nep-ain New Economics Papers
on Artificial Intelligence
Issue of 2023‒10‒23
twelve papers chosen by
Ben Greiner, Wirtschaftsuniversität Wien

  1. Generation Next: Experimentation with AI By Gary Charness; Brian Jabarian; John A. List
  2. Recruiters' Behaviors Faced with Dual (AI and human) Recommendations in Personnel Selection By Alain Lacroux; Christelle Martin Lacroux
  3. Consumer Acceptance of Mobile Shopping Apps, From Basic Apps to AI-Conversational Apps: A Literature Review By Taoufiq Dadouch; Bouchra Bennani; Malika Haoucha
  4. Human-AI Interactions and Societal Pitfalls By Francisco Castro; Jian Gao; S\'ebastien Martin
  5. A Comprehensive Review on Financial Explainable AI By Wei Jie Yeo; Wihan van der Heever; Rui Mao; Erik Cambria; Ranjan Satapathy; Gianmarco Mengaldo
  6. Startup success prediction and VC portfolio simulation using CrunchBase data By Mark Potanin; Andrey Chertok; Konstantin Zorin; Cyril Shtabtsovsky
  7. Transformers versus LSTMs for electronic trading By Paul Bilokon; Yitao Qiu
  8. Automatic Product Classification in International Trade: Machine Learning and Large Language Models By Marra de Artiñano, Ignacio; Riottini Depetris, Franco; Volpe Martincus, Christian
  9. Responsible artificial intelligence in Africa: Towards policy learning By Plantinga, Paul; Shilongo, Kristophina; Mudongo, Oarabile; Umubyeyi, Angelique; Gastrow, Michael; Razzano, Gabriella
  10. AI Adoption in America: Who, What, and Where By Kristina McElheran; J. Frank Li; Erik Brynjolfsson; Zachary Krof; Emin Dinlersoz; Lucia Foster; Nikolas Zolas
  11. AI Adoption and Productivity of Japanese Firms: Spillover and innovation effects (Japanese) By IKEUCHI Kenta; INUI Tomohiko; KIM YoungGak
  12. Exporting the Surveillance State via Trade in AI By Martin Beraja; Andrew Kao; David Y. Yang; Noam Yuchtman

  1. By: Gary Charness; Brian Jabarian; John A. List
    Abstract: We investigate the potential for Large Language Models (LLMs) to enhance scientific practice within experimentation by identifying key areas, directions, and implications. First, we discuss how these models can improve experimental design, including improving the elicitation wording, coding experiments, and producing documentation. Second, we discuss the implementation of experiments using LLMs, focusing on enhancing causal inference by creating consistent experiences, improving comprehension of instructions, and monitoring participant engagement in real time. Third, we highlight how LLMs can help analyze experimental data, including pre-processing, data cleaning, and other analytical tasks while helping reviewers and replicators investigate studies. Each of these tasks improves the probability of reporting accurate findings.
    JEL: C0 C1 C80 C82 C87 C9 C90 C92 C99
    Date: 2023–09
  2. By: Alain Lacroux (UP1 EMS - Université Paris 1 Panthéon-Sorbonne - École de Management de la Sorbonne - UP1 - Université Paris 1 Panthéon-Sorbonne); Christelle Martin Lacroux (Université Grenoble Alpes IUT2)
    Abstract: Artificial Intelligence (AI) is increasingly used for decision-making support in organizations, and especially during the recruitment process. Consequently, recruiters may sometimes find themselves having to process different sources of information (human vs. algorithmic decision support system, ADSS) before deciding to preselect an applicant. Our study aims to explore the mechanisms that lead recruiters to follow or not the recommendations made by human and non-human experts, in particular when they receive contradictory or inaccurate information from these sources. Drawing on results obtained in the field of automated decision support, we make a first general hypothesis that recruiters trust human experts more than ADSS and rely more on their recommendations. Secondly, based on the Judge Advisor System Paradigm (Sniezek & Buckley, 1995), we make a second general hypothesis that the accuracy of the recommendations provided by the dual source of advice influences in different ways the accuracy of recruiters' preselection decisions. We conducted an experiment involving the screening of resumes by a sample of professionals (N=746) responsible for screening job applications in their work. As hypothesized, the recommendations made to recruiters do influence the accuracy of their decisions. Our results suggest that recruiters comply more with ADSS than human recommendations even if they declare a higher level of trust in human experts. Finally, implications for research and HR policies are discussed.
    Keywords: Artificial intelligence (AI), trust, resume screening, augmented recruitment
    Date: 2023–08–05
  3. By: Taoufiq Dadouch (University Hassan II [Casablanca]); Bouchra Bennani; Malika Haoucha
    Abstract: The rapid proliferation of Digital marketing, due to recent digital transformation, has been accentuated by the effects of the Covid-19 pandemic. This can be noticed with changes in customer shopping behavior while adopting various digital marketing tools such as social media, E & M-commerce, and very recently AI enablers such as conversational agents/apps (Virtual Assistants & Chatbots). The purpose of this paper is to present some literature findings on consumer behavior toward mobile shopping via AI-Conversational-apps (Virtual Assistants & Chatbots), as compared to Mobile basic apps. Indeed, Mobile Shopping via AI-Conversational apps and their consumer acceptance behavior have become an important research issue worldwide in terms of involved predictors, theories, and methodologies. In summary, the literature showed that Anthropomorphism Construct (i.e., the degree to which a user perceives AI-Conversational apps to be humanlike) emerged as the primary additional predictor for acceptance of M-Shopping via AI-Conversational apps (AI-CA), in addition to mobile primary apps determinants. These determinants consist of utilitarian, hedonic & social antecedents adapted mainly from the UTAUT2 model (Unified theory of acceptance & use of technology), including mainly; performance expectation & effort expectation, hedonic motivation, social influence, and facilitating conditions. Literature findings also clarified the lack & importance of multimarket & multicultural research on M-Shopping apps' acceptance (mainly AI-CA). Indeed, not only developed markets but also developing ones, have seen surging rates of smartphone penetration conditions & mobile internet connectivity, along with changing consumer behaviors and dominating M-Shopping-apps activities. This offers great potential for research on M-Shopping-AI-CA acceptance behaviors in such developing countries, mainly in Morocco.
    Keywords: Digital Marketing Consumer behavior Mobile Shopping Artificial Intelligence AI-Conversational apps. JEL Classification : M21 M31 M37 M39 Paper type: Theoretical Research, Digital Marketing, Consumer behavior, Mobile Shopping, Artificial Intelligence, AI-Conversational apps. JEL Classification : M21, M31, M37, M39 Paper type: Theoretical Research
    Date: 2023–08–29
  4. By: Francisco Castro; Jian Gao; S\'ebastien Martin
    Abstract: When working with generative artificial intelligence (AI), users may see productivity gains, but the AI-generated content may not match their preferences exactly. To study this effect, we introduce a Bayesian framework in which heterogeneous users choose how much information to share with the AI, facing a trade-off between output fidelity and communication cost. We show that the interplay between these individual-level decisions and AI training may lead to societal challenges. Outputs may become more homogenized, especially when the AI is trained on AI-generated content. And any AI bias may become societal bias. A solution to the homogenization and bias issues is to improve human-AI interactions, enabling personalized outputs without sacrificing productivity.
    Date: 2023–09
  5. By: Wei Jie Yeo; Wihan van der Heever; Rui Mao; Erik Cambria; Ranjan Satapathy; Gianmarco Mengaldo
    Abstract: The success of artificial intelligence (AI), and deep learning models in particular, has led to their widespread adoption across various industries due to their ability to process huge amounts of data and learn complex patterns. However, due to their lack of explainability, there are significant concerns regarding their use in critical sectors, such as finance and healthcare, where decision-making transparency is of paramount importance. In this paper, we provide a comparative survey of methods that aim to improve the explainability of deep learning models within the context of finance. We categorize the collection of explainable AI methods according to their corresponding characteristics, and we review the concerns and challenges of adopting explainable AI methods, together with future directions we deemed appropriate and important.
    Date: 2023–09
  6. By: Mark Potanin; Andrey Chertok; Konstantin Zorin; Cyril Shtabtsovsky
    Abstract: Predicting startup success presents a formidable challenge due to the inherently volatile landscape of the entrepreneurial ecosystem. The advent of extensive databases like Crunchbase jointly with available open data enables the application of machine learning and artificial intelligence for more accurate predictive analytics. This paper focuses on startups at their Series B and Series C investment stages, aiming to predict key success milestones such as achieving an Initial Public Offering (IPO), attaining unicorn status, or executing a successful Merger and Acquisition (M\&A). We introduce novel deep learning model for predicting startup success, integrating a variety of factors such as funding metrics, founder features, industry category. A distinctive feature of our research is the use of a comprehensive backtesting algorithm designed to simulate the venture capital investment process. This simulation allows for a robust evaluation of our model's performance against historical data, providing actionable insights into its practical utility in real-world investment contexts. Evaluating our model on Crunchbase's, we achieved a 14 times capital growth and successfully identified on B round high-potential startups including Revolut, DigitalOcean, Klarna, Github and others. Our empirical findings illuminate the importance of incorporating diverse feature sets in enhancing the model's predictive accuracy. In summary, our work demonstrates the considerable promise of deep learning models and alternative unstructured data in predicting startup success and sets the stage for future advancements in this research area.
    Date: 2023–09
  7. By: Paul Bilokon; Yitao Qiu
    Abstract: With the rapid development of artificial intelligence, long short term memory (LSTM), one kind of recurrent neural network (RNN), has been widely applied in time series prediction. Like RNN, Transformer is designed to handle the sequential data. As Transformer achieved great success in Natural Language Processing (NLP), researchers got interested in Transformer's performance on time series prediction, and plenty of Transformer-based solutions on long time series forecasting have come out recently. However, when it comes to financial time series prediction, LSTM is still a dominant architecture. Therefore, the question this study wants to answer is: whether the Transformer-based model can be applied in financial time series prediction and beat LSTM. To answer this question, various LSTM-based and Transformer-based models are compared on multiple financial prediction tasks based on high-frequency limit order book data. A new LSTM-based model called DLSTM is built and new architecture for the Transformer-based model is designed to adapt for financial prediction. The experiment result reflects that the Transformer-based model only has the limited advantage in absolute price sequence prediction. The LSTM-based models show better and more robust performance on difference sequence prediction, such as price difference and price movement.
    Date: 2023–09
  8. By: Marra de Artiñano, Ignacio; Riottini Depetris, Franco; Volpe Martincus, Christian
    Abstract: Accurately classifying products is essential in international trade. Virtually all countries categorize products into tariff lines using the Harmonized System (HS) nomenclature for both statistical and duty collection purposes. In this paper, we apply and assess several different algorithms to automatically classify products based on text descriptions. To do so, we use agricultural product descriptions from several public agencies, including customs authorities and the United States Department of Agriculture (USDA). We find that while traditional machine learning (ML) models tend to perform well within the dataset in which they were trained, their precision drops dramatically when implemented outside of it. In contrast, large language models (LLMs) such as GPT 3.5 show a consistently good performance across all datasets, with accuracy rates ranging between 60% and 90% depending on HS aggregation levels. Our analysis highlights the valuable role that artificial intelligence (AI) can play in facilitating product classification at scale and, more generally, in enhancing the categorization of unstructured data.
    Keywords: Product Classification;machine learning;Large Language Models;Trade
    JEL: F10 C55 C81 C88
    Date: 2023–07
  9. By: Plantinga, Paul; Shilongo, Kristophina; Mudongo, Oarabile; Umubyeyi, Angelique; Gastrow, Michael; Razzano, Gabriella
    Abstract: Several African countries are developing artificial intelligence (AI) strategies and ethics frameworks with the goal of accelerating responsible AI development and adoption. However, many of these governance actions are emerging without consideration for their suitability to local contexts, including whether the proposed policies are feasible to implement and what their impact may be on regulatory outcomes. In response, we suggest that there is a need for more explicit policy learning, by looking at existing governance capabilities and experiences related to algorithms, automation, data and digital technology in other countries and in adjacent sectors. From such learning it will be possible to identify where existing capabilities may be adapted or strengthened to address current AI-related opportunities and risks. This paper explores the potential for learning by analysing existing policy and legislation in twelve African countries across three main areas: strategy and multi-stakeholder engagement, human dignity and autonomy, and sector-specific governance. The findings point to a variety of existing capabilities that could be relevant to responsible AI; from existing model management procedures used in banking and air quality assessment, to efforts aimed at enhancing public sector skills and transparency around public-private partnerships, and the way in which existing electronic transactions legislation addresses accountability and human oversight. All of these point to the benefit of wider engagement on how existing governance mechanisms are working, and on where AI-specific adjustments or new instruments may be needed.
    Date: 2023–09–26
  10. By: Kristina McElheran; J. Frank Li; Erik Brynjolfsson; Zachary Krof; Emin Dinlersoz; Lucia Foster; Nikolas Zolas
    Abstract: We study the early adoption and diffusion of five AI-related technologies (automated-guided vehicles, machine learning, machine vision, natural language processing, and voice recognition) as documented in the 2018 Annual Business Survey of 850, 000 firms across the United States. We find that fewer than 6% of firms used any of the AI-related technologies we measure, though most very large firms reported at least some AI use. Weighted by employment, average adoption was just over 18%. Among dynamic young firms, AI use was highest alongside more-educated, more-experienced, and younger owners, including owners motivated by bringing new ideas to market or helping the community. AI adoption was also more common in startups displaying indicators of high-growth entrepreneurship, such as venture capital funding, recent innovation, and growth-oriented business strategies. Adoption was far from evenly spread across America: a handful of “superstar” cities and emerging technology hubs led startups’ use of AI. These patterns of early AI use foreshadow economic and social impacts far beyond its limited initial diffusion, with the possibility of a growing “AI divide” if early patterns persist.
    Date: 2023–09
  11. By: IKEUCHI Kenta; INUI Tomohiko; KIM YoungGak
    Abstract: The use of Artificial Intelligence (AI) in business has been expanding in recent years, and there is growing interest in the mechanisms and extent to which AI affects firm performance. In this study, we analyze the impact of firms’ introduction of AI on their performance using the "Basic Survey of Japanese Business Structure and Activities" of the Ministry of Economy, Trade and Industry (METI), Japan, TSR's business-to-business transaction data, press release data from NIKKEI, and IIP patent database 2020. In addition to the introduction of AI-related patents generated by a company's own R&D activity, we also try to analyze the impact of the introduction of AI within their trading partners (suppliers and customers). In addition to the efficiency gains in production processes (process innovation), the study analyzes the creation of new products and improvements in existing products (product innovation). The main results from the analyses are as follows. (1) AI-related patents positively correlate with firm productivity and have a stronger relationship with productivity than non-AI patents. (2) The relationship between AI-related patents and firm productivity strengthened even after 2009 when the number of patent applications began to decline. (3) AI-related patents mainly contribute to the productivity of firms with productivity in middle or higher status within the industry, while AI-related patents have a negative impact on the productivity of firms with low productivity. (4) We cannot confirm that the introduction of AI by a firm's business partner has a positive or negative spillover effect on the productivity of that firm. (5) AI-related patents are strongly related to product innovation, process innovation, and technological innovation of the firms, and especially high-quality AI-related patents have a mid-term and important impact on innovation.
    Date: 2023–09
  12. By: Martin Beraja; Andrew Kao; David Y. Yang; Noam Yuchtman
    Abstract: We document three facts about the global diffusion of surveillance AI technology, and in particular, the role played by China. First, China has a comparative advantage in this technology. It is substantially more likely to export surveillance AI than other countries, and particularly so as compared to other frontier technologies. Second, autocracies and weak democracies are more likely to import surveillance AI from China. This bias is not observed in AI imports from the US or in imports of other frontier technologies from China. Third, autocracies and weak democracies are especially more likely to import China’s surveillance AI in years of domestic unrest. Such imports coincide with declines in domestic institutional quality more broadly. To the extent that China may be exporting its surveillance state via trade in AI, this can enhance and beget more autocracies abroad. This possibility challenges the view that economic integration is necessarily associated with the diffusion of liberal institutions.
    JEL: E0 L5 L81 O30 P0
    Date: 2023–09

This nep-ain issue is ©2023 by Ben Greiner. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.