Funded Research

Stephanie Creary, Assistant Professor, Management

Recently, Wharton AI and Analytics Initiative funded a project to study “Reducing Gender and Racial Disparities in the Use of Generative AI in the Workplace.” While this research intends to enhance both academic research and practice over the longer-term, there is an immediate need to ensure that the “digital divide” in AI adoption does not manifest (or widen) among students enrolled in the courses, MGMT 2240 and MGMT 6240, Leading Diversity in Organizations. Thus, this project involves Education Innovation which aims to complement the aforementioned research study. Specifically, the study updates and expands the course materials to include the adoption of ChatGPT in the main course project and in-class activities.

Barbara Mellers, Professor, Psychology and Marketing
Philip Tetlock, Professor, Management

Recent research on the forecasting abilities of Large Language Models (LLMs) has shown great promise. Studies have focused on general capability evaluation, human-AI interactions and aggregation methods involving multiple models. A notable gap is the systematic exploration of theory-grounded prompting to enhance LLM reasoning in forecasting tasks. The proposed project aims to fill this gap by conducting a comprehensive study of prompting strategies that are grounded in the vast literature on best forecasting practices. We will test a comprehensive set of prompting techniques against a control condition, a naïve baseline, and the human benchmark. By systematically testing and comparing a wide range of prompting techniques, we aim to identify methods that can improve LLM reasoning and forecasting accuracy in a gradually increasing range of domains.

Wharton People Analytics

Wharton People Analytics is developing the Upskilling Lab, a new initiative designed to address the global need for workforce planning and development in the age of AI. As business leaders grapple with the opportunities that AI presents, they are also concerned with the way that AI will impact the workforce. What new jobs will emerge? What jobs will go by the wayside? How will jobs change? And how can employees prepare for these changes to build a prepared, sustainable, and more equitable workforce? To answer these questions, organizations must be able to assess the skills necessary for organizational effectiveness and success in the age of AI. They must also develop evidence-informed processes to upskill and reskill their employees.

Jonah Berger, Associate Professor, Marketing

From salespeople and entrepreneurs to politicians and managers, speakers often communicate through their hands. But despite how integral these movements are to communication, little work has examined their effects. Might how speakers move their hands while speaking shape communication’s impact? And if so, what types of movements are more impactful and why? A multimethod investigation, including automated video analysis of thousands of presentations, application of a large multimodal model on almost 200,000 video segments, and a preregistered controlled experiment, explore how different types of hand movement shape impact.

Leandro Pongeluppe, Assistant Professor, Management

One of the most pressing challenges in the provision of healthcare for individuals living with HIV/AIDS is predicting which patients are at higher risk of lapsing from or discontinuing antiretroviral therapy (ART). This challenge is particularly pronounced in resource-limited settings, where healthcare infrastructure and access to continuous care are often strained. This study seeks to address this issue by comparing various artificial intelligence (AI) and machine learning (ML) models to determine the patient characteristics most strongly associated with ART default and HIV/AIDS mortality, based on data collected at the time of patient enrollment. By identifying the specific factors that increase the likelihood of treatment default or death, this research aims to inform and enhance clinical protocols, enabling healthcare providers to more accurately predict at-risk patients and intervene proactively to keep them in care. The analysis is based on de-identified, individual-level data from healthcare clinics in Malawi, offering valuable insights into the challenges faced in similar developing contexts and paving the way for more targeted, data-driven interventions.

Manav Raj, Assistant Professor, Management

This research project seeks to understand: a) how investing in AI human capital (i.e., individuals with AI-related skills) affects the characteristics of firms’ innovation and competitive advantage; and b) how securing intellectual property, through patents, affects firms’ innovation and advantage.

Prasanna Tambe, Associate Professor, Operations, Information and Decisions
David Hsu, Professor, Management

We research the management of multiple agentic LLMs (large language models) such as open-source Llama or GPT (Generative Pre-trained Transformer) models for effective problem-solving when used collaboratively. Specialized LLM systems, each focusing on different domains, may offer more accurate, creative, and more balanced solutions than single-agent models. Such an approach could have significant applications in industries where (human) experts from various backgrounds collaborate, such as healthcare—where diverse care teams are needed for complex diseases like autoimmune conditions and cancer—and software development, where large projects often require the coordination of designers, front-end and back-end developers, database engineers, and data scientists.

Natalie Carlson, Assistant Professor, Management

With a GDP per capita 45 times smaller than that of the United States, Zimbabwe is a nation of entrepreneurs by necessity (Mazikana, 2023), and its business context is vastly underrepresented in the training data of foundational models compared to entrepreneurs in wealthy Anglophone countries. We conduct an entrepreneurial workshop with 750 participants, where entrants are randomized to search and refine their entrepreneurial business idea with either a peer or a generative AI tool. To evaluate whether investing in local data may help overcome potential limitations in the model’s knowledge base, we test two versions of the genAI tool: one standard “global” model and another fine-tuned to be more localized by providing additional codified knowledge about the local environment and entrepreneurial opportunities. We document how the ideas received from the AI tools shift the entrepreneurial search process relative to the baseline: do they find the ideas feasible? High-quality? Do they change which business idea they ultimately decide to pursue? We also follow up with entrepreneurs three months after the completion of the program to examine outcomes in practice: what percentage of entrepreneurs initiated business start-up, and whether there were any differences in business performance.

Lorin Hitt,Professor of Operations, Information and Decisions
Lynn Wu, Associate Professor of Operations, Information and Decisions

This project investigates the extent to which AI and the construction of analytics datasets exposes a firm to a greater risk of data breaches. We are studying this using a combination of available public domain data and dark web intelligence tools. The goal is to understand the prevalence and determinants of AI-related data breaches and to study how the value (to bad actors) of these datasets evolve over time and compare to data that is commonly transacted on the dark web (e.g. credit card numbers or other personally identifying information (PII)). By using dark web offer (and perhaps transaction) prices and how these evolve over time, we hope to assess the incentives for bad actors to target these types of data, and the extent to which breached data may create ongoing exposure to the targeted firm or, in the case of PII, the individuals described by the data.

Neha Sharma, Assistant Professor, Operations, Information and Decisions
Sophie Yu, Assistant Professor, Operations, Information and Decisions

The rise of LLMs and AI have made it extremely easy for people to code in various programming languages. As a result, StackOverflow, one of the largest online Q&A community for coders and developers has seen a significant decline in its traffic and user participation metrics. However, despite these events, the community still has about 1 million active users per month. In this work, we try to understand why?

Winston Dou, Assistant Professor, Finance
Itay Goldstein, Professor, Finance

This project will investigate, theoretically, numerically, and empirically, the following three related research questions with important regulatory implications: Whether AI-powered trading algorithms will learn and interact in ways that autonomously achieve coordination, leading to the creation of asset price bubbles, and subsequently exploit these bubbles to extract trading profits at the expense of others. Whether AI-driven bubbles trigger severe market instability, with AI-powered trading algorithms executing coordinated exits that amplify the disruption. Whether market manipulation laws need reform and what new regulatory policies and tools should be introduced to prevent AI-driven bubbles and safeguard financial market stability.

Marius Guenzel, Assistant Professor of Finance
Shimon Kogan, Adjunct Associate Professor, Finance

This project seeks to develop machine learning (ML) algorithms that extract both facial features and personality traits from facial images. By leveraging publicly available LinkedIn images and data from our surveys, we aim to create models that will help explain the link between personality characteristics and important life and career outcomes, such as job compensation and seniority, credit market behaviors, and team performance. These algorithms will be applied across various fields, including labor economics, corporate finance, and household finance. This work will support the development of general facial feature extraction models, personality extraction models, as well as their application in large-scale analyses of individual decisions and market outcomes. This research will offer novel insights into the role of facial and personality traits in shaping professional and financial success.

Dean Knox, Assistant Professor, Operations, Information, and Decisions

The proposed research develops and demonstrates new techniques for incorporating novel data sources such as red-light cameras, traffic sensors, mobile location data, and body-worn camera footage into the management of police organizations. These new sources of information are fused with other police datasets, such as dispatch records and daily shift assignments, to provide external checks on officer self-reports. The goal of the project is to demonstrate how these novel data sources make it possible to conduct improved analyses that address three important policy questions in police oversight: how often officers deviate from their agencies’ use-of-force policy, whether force is used discriminatorily against minority civilians, and if there are “bad apple” officers that use an anomalous amount of force compared to their peers.

Tsai-Hsuan Chung, PhD Candidate, Operations, Information and Decisions
Hamsa Bastani, Associate Professor, Wharton OID (Faculty Advisor)
Osbert Bastani, Assistant Professor, Penn CIS (Faculty Advisor)
Haosen Ge, Data Scientist, Wharton

This project includes two innovative research studies designed to harness the transformative power of artificial intelligence (AI) to enhance healthcare outcomes. In close collaboration with the Somaliland Ministry of Health and Development (MoHD), the Taiwan International Cooperation Development Fund (ICDF), and Penn researchers, we aim to tackle critical healthcare challenges in Somaliland, one of East Africa’s most impoverished regions. Our main goal is to develop effective and safe AI methodologies to improve healthcare accessibility, quality, and efficiency. These projects will deepen our understanding of how AI can be applied in safety-critical scenarios and resource-constrained environments, facilitating healthcare advancement on a global scale.

Stephanie Creary, Assistant Professor of Management

Insights from practice (Charter, 2023) suggest that as more and more organizations are embracing the promise of generative artificial intelligence (GenAI), Black workers and managers are likely to feel the most concerned about AI replacing them in their jobs in the next five years. Further, women are less likely to be using GenAI in their work and to be excited about the prospect of using it. In addition, people who are 55+ are less likely to have used generative AI in their work than individuals 18 to 44. These unfortunate trends suggest that bridging the “digital divide” among workers to support gender, racial, and age equity is of critical importance amid AI adoption in the workplace. Yet, to date, little is known about interventions that can meaningfully accomplish this outcome. Further, it is also unclear what impact, if any, generative AI adoption in workplace practices can have on issues of gender, racial, and age equity in the workplace. Through a multi-method field study on working professionals, my research aims to address these questions.

Edgar Dobriban, Associate Professor of Statistics and Data Science, with secondary appointment in Computer and Information Science, Statistics and Data Science
Ian Barnett, Assistant Professor, Penn Biostatistics (PSOM)
Pratik Chaudhari, Assistant Professor, Penn Electrical and Systems Engineering (SEAS)

Artificial Intelligence (AI) has been making spectacular progress in the past few years, leading to popular consumer products such as intelligent chatbots powered by large language models (e.g., ChatGPT), image generators powered by diffusion models (e.g., Dall-E), and computer vision systems powered by deep learning used in self-driving cars (e.g., Waymo). At the same time, AI has proved to be unreliable and hard to measure, such as in language model hallucinations and unrealistic generated images. Moreover, progress in AI is often measured in ad hoc ways based on performance on various benchmark datasets (e.g., MMLU, BigBench), and it can sometimes be hard to know how much new models improve compared to existing ones and how good they really are. In this project, we aim to develop methods for quantifying AI performance. We will develop rigorous and trustworthy methods, by leveraging established principles from statistical inference such as confidence intervals. Specifically, we will develop methods to quantify improvement in performance of new AI models compared to existing ones, and methods for measuring the trustworthiness of AI models. We will provide theoretical guarantees on the correctness of our methods and evaluate them in a broad range of experiments with state-of-the-art AI models including language and diffusion models.

Bethany Hsiao, PhD Student, Operations, Information and Decisions
Hamsa Bastani, Professor, OID (Faculty Advisor)
Duncan Watts (Faculty Advisor)

What does “fairness” actually mean? In algorithmic decision-making, different definitions of “fairness” can often lead to contradictory conclusions about whether an algorithm is fair or not. This research seeks to standardize operationalizations of fairness and create a framework that elicits individuals’ preferences in fairness across various domains, including loan and bail decisions. By understanding stakeholder priorities and tradeoffs, we can then use these results to inform the development of fairer policies and improvement of decision-making processes that integrate algorithms.

Judd Kessler, Howard Marks Professor, Business Economics and Public Policy
Flavia Hug, PhD Student, University of Zurich
Stephan Meier, James P. Gorman Professor of Business, Columbia Business School

There are robust and well-documented gender gaps in labor market outcomes (Blau & Khan, 2017; Goldin, 2021). One set of explanations for the gender gaps relates to gender differences in traits. Women have been shown to be less competitive (Niederle & Vesterlund, 2007), less willing to negotiate (Bertrand, 2011), and to talk less favorably about their ability and performance (Exley & Kessler, 2022). This last set or results on self-promotion is particularly troubling, since it is very hard to mitigate — even when men and women are aware of how well they performed on a math and science test, they still describe the same performance differently. These gender gaps are also unlikely to be corrected by the market — even when others are aware of gender gaps, they fail to account for them (Exley & Nielsen, 2024). The emergence of generative AI has allowed for rapid advancements in helping people to communicate (Noy & Zhang, 2023). This project will explore whether the availability of AI exacerbates or mitigates gender gaps in how men and women talk about their ability and performance. It will also test the efficacy of interventions that might help to close the gap.

Xinyu Ma, Doctoral Student, Operations, Information and Decisions
Bowen Lou, Assistant Professor, University of Southern California
Lynn Wu, Associate Professor, The Wharton School (Faculty Advisor)

Although it is well documented that disruptive events, such as an inventor’s premature death, cause a large and persistent decline in their co-inventor’s innovation performance, strategies to mitigate these negative effects remain unexplored. This project aims to understand how an inventor’s skill and proficiency in artificial intelligence (AI) could counteract the productivity decline resulting from a co-inventor’s demise. We intend to first identify the specific challenges facing inventors in staying innovative after the unexpected death of their close collaborators. Furthermore, we will examine how AI can help inventors overcome these challenges. Although our research setting is premature deaths of close collaborators, our study will offer boarder insights inventors on making strategic human capital investment amidst various uncertain and turbulent events. It will also provide a roadmap for firms in optimally allocating AI resources to foster a resilient ecosystem for innovation.

Gideon Nave, Carlos and Rosa de la Cruz Associate Professor of Marketing
Steven Shaw, Postdoctoral Researcher, Wharton Marketing

In recent years, researchers have made significant advances to improve the replicability of research findings in the behavioral and social sciences. While replicability initiatives focus on establishing that experimental outcomes represent true scientific effects, the generalizability of these effects often remains limited to the stimuli directly used in the experiment (and these same stimuli are used in replications). Here, we aim to demonstrate the usefulness of AI tools for evaluating the generalizability of replicable scientific findings in the behavioral sciences. For example, using prompts that emulate information from replicable studies, Large Language Models such as ChatGPT can help generate a wide range of stimuli that can be used to test the generalizability of such studies (to the extent that each participant views unique stimuli for a study). This diversity ensures that effect sizes observed are not limited to idiosyncratic aspects of the few options that researchers came up with or pretested, reducing the risk of bias tied to a narrow set of stimuli.

Alina Song, PhD student, Finance
Joao Gomes (Faculty Advisor)

This research project aims to investigate the impact of artificial intelligence (AI) investments by non-tech firms (AI-adopters) on product market competition, innovation, and industry dynamics. By leveraging novel datasets, including firm-level AI-workers, product-level pricing and quantities, and textual information from company filings, this study seeks to provide new insights into how AI adoption shapes product differentiation, firm entry and exit, and the intensity of competition within industries. By providing empirical evidence on these issues, this research project aims to inform policy discussions and contribute to shaping regulatory frameworks that balance fostering innovation with protecting consumer interests in a rapidly growing AI economy.

Weijie Su, Associate Professor, Statistics and Data Science

Large language models (LLMs) have shown immense potential in revolutionizing business applications and healthcare by providing efficient access to personalized information, supporting decision-making processes, and synthesizing vast amounts of relevant literature. These powerful AI tools can bridge the knowledge gap between healthcare professionals and underserved communities, offer preliminary advice, aid in decision support, and streamline administrative tasks. Additionally, LLMs can enhance research efforts by identifying gaps and suggesting future directions. However, the rapid adoption of LLMs in healthcare and business settings has revealed significant challenges that limit their use in high-stakes applications, such as bias, hallucinations, misinformation, and overconfidence. This project aims to address these issues by developing robust statistical methods for accountable and responsible LLM applications in healthcare and business. The objectives include developing methods to address bias, enhance calibration, reduce hallucinations, create watermarking techniques for authentication, and develop reliable methods for uncertainty quantification. The proposed methods will be assessed and validated using structured and unstructured electronic health records data and real data from other sources. By fostering interdisciplinary collaboration, prioritizing ethical considerations, and leveraging technology for the advancement of human health and business efficiency, this proposal seeks to mitigate potential risks and harms associated with adapting and applying LLMs to high-stakes decision-making, ultimately strengthening the rigor and reproducibility of research in these domains.

Duncan Watts, Stevens University Professor and twenty-third Penn Integrates Knowledge University Professor at the University of Pennsylvania, Operations, Information and Decisions
Mark Whiting, Senior Scientist, OID & CIS

Machines with common sense have been a long-term goal of the AI community, yet common sense has tended to elude formal study. While countless efforts have been made to generate corpora of common sense ideas, they seemingly all hinge on the assumption that common sense is simply a body of obvious truths, as opposed to a more complex reality that we have found in our prior work (paper attached in application), that common sense is in fact not particularly common at all. In this work, we empirically and systematically measure common sense in human subjects and large language models. We extend our earlier work from a demonstration of quantifying common sense to an extensible and general-purpose infrastructure for evaluating common sense and related systematic questions with world-scale samples. Our work includes deploying this infrastructure and acquiring samples to produce the largest mega-study on common sense across types of knowledge, disparate populations, social connections, and a world of languages.

Shun Yiu, PhD Candidate, Management
Matthew Bidwell (Faculty Advisor)

Algorithms play an increasingly important role in today’s digital economy. An emerging view highlighted the viability of utilizing algorithms as a governance device. With innovations in blockchain infrastructure and smart contracts algorithms, platforms can delegate formal authority to exchange partners and involve them in organizational decision-making processes. While some are optimistic that this governance innovation can bring about a more collaborative and democratic digital economy, this paper highlights potential challenges associated with decentralized governance systems. I argue that decentralized governance shifts the locus of opportunism from platform owners to a body of diffuse exchange partners. Appropriation no longer come from platform owners only, but also existing and potential exchange partners who have or could acquire formal authority. This introduces heterogeneity in exchange partners’ exposure to appropriation hazard, leading to selective disengagement from platform participation. I examine these arguments with the staggered adoption of decentralized governance structures among a large sample of exchange platforms in the decentralized finance industry. I analyze depositors’ response to policy change from exchange platforms and find that the adoption of decentralized governance has a negative effect on participation among depositors. This negative effect is primarily driven by depositors with intermediate levels of investments. As a result of the overall decline in participation and hollowing out dynamics, participation became more concentrated following the adoption of decentralized governance. These results highlight the perhaps surprising centralization tendency of a decentralized governance system.

Gad Allon, Jeffrey A. Keswin Professor, Professor of Operations, Information and Decisions

This project aims to systematically study the learning processes among gig economy drivers. By understanding how drivers adapt and improve over time, we can gain valuable insights into the workings of the gig economy, paving the way for more efficient and driver-friendly practices. This research holds the potential to inform platform design, policy making, and the support structures necessary for the well-being and success of gig economy workers.

Henrique Laurino dos Santos, PhD Candidate, Marketing

Despite the magnitudes of business budgets assigned to digital marketing channels, it remains challenging to benchmark the effectiveness of actions in that space (McMahan et al. 2013, Blake et al. 2015, Dai et al. 2023). A major barrier to better web traffic analytics is that website traffic data – or “clickstream” – present multiple undesirable properties for traditional marketing models. Datapoints are both clumpy at high frequencies and sparse over long horizons, underpinned by a myriad of state-dependent mechanisms, and often truncated for multiple intervals. Compounding those are a variety of new data protection laws around the globe (e.g. the European GDPR and the Brazilian LGPD) which impose strong restrictions on data access for advertisers. Multiple studies have argued that those restrictions may lead to undesirable outcomes for consumer welfare, such as weakening the reach of smaller digital retailers or inducing excessive advertising from major retailers (Gal and Aviv 2020, Ke and Sudhir 2023). In this project, the researcher attempts to mitigate those effects by directly modelling undesirable properties of the clickstream without commonly used techniques which violate users’ privacy.

Shiri Melumad, Assistant Professor, Marketing
Jin Ho Yun, Post Doc, Wharton Neuroscience Initiative

The purpose of this research is to empirically explore the effects of using LLMs such as ChatGPT versus traditional search engines (e.g., Google search) on consumers. The central thesis of this work is that, as tools for knowledge acquisition, LLMs act as a double-edged sword: specifically, the very trait that makes them potentially appealing to users–their ability to aggregate and synthesize information–may also lead users to feel diminished ownership over the inferences they draw from the information provided. This reduced sense of ownership, in turn, is predicted to reduce users’ willingness to act on the information provided by LLMs compared to information gained through traditional search engines like Google–which typically require users to sort through search results and synthesize the information themselves. The research will also explore how consumers use the two tools in tandem, and how this may affect their confidence in the knowledge they acquire.

Leon Musolff, Assistant Professor, Business Economics and Public Policy 

(1) The Artificial Salesman: Ad Copy by Generative AI
What is the effect of Generative AI on the $280 billion search advertising industry, which currently relies on the work of professional copywriters to come up with titles and short excerpts that convince users to visit an advertiser’s website? In this project, researchers analyze two experiments at a major search engine to study the effect of machine-generated ad copy.

(2) The Productivity Effects of Generative AI: Evidence from a Field Experiment with GitHub Copilot
This research studies the impact of Generative AI tools on software development by analyzing two large-scale randomized controlled trials in a real-world environment.

(3) Privacy & Advertising on Mobile Phones
Since Apple introduced App-Tracking Transparency (ATT) in 2021, mobile app developers have been required to ask users for permission to track them across apps and websites owned by other companies. This requirement has led to a significant reduction in the amount of data available to advertisers and, hence, the effectiveness of targeted advertising. How large are these losses? And to what extent does the reduction in data availability lead to a concentration of advertising platforms? To answer these questions, the researchers plan to obtain extensive bidding data from a major advertising SDK used by mobile app developers. They will then use this data to estimate the causal effect of ATT on advertising revenue and concentration.

Serguei Netessine, Senior Vice Dean, Professor of Operations, Information and Decisions 

While there is a plethora of studies for consumer transportation platforms (e.g., Uber, Lyft), relatively little is known about challenges faced by freight platforms. Recent bankruptcy of the largest such platform in US (Convoy) has been in the news. We provide first study of trucking platform efficiency using detailed data from the larges Chinese platform and we demonstrate unique challenges faced by platform intermediaries.

R. Jisung Park, Assistant Professor, Business Economics and Public Policy 

In this project, the researchers consider the case of workplace heat safety standards in the state of California. The number of days with potentially damaging heat is increasing rapidly and will continue to increase even with aggressive greenhouse gas mitigation.

Ruben Ramirez Salas, Phd Candidate, Operations, Information and Decisions 

This project examines the impact of Artificial Intelligence (AI) and Large Language Models (LLMs) on creative industries, highlighting their potential to revolutionize these sectors. The researchers investigate the dual nature of generative AI’s influence: its capacity to enhance human creativity and its challenges, including threats to human-centered views and market acceptance. Focusing on the U.S. humor market, we analyze AI’s potential to disrupt traditionally human-dominated creative spaces.

Vishrut Rana, PhD Candidate, Operations, Information and Decisions 

This project develops a structural model for electricity price formulation using historical price, demand, renewable generation, and transmission data from various nodes in Texas. Using this model, the researchers estimate revenues from arbitrage and predict future revenues under changing market conditions. Results from theirmodel can help battery operators identify optimal locations for storage investments and help system operators study market performance under different demand and renewable generation growth scenarios in the presence of storage technologies.

Alina Song, PhD Candidate, Finance

Trade credit (firms selling goods to customers on credit rather than requiring immediate cash payment, empirically measured as account receivables) is the most important form of short-term financing, yet its variation across firm and over time is not well-understood. Moreover, it is not clear how trade credit interacts with a firm’s financing from banks (especially credit lines). Existing research faces constraints due to the absence of granular data delineating customer-specific trade credit and the detailed allocation of bank credit lines. This research seeks to bridge this gap by employing advanced AI methodologies, specifically the ChatGPT API, to extract and analyze novel datasets from corporate filings, such as Form 10-Ks. The objective is to construct a detailed panel dataset that elucidates trade credit disaggregated by customer and credit lines segmented by lending banks. This research allows us to better understand supply chain fragility and inform policymakers on how to regulate payment delays.

Hummy Song, Assistant Professor of Operations, Information and Decisions
Christian Terwiesch, Andrew M. Heller Professor at the Wharton School; Professor of Operations, Information and Decisions; Professor of Health Policy, Perelman School of Medicine; Co-Director, Mack Institute of Innovation Management Department Chair
Xufei Liu, PhD Candidate, Operations, Information and Decisions

Since COVID-19, telehealth has risen in popularity, especially with respect to mental and behavioral health services. Telehealth therapy provides access to these services for clients who prefer to stay in the comfort and safety of their own home while receiving access to care. However, as the demand for telehealth therapy rises, so does the workload of practicing therapists. Their shift does not end when client sessions are over – rather, they must create an assortment of documentation ranging from diagnostic assessments, treatment plans, severity scores, etc. for every client. These unpaid hours can increase therapist burnout and limit their capacity for taking on new patients.

Recent developments in large language models (LLMs; e.g., OpenAI’s ChatGPT) provide new possibilities for improving productivity and decreasing workload for telehealth providers. Allowing LLMs to pre-generate the required documentation as a draft for therapists can significantly reduce the work needed after client sessions. In this project, researchers measure and quantify the accuracy of LLM-created documentation, potential time-savings that occur with this approach, and whether LLM-generated documentation can be of comparable quality to therapist-created documentation.

Lynn Wu, Associate Professor of Operations, Information and Decisions

At the core of the innovation ecosystem is the dedicated innovation teams. These teams not only generate creative ideas but also transforms embryonic thoughts into tangible patents through rigorous experimentation and prototyping. The significance of team-based collaboration in both innovation and scientific inquiries has been on the rise, becoming increasingly crucial for enhancing overall team effectiveness. Recent advances in machine learning have made promising breakthroughs in spurring innovation prompting numerous organizations to invest in AI to maintain a competitive edge in innovation. This research investigates the extent to which AI can either mitigate or potentially exacerbate risks in team-based innovation.

Winston Dou, Assistant Professor, Finance
Itay Goldstein, Joel S. Ehrenkranz Family Professor, Finance

The integration of algorithmic trading and reinforcement-learning (RL) algorithms, commonly known as AI-powered trading, has the potential to reshape capital markets fundamentally and presents new regulatory challenges. Policymakers, regulators, and financial market supervisors worldwide have recognized AI as a regulatory priority, directing their attention to how AI techniques are applied in financial markets to comprehend the associated implications and assess potential systemic risks. This paper aims to analyze the behavior of AI-powered trading algorithms that possess private information, investigating the significant effects they have on the market power of informed AI traders, the market liqudity of capital markets, and the informativeness of asset prices. The findings of this research unveil that the phenomenon of AI collusion can sporadically surface across a diverse array of market structures and informational environments. Importantly, this study discovers that the emergence of AI collusion can be attributed to two distinct sources or mechanisms.

Marissa King, Alice Y. Hung President’s Distinguished Professor, Healthcare Management

More than 40 million adults and adolescents in the US need treatment for substance use disorders. Yet, only 6.5% of those needing treatment received it. Moreover, far too little is known about what treatments are being delivered and how patterns of treatment vary by gender, race, and insurance status. This study will examine differences in modalities of treatment for substance use disorder by race, gender, and insurance status.

Sophia Pink, PhD Candidate, OIDD
Jose Cervantez, PhD Candidate, OIDD

Does telling women about gender differences in willingness to compete increase their likelihood of entering a competition? This research conducts a field experiment on an executive job search platform where researchers find that telling women about the gender gap in willingness to compete causes them to apply to over 20% more leadership roles. The researchers also plan to conduct incentive-compatible lab experiments conceptually replicate the effect and test for psychological mechanisms. This work contributes to theory on stereotype reactance, and has practical implications for closing a policy-relevant gender gap.

Alp Sungu, Assistant Professor, OIDD
Hamsa Bastani, Associate Professor, OIDD

The widespread adoption of ChatGPT has given rise to concerns about the future of education. Numerous educational institutions are employing guardrails against ChatGPT-aided cheating to ensure authentic learning; at the same time, this technology holds the promise to democratize education by providing personalized tutors to guide students to better educational outcomes, particularly in areas where access to high-quality teachers is limited. In collaboration with a large high school and several technology-embracing high school teachers, the researchers are piloting an active learning approach in the classroom, where students are guided by a custom mentoring interface powered by GPT-4. The researchers are conducting a randomized controlled trial, where classrooms are randomly assigned to one of three treatment arms, in order to rigorously evaluate students’ educational outcomes in the field.

Shun Yiu, PhD Candidate, Management

This project aims to study how recent advances in generative AI are affecting the skill and task content of jobs and the strategic positioning of workers (i.e., how they describe themselves, the skills they develop, or the jobs they seek out) in response to those changes. Despite widespread interest in how generative AI affects jobs and careers, there has been little research that addresses this question. One reason is that it is difficult to track the job demands and career trajectories for a large number of jobs and workers who are differentially exposed to generative AI. With proprietary data from Upwork Inc., the researchers are able to access full information on a large sample of jobs and workers over time, thus obtaining findings that can inform the effects of generative AI on the labor market and careers.

Linda Zhao, Professor, Statistics and Data Science Department

This project aims to address the joint problem of product assortment and pricing optimization by developing a new machine learning model and algorithm. The project will result in two key deliverables: research articles and software.

Anna Helmke, Doctoral Candidate of Finance

What is the optimal disclosure policy for swing pricing by open-end mutual funds? Specifically, which disclosure rule, with respect to the swing factor and – thresholds, minimizes run risk in open-end mutual funds? Does the disclosure policy rule that is optimal from a financial stability perspective also preclude the possibility of front running by fund investors? Answering this question will likely also require a model of the optimal swing factor. By addressing this critical gap in the literature, this research aims to contribute to inform policy decisions regarding the proposed SEC rule.

Joseph Simmons Dorothy Silberberg Professor of Applied Statistics Professor of Operations, Information, and Decisions, Beidi Hu, Doctoral Candidate of Operations, Information and Decisions

The proposed research focuses on how people use and perceive innovative decision aids, for example, AI tools like ChatGPT. Individuals and organizations are increasingly outsourcing part of their labor or decision-making process to these tools. The proposed research aims to investigate the perception around using AI decision aids, both in terms of its impact on individuals’ metacognitive assessments and observers’ social judgments.

Shun Yiu, Doctoral Candidate of Management

This project studies how algorithmic governance affect economic coordination in the context of blockchain-based organizations. Specifically, the project studies the algorithmic governance structures adopted among decentralized exchange protocols deployed on blockchains. The project documents and develops frameworks to understand variation in the kinds of structures adopted. In addition, the project examines how the adoption of different kinds of governance structures affect the decentralized coordination among individuals.

Xinyu Ma, Doctoral Candidate of Operations, Information and Decisions

In this project, the researchers aim to investigate the extent to which acquiring AI capabilities can ease firms’ transition of innovation directions in uncertain times of leadership change. To achieve this goal, they plan to identify the challenges that firms face when they attempt to pivot the innovation direction during CEO turnover. Furthermore, they plan to examine how AI can help mitigate these challenges and provide new opportunities for certain types of innovation development. This study will provide a roadmap for firms to make optimal decisions when allocating AI resources for strategic changes in innovation.

Gad Allon, Jeffrey A. Keswin Professor, Professor of Operations, Information and Decisions

This project proposes a data-driven exploration of the main factors behind basket abandonment in collaboration with an online e-grocer. The study aims to identify the key factors that contribute to cart abandonment and provide insights into how online e-grocers can reduce abandonment rates and improve customer satisfaction.

Cait Lamberton, Alberto I. Duran Presidential Distinguished Professor of Marketing, Dave Reibstein, Professor of Marketing

The Consumer Cryptoconfidence Index offers the first ongoing, consumer-centered data source specifically intended to capture sentiment related to this market exchange tool. Offering both a novel “snapshot” understanding of the typical consumer’s perception of and interaction with cryptocurrency as well as data that may be fruitfully connected to other marketplace changes, the C3i index will allow us to build our knowledge as the cryptocurrency market evolves. By developing this understanding, we hope to support decision-making among consumers, regulators, investors, and entrepreneurs, while also laying the groundwork for ongoing consumer-centered research in this domain.

Hamsa Bastani, Assistant Professor of Operations, Information and Decisions, Angel Chung, Doctoral Candidate of Operations, Information and Decisions

This proposal aims to develop and deploy novel methods that integrate machine learning and optimization to improve the efficiency and equity of global health supply chains in underserved communities. Current effort is targeted specifically towards distributing a limited budget of various essential medicines to 1000+ health facilities throughout Sierra Leone; we are in the process of piloting our approach. This work is in close partnership with the Sierra Leone National Medical Supplies Agency, the Sierra Leone Ministry of Health and Sanitation, an AI health startup Macro-Eyes, and Penn researchers.

Allison Nicoletti, Assistant Professor of Accounting, Christina Zhu, Assistant Professor of Accounting

The recent failure of Silicon Valley Bank (SVB) and subsequent volatility in seemingly healthy institutions raises concerns that social media has changed the nature of deposit markets and depositors’ typical responses to information. We provide insight into this concern by examining whether social media activity about banks, and in particular, activity that is does not directly relate to bank fundamentals, is associated with banks’ deposit growth. We also examine whether this association differs for insured and uninsured depositors, who differ on dimensions such as information processing ability and incentives to monitor. To the extent that deposit growth is related to social media information in a broad sample of banks, the results of this study have implications for banking system stability and how regulators supervise banks.

Leandro Pongeluppe, Assistant Professor of Management

The primary objective of this research project is to develop a toolkit that can identify greenwashing practices in the sustainability reports of US Fortune 500 firms. The project will be achieved through three dimensions: a textual analysis of environmental terms, collocation of these terms in sentences with emotional (sentiment) content, and visual analysis of the reports (count of green pixels). The secondary objective is to compare the firms’ disclosure with their ESG scores and penalties to identify the degree of greenwashing.

Bingxin Zhao, Assistant Professor of Statistics and Data Science, Weijie Su, Associate Professor of Statistics and Data Science

The fast-growing generative artificial intelligence (AI) industry is experiencing severe legal challenges from copyright holders, as AI models often use copyright-protected materials as training data without authorization. Several recent lawsuits against Midjourney and Stability AI have attracted a lot of attention, and these could only be the beginning of the AI lawsuit storm in 2023. As a result, new algorithms are needed to protect the copyright of training data in generative AI applications. This project focuses on developing algorithms to protect copyright in generative models, which will be essential for addressing the legal complexities of generative AI.

Lynn Wu, Associate Professor of Operations, Information and Decisions, Gavin Wang, Doctoral Candidate of Operations, Information and Decisions

In this study, the researchers plan to examine the business impact of ChatGPT adoption on e-commerce vendors. The research goal is twofold. First, they examine whether ChatGPT can improve sales performance of e-commerce vendors. Second and more importantly, they study who will benefit more from adopting ChatGPT.

Manav Raj, Assistant Professor of Management

The emergence of generative AI technologies, such as OpenAI’s ChatGPT chatbot, has expanded the scope of tasks that AI tools can accomplish and enabled AI-created creative content. In this study, researchers explore how disclosure regarding the use of AI in the creation of creative content affects human evaluation of such content. The researchers discuss the implications of this work and outline planned pathways of research to better understand whether and when AI disclosure may affect the evaluations and reactions to the use of AI in different kinds of task.

Pinar Yildirim, Associate Professor of Marketing

This project aims to estimate how the usage of online dating platforms in the United States has impacted high-level relationship outcomes (e.g., marriage and divorce rates, time to marry and stay married) as well as user mental and physical health. To estimate the causal impact of online dating platform usage on such outcomes, the researchers use variation in the penetration of online dating apps and websites across locations (e.g., counties). The idea is the following: in locations where we see a large increase in the usage of online dating platforms, how do relationship outcomes change in the following year?

Serguei Netessine, Senior Vice Dean, Professor of Operations, Information and Decisions 

The arrival of the gig economy has led to an unprecedented explosion of person-to-person task outsourcing: driving, food pickup, and shopping can all be done by someone other than the consumer. Such outsourcing potentially creates new challenges for gig workers: knowing the most efficient route, determining the entrance to the customer’s home, or knowing where to find the product they are shopping for. Overall, AI improves the effectiveness of gig workers by helping less experienced workers achieve order outcomes that are more comparable to those of more experienced workers, thus increasing both customer satisfaction and revenue per order. However, there are boundary conditions for technology adoption and overuse of technology can even lead to lower productivity.

Ron Berman, Assistant Professor of Marketing, Hangcheng Zhao, Doctoral Candidate of Marketing

The US government recently required healthcare price transparency from hospitals and insurers in an attempt to control spiraling healthcare costs. On the one hand, price transparency and online search tools have been very effective in increasing competition for other products in the e-commerce space. On the other hand, healthcare is a complex product, often bought through insurance, which makes it unclear how much transparency will affect realized prices. Further, public disclosure of prices might allow healthcare providers to better negotiate with insurance companies, which might even lead to an increase in healthcare prices. This project uses a large-scale nationwide dataset of healthcare insurance claims to analyze the potential benefits and impact of these regulations, and inform policymakers and the public about the values and risks of these reforms.

Robert Inman, Richard K. Mellon Professor, Emeritus, Finance Department

The federal government has, over the past decade, allocated nearly $1 trillion dollars to state governments for managing the fiscal consequences of the Great Recession and the Covid pandemic. This project will evaluate the incentives embedded in such assistance for the efficient budgeting of state and local governments. Specifically does such aid create perverse incentives for deficit financing of state government services?

Angela Chen, MD/Doctoral Candidate of Healthcare Management

This project develops a machine learning model to identify low acuity visits to emergency departments (EDs) using large administrative and claims-based datasets. We show that an ML-based, multivariate approach to identifying these visits outperforms traditional International Classification of Disease (ICD) code-based methods and has more consistent, improved performance across demographic subgroups. Results highlight the potential for machine learning models to improve research quality and demonstrate the potential for eventual incorporation of these models in urgent care provision. Such work can help reduce unnecessary resource utilization and improve patient care outcomes and experience, while also demonstrating the importance of considering demographic factors in model development and evaluation.

Shannon Duncan, Doctoral Candidate of Marketing

Can companies providing goal-tracking technology innovate their products to better enable consumers to keep going when they inevitably fail (i.e., skip a workout or overeat calories)? This research aims to fill this gap by testing the effectiveness of one type of goal tracking design: dynamic goal setting. Most traditional goal tracking apps focus on static goals – goals where the consumer aims to complete the same amount of goal behavior each day (i.e., working out 30 minutes a day or walking 10,000 steps a day). However, the researchers suggest behavior tracking apps that use AI technology to vary the amount of goal behavior the consumers engage in from day to day (i.e., work out an average of 30 minutes a day, with some days being 15, some days 45, etc.) will be more effective, a type of goal tracking they term dynamic goal setting.

Susanna Berkouwer, Assistant Professor of Business Economics and Public Policy

Tropical cyclones have tragic effects on human lives and livelihoods, especially for those without the economic means to protect themselves. Climate change is expected to increase the frequency and severity of hurricanes and exacerbate their damages. Climate risk and uncertainty may also deter economic investment and slow economic growth even in the absence of a storm in a given year. We evaluate a novel financial product designed to make vulnerable communities more resilient to climate change: parametric insurance, which is being rolled out by credit unions across the Philippines.

Hummy Song, Assistant Professor of Operations, Information and Decisions, Harriet Jeon, Doctoral Candidate of Healthcare Management 

As hospitals face catastrophic level of waiting for an available inpatient bed, one strategy to improve productivity given fixed resources is specialization. While specialization in health care has primarily been along medical specialties (e.g., radiology or internal medicine), this project explores the impact of specialization in roles and its impact on hospital operations and patient care. The researchers study this question in the context of an “admitter” role, which embeds an internal medicine-trained physician within the Emergency Department to exclusively focus on best sorting patients into appropriate care settings. They use detailed electronic medical records from a large multi-site academic health system to examine this question.

Ulrich Doraszelski, Joseph J. Aresty Professor, Professor of Business Economics and Public Policy, Economics, and Marketing, Jordi Jaumandreu, Senior Academic Researcher, Boston University

Artificial intelligence, machine learning, robots, and automation have fundamentally changed firms’ production processes over the years. Yet, the measurement of productivity traditionally assumes that these new technologies have scaled up existing production processes without substantially affecting how firms combine the various inputs to produce output. This project develops methods for the measurement of productivity that account for these (and other) new technologies with the overarching goal of ensuring that their impact is fully reflected in the aggregate productivity statistics.

Bingxin Zhao, Assistant Professor of Statistics and Data Science

How does the brain of a manager differ from that of other people? Are manager’s brains born or made? In this project, a half-million large-scale biomedical datasets will be analyzed in combination with advanced statistical learning methods to uncover inter-subject variations in brain structure and function related to being a manager. Researchers will analyze the genetic endowment of manager-related brain differences by integrating imaging, genetics, and environmental information, and investigate their links to social activity, lifestyle, mental health, and physical health.

Winston Dou, Assistant Professor of Finance, David Reibstein, William Stewart Woodside Professor of Marketing

Environmental, Social, and Governance (ESG) issues are playing an increasingly important role in firms’ strategic decision-making in financing, investment, marketing, and industry competition. This project aims to provide theoretical insights and empirical evidence on how corporate ESG activities could be motivated by firms’ considerations about product market competition and the potential feedback effect between ESG activities and competitive advantages in the product market.

Leon Musolff, Assistant Professor of Business, Economics, and Public Policy, James Brand, Senior Researcher, Microsoft, Juan Camilo Castillo, Assistant Professor of Economics, Will Wang, Chief Economist, Operate, Unity 

Access to ample computing resources has become a key concern for leading firms across many industries. To investigate this concern, researchers empirically study the market design problem faced by cloud providers, which need to determine how to price and allocate fixed computing capacity across firms with differing needs and volatile demand. This investigation focuses on the prevailing “quality differentiation” strategy and its impact on market outcomes.

Duncan Watts, Stevens University Professor, Xinlan Emily Hu, Ph.D. candidate, Operations, Information and Decisions

Which aspects of a team’s interaction (the “team process”) predict team success? And how might the answer change across different types of teams, tasks, and contexts? This project answers these questions by computationally modeling theories about team processes, then testing the theories head-to-head on a variety of real teams — from political deliberation groups to freelance software engineers. This model will uncover the boundary conditions of what makes an interaction successful, producing both precise social scientific theories for academics and data-backed insights for managers, executives, and team leaders.

Britta Glennon, Assistant Professor of Management, Saerom (Ronnie) Lee, Assistant Professor of Management, Saerom (Ronnie) Lee, Assistant Professor of Management

How do entrepreneurs choose the location of their startups? Other dimensions of startup formation have long received extensive attention from scholars, but location choice—particularly, across national borders—remains under-explored. This project explores questions such as: are entrepreneurs more likely to establish their startups in other countries in response to more restrictive immigration or trade policies in the US? What types of individuals are most or least responsive? Who are the winners and losers?

Ravi B. Parikh, Assistant Professor of Health Policy and Medicine, Ezekiel J. Emanuel, Vice Provost for Global Initiatives

The U.S. government pays Medicare Advantage (MA) insurers a set amount for each person who enrolls, with higher rates paid for patients assigned more co-morbidities via the current risk adjustment methodology. This incentivizes “upcoding”, systematic over-billing by MA insurers that results in at least $10 billion in excessive and unjustified costs each year. Leveraging applied machine learning methods, this “ML-Guided Risk Scoring” project aims to validate a more accurate risk score with wide adoption potential that can reduce gaming and upcoding. The outputs of this research approach will have relevance for policy adoption, leading to fairer payment to providers, incentives to care for medically vulnerable patients and parity between MA and traditional Medicare.

Amy Sepinwall, Associate Professor of Legal Studies and Business Ethics

This project seeks to gain clarity on whether AI can satisfy the requirements of moral agency and how this impacts corporations.

Duncan Watts, Stevens University Professor of Computer and Information Science, Communication, and Operations, Information and Decisions
Linnea Gandhi, Doctoral Candidate, OID

This project seeks to build a map of “nudge” or “choice architecture” interventions, enabling practitioners and academics alike to navigate the theoretical space easily and effectively. The map will be seeded with historical studies and enriched with data from new lab and field experiments to help validate what we, as a field, do and don’t yet “know” about the efficacy of these interventions across contexts.

Susanna Berkouwer, Assistant Professor of Business, Economics and Public Policy
Arthur van Benthem, Associate Professor of Business, Economics and Public Policy

This project builds a research portfolio that gathers large data sets from the U.S. and across the world and uses sophisticated econometric tools to analyze this data with the goal of quantifying the inefficiency and unintended consequences from inefficient regulations, and to propose improved energy policy.

Matthew Bloomfield, Assistant Professor of Accounting
Christina Zhu, Assistant Professor of Accounting

This project seeks to provide novel evidence regarding the link between firms’ public financial disclosures and their product market pricing decisions.

Saerom (Ronnie) Lee, Assistant Professor of Management
Exequiel (Zeke) Hernandez, Associate Professor of Management

Using multiple large-scale datasets on the U.S. labor market, this project will examine the firm-level drivers of hiring immigrant workers.

Aviv Nevo, George A. Weiss and Lydia Bravo Weiss University Professor; Professor of Marketing; Professor of Economics
Yihao Yuan, Doctoral Candidate, Marketing

This project seeks to understand the role that exclusive contracts play in shaping market structure, consumer demand, and innovation. This project will develop a structural model and use data-driven methods to quantify the impact of vertical contracts on consumer welfare and profits of studios and platforms in the video streaming market, a fast-growing market that already accounts for more than a quarter of all time spent on television sets by Americans.

Lindsey Cameron, Assistant Professor of Management

This project is an inductive two-part multi-sourced qualitative study that focuses on the practices and community around the developers that write bots, scripts and automated programs that are designed to override algorithmic controls and how workers use these technologies to resist and counter algorithmic control.

Marius Guenzel, Assistant Professor of Finance

The goal of this project is to empirically study the effect of workplace and economic stress on health outcomes including aging and mortality.

Natalie Carlson, Assistant Professor of Management

This project supports the launch of the initial Wharton/Columbia Management, Analytics, and Data (M.A.D.) Conference. The conference is created to bring together academics working to understand the role of data and analytics in shaping managerial practice and the determinants of firm performance.

Robert P. Inman, Richard King Mellon Professor Emeritus of Finance; Professor Emeritus of Business Economics & Public Policy

This project uses data collection for a student-led evaluation of the health and economic consequences of the Covid-19 pandemic and the effectiveness of national and state-wide policy responses to contain the coronavirus and to mitigate its health and economic consequences. The format for this evaluation will be a Capstone Course (BEPP 401) entitled, Federal and State Management of the Pandemic: Money, Messages, Vaccinations, and State Policies.

Dean Knox, Assistant Professor of Operations, Information, and Decisions
Rachel Mariman, Research Project Manager 

The goal of this project is to create a tool to automate causal inference from incomplete or imperfect data. This tool will reach a broad audience of applied researchers across the social and medical sciences by developing an easy-to-use front-end interface and implement more efficient back-end optimizations. In addition, the project will create a series of data applications to illustrate its ease of use. This project is funded by AI for Business.

Benjamin Lockwood, Assistant Professor of Business Economics and Public Policy
Santosh Anagol, Associate Professor of Business, Economics and Public Policy

This project seeks to characterize and advertise a new statistical method for estimating the elasticity of taxable income in the presence of optimization frictions.

Philip Tetlock, Professor of Management
Dillon Bowen, Doctoral Candidate of Operations, Information, and Decisions

This project will compare the effectiveness of many judgmental forecasting techniques on a standard set of forecasting tasks, delivering a concise list of highly effective techniques and recommended best practices.

Raghu Iyengar, Miers-Busch, W’1885 Professor; Professor of Marketing; Faculty Director, Wharton Customer Analytics
Nicole Wang Trexler, Associate Director of Data Science and Research

This project conducts a quasi-experimental research study to gain an in-depth look at how an experiential learning pathway impacts students’ informal learning journey. This information will help WCA design and deploy better experiential learning pathways to better serve the students at Wharton and within the Penn community.

Duncan Watts, Stevens University Professor of Computer and Information Science, Communication, and Operations, Information and Decisions
Baird Howland, PhD Student, Annenberg School of Communication, Computational Social Science Lab at Penn,
Valery Yakubovich, Executive Director, Computational Social Science Lab at Penn

The goal of this project is to study our understanding of the COVID-19 pandemic – the remarkably varied conceptions of what is happening, why it is happening, and what should be done in response, with a rare combination of quantitative rigor and qualitative depth.

Hamsa Bastani, Assistant Professor of Operations, Information, and Decisions
Pia Ramchandani, Doctoral Candidate of Operations, Information, and Decisions

In collaboration with the Tellfinder Alliance for Global Counter-Human Trafficking, this project will leverage unstructured, massive deep web data from leading adult-services websites using a novel machine learning framework to construct the first global network view of sex trafficking supply chains.

Dan Taylor, Associate Professor of Accounting

This five-part case series will highlight recent accounting frauds (e.g., Wirecard, Luckin Coffee, etc.) and how each of the frauds could have been detected using business analytics.

Lorin Hitt, Zhang Jindong Professor; Professor of Operations, Information and Decisions
Lynn Wu, Associate Professor of Operations, Information and Decisions

This research explores how AI facilitates innovation by documenting specific cases and mechanisms on when AI technologies should be used to innovate and when they should not, and their implications on demand for different types of labor and productivity. This project is co-sponsored by AI for Business and Analytics at Wharton.

Michael Platt, James S. Riepe University Professor of Marketing, Neuroscience, and Psychology; Faculty Director, Wharton Neuroscience Initiative
Elizabeth Johnson, Executive Director, Wharton Neuroscience Initiative

Wharton Neuroscience Initiative will support two underrepresented undergraduate summer students focused specifically on Applied Neuroscience and Business Analytics for a 10-week summer internship program. These students will be part of a new, larger applied brain and cognitive science summer undergraduate internship program which aims to combat systemic inequalities and a lack of diversity that plague neuroscience, brain and behavioral science, analytics, and data science careers.

Susanna Berkouwer, Assistant Professor of Business Economics and Public Policy
Arthur van Benthem, Associate Professor of Business Economics and Public Policy

This project aims to understand and quantify inefficiencies in government regulations related to the environment and to provide tangible recommendations for designing smarter policies that address environmental concerns while driving economic growth.

Maurice E. Schweitzer, Cecilia Yen Koo Professor; Professor of Operations, Information and Decisions
T. Bradford Bitterly, Assistant Professor of Management, HKUST Business School
Alex Hirsch, Research Coordinator, Operations, Information and Decisions

This project will support the development and use of an AI-powered chatbot platform for negotiations. This project is co-sponsored by AI for Business and Analytics at Wharton.

Shing-Yi Wang, Associate Professor of Business Economics and Public Policy
Jing Cai, Assistant Professor at University of Maryland

This research explores how firms can improve their hiring decisions by employing state-of-the-art machine learning methods that use observable characteristics of applicants to predict their probability of staying in the job and their performance.

Dean Knox, Assistant Professor of Operations, Information, and Decisions
Rachel Mariman, Senior Research Project Manager, Analytics at Wharton

Using unique access to archives of administrative records on civilian complaints against police, this experimental study seeks to assess how city residents in Philadelphia, New York, and Chicago understand and perceive the current civilian complaint process, and systematically evaluates the impact of transparency initiatives on civic engagement and public trust in police.

J. Daniel Kim, Assistant Professor of Management
Saerom (Ronnie) Lee, Assistant Professor of Management

This project examines the scaling of startups by applying cutting-edge machine learning and econometric tools to assess a novel big dataset of more than 200 million jobs.

Eugene Katsevich, Assistant Professor of Statistics

Single cell CRISPR screen technology, proposed a few years ago, offers unprecedented opportunities to unravel the molecular mechanisms of human disease and guide drug development. The objective of this project is to produce a high-quality software implementation of Dr. Katsevich’s SCEPTRE methodology, with the goal of broad adoption among the genomics community.

Ryan Dew, Assistant Professor of Marketing

This research seeks to enhance our understanding of the role that aesthetics play in consumer decision-making and perceptions of companies and to develop effective tools that help companies craft data-driven visual brands, products, and platforms.

Robert P. Inman, Richard King Mellon Professor Emeritus of Finance; Professor Emeritus of Business Economics & Public Policy

This project uses data collection for a student-led evaluation of the health and economic consequences of the Covid-19 pandemic and the effectiveness of national and state-wide policy responses to contain the coronavirus and to mitigate its health and economic consequences. The format for this evaluation will be a Capstone Course (BEPP 401) entitled, Managing the Pandemic: Messages and Money.

Susan Wachter, Albert Sussman Professor of Real Estate, Professor of Finance
Shane Jensen, Professor of Statistics, Department of Statistics

This project seeks to identify the neighborhood amenity value of transforming blighted and vacant lots into maintained green open space by deploying spatial techniques and integrating multiple data sources to improve our understanding of the dynamics of urban change and identify how residents value “greener” neighborhoods.

Kartik Hosanagar, John C. Hower Professor of Technology & Digital Business; Professor of Marketing

This project seeks to conduct original academic research to better understand how real-world firms use A/B testing software and how consumers respond to online experiments.

Ken Moon, Assistant Professor in Operations, Information and Decisions

In collaboration with three highly sophisticated, intensive-care clinics operating within the University of Pennsylvania hospital system, this project equips medical providers with individually worn biometric sensors that closely track the workplace demands and stresses they experience while navigating swings in high acuity patients’ urgent needs.

Benjamin Keys, Rowan Family Foundation Professor; Professor of Real Estate; Professor of Finance
Maisy Wong, James T. Riady Associate Professor of Real Estate; Assistant Director, Grayken Program in International Real Estate at the Zell/Lurie Real Estate Center

This project procures access to the The Corelogic Multiple Listing Services (MLS) database and includes 10 million observations of property listing data, and more than 600 variables that describe listing details. Professor Ben Keys will use the dataset for his research agenda: How is Climate Change Reshaping Housing and Mortgage Markets? Professor Maisy Wong will use the dataset to study the returns to scale of MLS platforms.

Corinne Low, Assistant Professor of Business Economics and Public Policy
Judd B. Kessler, Associate Professor of Business Economics and Public Policy

This project expands upon the recently published paper in American Economic Review, Incentivized Resume Rating: Eliciting Employer Preferences without Deception, by building two platforms to disseminate the research tools and insights from the IRR method targeted at researchers/policymakers and firms.

Wharton Customer Analytics

This project will develop real-world case studies for teaching data analytics and data science in the classroom based on current datasets from Wharton Customer Analytics partners. In addition, Wharton Customer Analytics will develop a series of in-person workshops that would provide an overview of the analytics basics that are necessary to understand data science, data management, and data visualization. In addition, these workshops will be supplemented by online modules and professional video recorded content for use by students, faculty, and alumni.

Edgar Dobriban, Assistant Professor of Statistics
Eric J. Tchetgen Tchetgen, Luddy Family President’s Distinguished Professor, Professor of Statistics
Steven O. Kimbrough, Professor of Operations, Information and Decisions

Wharton Energy Analytics Lab brings to bear cutting-edge applications of machine learning techniques to position Wharton as an undisputed leader in connecting data analytics and energy markets to face societal challenges. The lab will develop research, teaching expertise, knowledge dissemination and outreach efforts to diverse audiences.

Michael R. Roberts, William H. Lawrence Professor of Finance

This funding supports the development of a new course: Data Science for Finance. The course will introduce students to data science for financial applications using the Python programming language and its ecosystem of packages (e.g., Dask, Matplotlib, Numpy, Numba, Pandas, SciPy, Scikit-Learn, StatsModels). To do so, students will investigate a variety of empirical questions from different areas within finance including: FinTech, investment management, corporate finance, corporate governance, venture capital, private equity, and entrepreneurial finance. The course will highlight how big data and data analytics shape the way finance is practiced.

Hamsa Bastani, Assistant Professor of Operations, Information, and Decisions

While modern data analytics are incredibly effective at extracting valuable insight from text data (e.g., product reviews, nurses’ notes, etc.), they require an enormous amount of data; this research seeks to increase the applicability of state-of-the-art text analytics algorithms by transferring word embeddings for large-scale data to domains with small- and medium-scale datasets.

Witold Henisz, Deloitte & Touche Professor of Management in Honor of Russell E. Palmer, former Managing Partner and Director, Wharton Political Risk Lab

This project focuses on analyzing the materiality of businesses’ environmental, social and governance risks and opportunities and promotes faculty research, teaching, and student learning in this area.

Laura Zarrow, Executive Director of Wharton People Analytics

People Analytics will produce a slate of instructive, engaging, 5-15-minute videos from previous conferences and events that illuminate aspects of people analytics for students, industry professionals, and alumni.

Prasanna Tambe, Associate Professor of Operations, Information, and Decisions

This research project conducts an empirical exploration of the relative costs and benefits of using machine learning based tools on video job application data during the hiring process.

Etan A. Green, Assistant Professor of Operations, Information, and Decisions

This research project trains an artificial intelligence to make optimal offers in negotiations on eBay.

Daniel Taylor, Associate Professor of Accounting

This project will develop research and teaching expertise on the application of Big Data and predictive analytics to issues related to insider trading, financial irregularities, and fraud. The Lab will aim to create new tools and technologies, academic research, and teaching and educational materials.

Mary Purk, Executive Director of Wharton Customer Analytics
Linda Zhao, Professor of Statistics

This conference, for Penn students, aims to inspire and educate data scientists, regardless of gender, and support women in analytics and data science-related careers. Planned for February 14, 2020, on Penn’s campus, this event is part of the larger Women in Data Science (WiDS) initiative originated at Stanford in November 2015 and includes a global conference, 150+ regional events, a datathon, and numerous podcasts.