rather than only what GPT regards as the single “best†reply. That motivates our decision to use Chat 3.5 Turbo instead of Chat 4 and to run each prompt we use 100 times. Asking GPT whether the statutory term “vehicle†includes a list of candidate objects (e.g., bus, bicycle, skateboard) allows us to test it against a benchmark, the results of a high-quality experimental survey (Tobia 2000) that asked over 2, 800 English speakers the same questions. After learning what prompts fail and which one works best (a belief prompt combined with a Likert scale reply), we use the successful prompt to test the effects of “informing†GPT that the term appears in a particular rule (one of five possible) or that the legal rule using the term has a particular purpose (one of six possible). Finally, we explore GPT’s sensitivity to meaning at a particular moment in the past (the 1950s) and its ability to distinguish extensional from intensional meaning. To our knowledge, these are the first tests of GPT as a tool for generating empirical data on the ordinary meaning of statutory terms. Legal actors have good reason to be cautious, but LLMs have the potential to radically facilitate and improve legal tasks, including the interpretation of statutes. |
| Date: | 2024–02 |
| URL: | http://d.repec.org/n?u=RePEc:mpg:wpaper:2024_05&r=ain |
| By: | Churchill, Alexander; Pichika, Shamitha; Xu, Chengxin (Seattle University) |
| Abstract: | Supervised content encoding applies a given codebook to a larger non-numerical dataset and is central to empirical research in public administration. Not only is it a key analytical approach for qualitative studies, but the method also allows researchers to measure constructs using non-numerical data, which can then be applied to quantitative description and causal inference. Despite its utility, supervised content encoding faces challenges including high cost and low reproducibility. In this report, we test if large language models (LLM), specifically generative pre-trained transformers (GPT), can solve these problems. Using email messages collected from a national corresponding experiment in the U.S. nursing home market as an example, we demonstrate that although we found some disparities between GPT and human coding results, the disagreement is acceptable for certain research design, which makes GPT encoding a potential substitute for human encoders. Practical suggestions for encoding with GPT are provided at the end of the letter. |
| Date: | 2024–01–25 |
| URL: | http://d.repec.org/n?u=RePEc:osf:socarx:6fpgj&r=ain |
| By: | Hauke Licht (University of Cologne); Ronja Sczepanski (Sciences Po Paris); Moritz Laurer (Hugging Face; Vrije Universiteit Amsterdam); Ayjeren Bekmuratovna (DHL) |
| Abstract: | As more and more scholars apply computational text analysis methods to multilingual corpora, machine translation has become an indispensable tool. However, relying on commercial services for machine translation, such as Google Translate or DeepL, limits reproducibility and can be expensive. This paper assesses the viability of a reproducible and affordable alternative: free and open-source machine translation models. We ask whether researchers who use an open-source model instead of a commercial service for machine translation would obtain substantially different measurements from their multilingual corpora. We address this question by replicating and extending an influential study by de Vries et al. (2018) on the use of machine translation in cross-lingual topic modeling, and an original study of its use in supervised text classification with Transformer-based classifiers. We find only minor differences between the measurements generated by these methods when applied to corpora translated with open-source models and commercial services, respectively. We conclude that “free” machine translation is a very valuable addition to researchers’ multilingual text analysis toolkit. Our study adds to a growing body of work on multilingual text analysis methods and has direct practical implications for applied researchers. |
| Keywords: | machine translation, multilingual topic modeling, multilingual Transformers |
| JEL: | C45 |
| Date: | 2024–02 |
| URL: | http://d.repec.org/n?u=RePEc:ajk:ajkdps:276&r=ain |
| By: | Leigh, Andrew (Parliament of Australia) |
| Abstract: | Artificial intelligence has the potential to be a valuable competitive force in product and service markets. Yet AI may also pose competitive problems. I identify five big challenges that AI poses for competition. (1) Costly chips. (2) Private data. (3) Network effects. (4) Immobile talent. (5) An 'open-first, closed-later' model. These are not just issues for our competition regulators, but also for competition reformers. Just as antitrust laws needed to be updated to deal with the misbehaviour of the oil titans and rail barons of nineteenth century America, so too we may need to make changes in competition laws to address the challenges that AI poses. |
| Keywords: | competition, antitrust, artificial intelligence |
| JEL: | L40 L63 |
| Date: | 2024–02 |
| URL: | http://d.repec.org/n?u=RePEc:iza:izapps:pp209&r=ain |
| By: | Mario Sanz-Guerrero; Javier Arroyo |
| Abstract: | Peer-to-peer (P2P) lending has emerged as a distinctive financing mechanism, linking borrowers with lenders through online platforms. However, P2P lending faces the challenge of information asymmetry, as lenders often lack sufficient data to assess the creditworthiness of borrowers. This paper proposes a novel approach to address this issue by leveraging the textual descriptions provided by borrowers during the loan application process. Our methodology involves processing these textual descriptions using a Large Language Model (LLM), a powerful tool capable of discerning patterns and semantics within the text. Transfer learning is applied to adapt the LLM to the specific task at hand. Our results derived from the analysis of the Lending Club dataset show that the risk score generated by BERT, a widely used LLM, significantly improves the performance of credit risk classifiers. However, the inherent opacity of LLM-based systems, coupled with uncertainties about potential biases, underscores critical considerations for regulatory frameworks and engenders trust-related concerns among end-users, opening new avenues for future research in the dynamic landscape of P2P lending and artificial intelligence. |
| Date: | 2024–01 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2401.16458&r=ain |
| By: | Helena Chuliá (Riskcenter- IREA and Department of Econometrics and Statistics, University of Barcelona.); Sabuhi Khalili (Department of Econometrics and Statistics, University of Barcelona.); Jorge M. Uribe (Faculty of Economics and Business Studies, Open University of Catalonia.) |
| Abstract: | SWe propose generative artificial intelligence to measure systemic risk in the global markets of sovereign debt and foreign exchange. Through a comparative analysis, we explore three novel models to the economics literature and integrate them with traditional factor models. These models are: Time Variational Autoencoders, Time Generative Adversarial Networks, and Transformer-based Time-series Generative Adversarial Networks. Our empirical results provide evidence in support of the Variational Autoencoder. Results here indicate that both the Credit Default Swaps and foreign exchange markets are susceptible to systemic risk, with a historically high probability of distress observed by the end of 2022, as measured by both the Joint Probability of Distress and the Expected Proportion of Markets in Distress. Our results provide insights for governments in both developed and developing countries, since the realistic counterfactual scenarios generated by the AI, yet to occur in global markets, underscore the potential worst-case scenarios that may unfold if systemic risk materializes. Considering such scenarios is crucial when designing macroprudential policies aimed at preserving financial stability and when measuring the effectiveness of the implemented policies. |
| Keywords: | Twin Ds, Sovereign Debt, Credit Risk, TimeGANs, Transformers, TimeVAEs, Autoencoders, Variational Inference. JEL classification: C45, C53, F31, F37. |
| Date: | 2024–02 |
| URL: | http://d.repec.org/n?u=RePEc:ira:wpaper:202402&r=ain |
| By: | LOMINE, LOYKIE |
| Abstract: | The role of Artificial Intelligence (AI) in higher education is generating considerable debate, including in business schools. Drawing insights from recent publications (both academic and journalistic) and from examples of business schools around the world, this paper explores the potential of AI as a catalyst for sustainable education. It is structured around the alignment of AI's educational benefits with four of the Sustainable Development Goals (SDGs): SDG 4 (Quality Education), SDG 9 (Industry, Innovation, and Infrastructure), SDG 12 (Responsible Consumption and Production) and SDG 17 (Partnerships for the Goals). Key findings suggest that AI's capabilities in offering personalized learning experiences, fostering innovation, promoting responsible consumption and bolstering sustainable partnerships position IA as an essential tool for business schools. This paper ultimately advocates for the deliberate and strategic integration of AI to further the mission of sustainability education of business schools worldwide. |
| Date: | 2024–02–03 |
| URL: | http://d.repec.org/n?u=RePEc:osf:osfxxx:64y38&r=ain |