Weaving AI into the Fiber of all Aspects of Business Education

Imagine a world without mass transit, electricity, the telephone or the internet. Over the past 200 years, each of these was considered a disruptive technology, as the horse and buggy were replaced by rail cars, candles were doused by lightbulbs, and almost no one today has a landline in their home, much less a set of encyclopedias. Each of these was an improvement, even if there were skeptics at the time. And many surely said, “What could possibly be next?” Right now, the answer is artificial intelligence (AI) and its all-encompassing impact on almost everything we do.

Today’s students at the Whitman School take much of it for granted, and little of it truly surprises them. They’ve never known a world without Siri checking the temperature or where Google can’t give them an answer in seconds. They will enter a workplace where AI is not just starting to be incorporated. It’s already there—and students must be prepared for it. AI is revolutionizing the way students learn, faculty research and teach, alumni work and universities revamp education. It is emphasizing critical thinking; the importance of data-driven decisions; fostering creativity, efficiency and technology; and the promise of creating a more inclusive, efficient and engaged learning environment.

The Whitman School is taking the challenge of this emerging technology head on, as it works AI into the fiber of its education and research to fully prepare students to be ahead of the curve in a competitive job market.

According to Interim Dean Alex McKelvie, Whitman revised its learning goals a few years ago to incorporate more technology. The point of this was to formalize in the curriculum that these emerging technological skills would be present. It also helped to match the expectations of companies looking to hire tech-savvy graduates. At that time, however, AI was not specifically identified, but it has quickly become an overarching technology that has been added to the top of that list.

To that end, Whitman has been supporting and incentivizing faculty and staff to use and learn more about AI as a boon to their careers and to provide the best education possible to students.

Many Whitman faculty are enthusiastic about the possibilities of AI in the classroom or to enhance research. Others are starting to dip their toes into the waters to stay current.

“AI has quickly become a vital component of a business education, so it has become a part of just about everything we do,” says McKelvie.

“It didn’t take long for us to realize that we either had to dig deep into it or be left behind. The challenge is not only implementing AI but also doing so when we know that six months from now, it’s already going to look different.” “AI in business education and practice is certainly exciting, and our interest is focused on how we are making sense of it and teaching it responsibly. AI will never replace solid business knowledge and good judgment, but it is going to challenge us by making sure we are using it as a tool for efficiency and innovation, rather than something that is simply doing work for us,” adds McKelvie. “If we want to become a top 25 business school, we have to have some realistic outcomes and train our faculty for AI.”

Surveys Show Faculty Has Mixed Approach to AI 

Last fall, the Whitman School surveyed faculty members under the direction of Associate Dean for Business Education Willie Reddic G’12, ’13 Ph.D., asking about their current level of AI use and class policies.

Results showed that 33% of Whitman faculty allowed unrestricted use of AI in their classrooms, 48% allowed it sometimes and 19% had a zero tolerance policy for using AI. Those who indicated a zero tolerance policy were asked to provide a rationale for their reasons. Sixty-seven percent of faculty had an interest in learning more about ways to integrate AI into their courses. And the survey showed no meaningful difference between use policies in undergraduate or graduate courses.

Willie Reddic

“In keeping with the mission of the University and Whitman’s own executive leadership, all courses need to have at least some component of AI incorporated,” says Reddic. “Whitman knows this is a necessary step to remain competitive, so we are working on ensuring AI is adopted across every major and that our faculty have the resources and confidence to implement various elements. The bottom line is that we are a business school, and businesses are using AI.”

To that end, the Whitman School has offered its faculty, staff and students ongoing training to learn the basics of Google AI Essentials. Starting last summer, faculty and staff went through an AI certification program to meet them where they were in terms of their AI knowledge and comfort level. The initiative was so popular that it has become ongoing, so that others at Whitman, including students, can participate. To date, approximately 125 people have gone through this training.

“As Whitman staff, we have been encouraged to challenge ourselves, particularly in the area of AI. Earlier this year, I took advantage of pursuing the Google AI Essentials, and now another AI certification is available. I hope to add that this summer,” says Tim Findlay G’22 (EDU), assistant director of graduate recruitment. “As the use of AI becomes standard in the world, I have built my knowledge to use, in my day-to-day work and encourage our students to do the same. I appreciate the support Whitman offers to help make that happen.”

AI in the Classroom

Last fall, Whitman held a teaching luncheon where several faculty members presented ways in which they have incorporated AI into their courses, demonstrating both simple and complex ways of using AI as a teaching tool benefitting professors and students. 

Creating Q/A With AI Personality Sets the Tone

Professor of Finance Anna Chernobai, who chairs Whitman’s Teaching Committee, is enthusiastic about her use of AI.

“I’ve always loved technology and different ways to use it in the classroom,” Chernobai says, “I was the first Whitman faculty member to record all of my lectures before COVID, so I was 100% prepared when we suddenly had to transition to remote learning on Zoom, which was fairly new to a lot of people. I helped to train other faculty during what was a really hectic time.”

In her graduate-level course Data Analysis and Decision Making (MBC 638), Chernobai continues to record her class lectures on Zoom, using Kaltura to create scripts that she uploads into Microsoft’s Azure OpenAI. Adding PowerPoint slides and other course information, as well, quickly helps create practice questions, quizzes or exams using a feature incorporated into Blackboard.

Anna Chernobai teaching one of her classes

She has also created an automated Q/A section to efficiently address general student questions based on information loaded into Blackboard. For example, if a student asks, “What does correlation mean in AI?” an answer is generated based on the information from Chernobai’s class materials. 

The feature also allows her to set the tone of the response—ranging from kind, happy and mean. “Setting the feature to kind is like getting an answer as if your grandmother is explaining it to you,”says Chernobai.

However, she has found that some students enjoy the mean version, too. For example, a student may enter a prompt about a mathematical problem and will be given a preliminary answer like, “Should I hold your hand while you calculate that?”

“I have tested out all three versions, and they all answer my questions in whichever tone I choose. I tend to stick with happy, but sometimes the mean version is fun to use. I do think this tool is handy since I can ask the chatbot about topics in our PowerPoints or lectures without searching for exactly when it was discussed. It has also helped explain topics that I am having a hard time picking up on,” says James Laquidara ’26 MBA.

The Q/A can be programmed to answer from different perspectives, a helpful feature when students are working on a case study and preparing to interview a company’s CEO, for example. AI allows the chat to play the role of a CEO using AI Conversation, so students can practice before going in front of an actual person, often helping them feel relaxed and better prepared.

Of course, AI is not perfect. Sometimes the answers the Q/A chat provides are not completely accurate, so Chernobai is careful to monitor the chat and tweak the information that has been entered. And there are limitations to AI use, which are clearly outlined in her syllabi (as required of all courses at Whitman and the University) that may differ from course to course.

Chernobai believes that most of her students are fascinated with AI and appreciate the use of it in her courses.

“My students tell me they love it, and the bottom line is that I’m just trying to keep them engaged and excited about learning,” Chernobai says.

Using AI to “Find Your Why”

Other faculty also offered examples of how they are using AI to increase student learning.

Arielle Newman teaching

Assistant Professor of Entrepreneurship Arielle Newman uses AI in Introduction to Entrepreneurship (EEE 370) for idea generation. Students are asked to think of a business idea and build it out throughout the semester. At times, many ideas are similar. So she has used AI to help students expand their ideas through a “Find Your Why” exercise where she gives each pair of students a detailed prompt via a QR code, so ChatGPT can help generate three to five business ideas that are more aligned with their true interests. Students can refine the prompt further to more closely match their ideas using a derivative feedback loop in ChatGPT.

“This has helped my students create much more interesting ideas than I’ve seen in a while, and it’s helping them become more comfortable with AI and understand how essential a specifically targeted prompt is to getting the output you need,” Newman says, noting that students are required to send her a link to their chats to see the depth and quality of their prompts. 

Sweet Ways to Keep Students Engaged

Assistant Teaching Professor Christie Novak ’10, ’11 M.S. teaches Introduction to Accounting (ACC 151), required of all undergraduate students. “It can be a difficult course to keep everyone engaged because some have no plans of going into accounting,” she says.

Christie Novak

However, Novak recently turned to ChatGPT for ideas to create in-class activities to solve practice problems in ways that hold almost every student’s attention.

One of the activities to come out of that query is a fixed asset challenge based on the iconic board game Candy Land. Using individual colors as different types of transactions, Novak puts construction paper across the floor with questions for the team. Students move around the game board learning accounting and can’t help but be engaged.

“The idea is not to take away the practice of fixed assets but instead integrate the content in an exciting way that keeps their attention,” Novak says.“You lose them if they’re not interested, and asking ChatGPT for ideas to apply to my lessons keeps them excited about coming to class.”

Students Use AI to Prepare for Experiential Learning in Asia

Of course, AI use is not limited to inside the classroom. When Associate Dean for Global Initiatives and Professor of Marketing Eunkye Lee took a group of students to Singapore and Malaysia for Business in East Asia (INB 400/600) in May 2024, he gave them an assignment using AI to prepare for the trip.

Kyu Lee

“Imagine if you were going on a business trip, and you had an assistant who could put together a basic report on the business environment of your destination,” says Lee.

“AI is a smart assistant that can do just that. However, part of the challenge was for students to verify that the information they were getting was correct, as well as identify critical holes and gaps.” he adds.

Students were required to create a pre-departure report before arriving in Asia with a brief about the business atmosphere in both countries using ChatGPT or any AI tool. The six- to eightpage brief was required to include information on the countries’ economic histories and current geopolitical, economic, social/ demographic, cultural and technological environments; as well as all sources used by their AI tools. In addition, they were required to add a reflection listing major points learned from the exercise, as well as important questions not answered in the report, which they had to investigate through company visits and other immersive activities on the trip.

“Overall, I think the assignment was an excellent introduction to ChatGPT. It was able to reveal many of the capabilities and limitations of generative AI, as I tried to craft prompts that would generate text adhering to the parameters of the assignment,” says Qiong Wu ’26 (WSM/A&S/MAX), an accounting, business analytics and finance major.

“I think it shows that Professor Lee is not only adapting to emerging technologies himself but also trying to pique students’ interest, so that they are motivated to gain skills that could be useful in the future.”

Use of AI Crucial to Whitman Competitions, Business Ideation

Students Use AI to Organize and Develop Capstone Ideas

The Whitman capstone course, Strategic and Entrepreneurial Management (EEE 457), is the culmination of four years of undergraduate work and usually leaves a lasting impression. Capstone is a required course taught each semester by faculty from the Department of Entrepreneurship and Emerging Enterprises during students’ senior year. Each student is put into a group challenged to create a product, service or business model that would be feasible in the real world and include a novel approach to making economic value. These ideas must be scalable with the ability to achieve $3 million in revenue within five years, as well as attract external investments exceeding $100,000. At the end of the semester, students present their plans to alumni and other industry professionals for awards, feedback and bragging rights.

Ken Walsleben (left) and Interim Dean McKelvie (right) posing with the Fall 2024 Capstone winning team, LaceMate.

According to Associate Professor of Entrepreneurship David Park, the faculty lead for capstone, AI was incorporated into this course in late 2022. Students are encouraged to utilize AI throughout the entrepreneurial process, from ideation to pitching. Park notes that students have been urged to experiment with various AI tools, not just ChatGPT or Gemini.

For their capstone project, Laurenne Yomi Mvete ’25, accounting major, and Maya Layton ’25, a double major in retail management and management, came up with the idea of ConnectUp, a platform designed to empower student entrepreneurs by connecting them with alumni entrepreneurs, fellow students and potential investors through AI. It would serve as a hub for networking, mentorship and freelance services, making it easier for students to navigate the challenges of launching a startup.

As the idea developed, it became apparent that AI was a part of the project in a way the team didn’t initially expect, according to Layton. In their business plan development, AI helped refine their proposal and structure ideas, as well as ensured they had a compelling pitch that included all main points for the judges. In addition, they used Gemini to streamline the creation task templates that kept them on track, and it transformed ConnectUp’s business plan by creating customer discovery questions to understand the pain points and needs of student entrepreneurs.

“Capstone was an experience that pushed us to think critically, refine ideas and embrace new technologies, especially AI,” says Mvete. “It reinforced a belief I’ve always had that entrepreneurship isn’t just about having an idea. It’s about taking action. And, with AI as a tool, I’m more confident that students have what they need to get started on their ventures.”

Drone Business Goes Sky High With AI

Angelo Niforatos ’20 M.S. (ECS), ’25 MBA, is using AI to further another disruptive technology—drones—through his startup Niffy Drone Solutions LLC.

A full-time systems engineer at ResilienX with a master’s degree in aerospace engineering from Syracuse University’s College of Engineering and Computer Science, Niforatos already had an interest in advancing drone technology before starting his MBA at Whitman. The program helped him turn his idea into reality with faculty entrepreneurs guiding him.

Angelo Niforatos

In spring 2024, he participated in the Orange Innovation Grant. He used the prize money to build a formal business plan and prototype for Niffy Drone Solutions. With a plan and proof of concept in place, he competed in Whitman’s Orange Tank competition last fall, supported by Linda Dickerson Hartsock, founder of the University’s Blackstone LaunchPad, and Indaria Jones, manager of Whitman’s Couri Hatchery Incubator.

“They helped me put together a great presentation that communicated my ideas to the judges properly, as I tend to get a little too nerdy with it,” he says. His hard work paid off, placing second and receiving feedback from angel investors, alumni, faculty and business leaders who served as judges.

Today, Niffy Drone is at the forefront of AI-driven autonomy, specializing in developing state-of-the-art technology that acts as a co-pilot for unmanned systems across commercial, defense and industrial applications using real-time sensor data processing, natural language processing and machine learning (ML) to create greater operational efficiency, according to Niforatos, the company’s CEO.

“Niffy Drone is bridging the gap between autonomous control and human oversight by reducing operator data overload and enabling seamless mission execution even in areas with limited internet connectivity,” he says, noting that this technology is revolutionizing not only drones but also unmanned aerial, ground and maritime vehicles with solutions ideal for military, defense and infrastructure sectors where reliability and ease of use are essential.

The company recently won a Small Business Innovation Research contract from the U.S. Air Force to perform feasibility research on this technology. “I’m grateful for those at Whitman who helped me get my business off the ground. With AI technology evolving so quickly, I am excited to see what new capabilities are ahead,” says Niforatos

AI-Sensored Sock Provides Life Saving Data for Diabetics

Establishing her startup, DiabeTech, is very personal for Tosin Alabi ’25 MBA, who lost her father to diabetes. Her loss led to an idea to create a revolutionary mobile health platform using wearable sensors and AI analytics designed to manage diabetic foot ulcers, which can result in gangrene or amputation.

Alabi, who has a background in software engineering and a master’s degree in information technology in health care from the University of Greenwich in London, came to Whitman to earn an MBA and focus on the business/entrepreneurial side of health care innovation. While she had a solid idea, it has been her time at Whitman, particularly through the resources of the Couri Hatchery Incubator, competitions that added to her start-up funding, and the mentorship and feedback from faculty and alumni entrepreneurs, that made her product a reality.

Her original concept was a bandage for diabetic foot ulcers with wearable sensors. However, she recently pivoted to a sock using the same sensors and AI technology. Today, DiabeTech offers advanced foot monitoring through a revolutionary mobile health alert platform that preempts complications with real time notifications. If any changes in the foot are detected, comprehensive AI data analytics enable patients and their physicians to monitor health trends for better treatment decisions, and personal care recommendations tailored to each person based on the data conveyed.

Alabi has won a number of competitions at Whitman with her idea. In November, she placed first in the 10th Annual Orange Tank Pitch competition, taking home a $25,000 prize in a “Shark Tank”-style event that is an entrepreneurial highlight at Whitman each year. 

An entrepreneur-in-residence at Whitman’s Couri Hatchery, Alabi helps other students across the University develop their own ideas and recently rolled out a website (www.diabetech.io) to introduce her socks and the lifesaving technology they can provide to consumers.

“My father’s diagnosis shattered our world, and I’ve wanted to create something that could prevent pain for other families facing similar challenges,” says Alabi. “The mentorship and resources at Whitman empowered me to turn my ideas into a viable startup with a focus on innovation and entrepreneurship that gave me the confidence to push boundaries and bring a product to market that can truly make a difference.”

Graduate Student Competition Offers Tasty AI Solutions for Local Business

If you’ve never tasted a chocolate pizza—a combination of chocolate, English toffee and sweet toppings popped into a pizza box—you’re missing out. And while Ryan Novak ’11 knows the business of making chocolate pizza and other delicacies inside and out, he thought he might be missing out on opportunities to use AI to further his Marcellus, New York, small business.

Ryan Novak

Novak purchased the Chocolate Pizza Co. in 2010 and has expanded its online business to 50 states and 17 countries, shipping as many as 1,000 chocolate pizzas daily during the holiday season. However, much of his focus has been on creating more manufacturing space, expanding e-commerce, shipping and managing the retail store—not on AI.

So, when Whitman approached him about using his company as a case study for the 2024 Graduate Case Competition, Novak was all in. The challenge for the graduate students was to develop a comprehensive AI strategy with feasible solutions, a budget and a detailed action plan for the Chocolate Pizza Co.

Of the 11 teams that participated, five were selected to make final presentations to Novak last October, giving him many ideas for implementing AI. The first-place winners were Trang Nguyen ’25 MBA, Sinduri Vangala ’25 M.S. and Ahan Kent ’25 M.S., who proposed a combination of Hubspot and Breeze AI tools to increase sales, service and marketing.

However, it was the second-place team whose idea was incorporated into the business recently. Allison Hellman ’26 MBA, ’26 M.S (A&S); Vaijayanthi Kadhiravan ’26 MBA, ’26 M.S. (MAX); Zhen Shi ’26 MBA; Shaurya Jain ’25 M.S. (IST); and Prasad Ranka ’25 M.S. (ECS) proposed a chatbot to answer routine questions on Novak’s recently launched new website (www.chocolatepizza.com), saving time for both his team and customers. The website’s analytics have shown that the chatbot has positively impacted customer satisfaction and is keeping the bounce rate down, which is great for business.

“This competition opened my eyes to a whole new world of AI possibilities,” Novak says, adding that he may consider incorporating other teams’ suggestions in the future.

“The chatbot was a great jumping-off point to give more thought to how AI can improve our business and allow our staff of only eight to concentrate on other goals. I am grateful for the ideas all the students came up with to make our business even better and am very impressed with how Whitman is willing to use its resources to help local businesses.”

Whitman’ s Faculty Research Encompasses AI/ML Across Disciplines

With the popularity of AI and the enhanced technological advances it continuous to bring, numerous faculty members at the Whitman School are exploring what the various types of AI can do to make their research more efficient and also benefit their areas of expertise.

“AI is here to stay, and our faculty is embracing it in new and interesting ways that not only benefit their own work but also advance the reputation of the Whitman School,” says Michel Benaroch, associate dean for research and Ph.D. programs and professor of management information systems.

“With the support of Whitman’s administration, we are seeing a lot of exciting ideas and advances coming from our faculty and their doctoral students from research using AI as a methodology for investigating instant return credit policies in the supply chain space, to using deep ML to study how image manipulation affects guest reviews on the Airbnb platform in marketing, to the ability to improve health care using Machine Learning for Health, and investors’ response to financial advice from a human versus a machine. We are welcoming AI, as it is opening doors to new and more efficient ways of doing things and a greater level of discovery here at Whitman, “ explains Benaroch.

AI Tools Help Examine Impact of Instant Credit Return on Online Purchases

One example is the work of Associate Professor of Supply Chain Management Rong Li, who has published papers in Productions and Operations Management. In a recent working paper “Online Retailing With Instant Return Credit” (with Chenxin Liao of the Chinese University of Hong Kong Business School and Duo Shi of the Chinese University of Hong Kong, Shenzhen, School of Management and Economics), Li uses predictive analytics enabled by AI tools to examine the strategy of instant return credit (IRC) in online retailing. Many online retailers have recently adopted instant return credit, where store credit is offered immediately upon a return claim without requiring it to be received or verified (similar to when a consumer starts a return with Amazon through an online account). IRC helps resolve the mismatches between products and consumer tastes and converts online returns into new sales by improving customers’ shopping budget through instant credit. However, IRC is susceptible to misuse, as some consumers may unintentionally delay returns, forget to send items back or event attempt fraudulent returns.

Rong Li

In the paper, Rong and her co-authors study IRC’s fundamental dynamics and implications under two settings, one of which assumes that dishonest customers can be identified using predictive analytics enabled by AI tools. In the baseline setting, the retailer offers IRC uniformly to all consumers under three schemes: 

• Offer partial IRC coupled with symmetric pricing under low market risks;

• Offer full IRC coupled with asymmetric pricing under medium market risk, and;

• Not offer IRC under high market risk.

In the advanced setting, the retailer offers IRC contingently; full IRC to honest consumers and partial or no IRC to dishonest customers. The researchers demonstrate that uniform IRC policy is more valuable for low-risk markets, while making IRC contingent on consumer type (honest or dishonest) is more valuable for medium-risk markets. They also demonstrate that uniform IRC may hurt both types of consumers for low-cost products, while making IRC contingent normally rewards honest consumers and penalizes dishonest customers. 

Do Gen-AI Ads Enhance Salesperson Performance?

In The Effects of Gen-AI Marketing Systems on Salesperson Performance By Turnover by Associate Professor of Marketing Guiyang Xiong, underperformance is prevalent in the sales occupation, and pressure to perform well is paramount. To empower salesforce success in a dynamic business environment, companies increasingly invest in generative-AI marketing systems, which automatically create marketing materials that salespeople can deliver to the customers.

However, empirical research on the effectiveness of such gen-AI systems is scarce, according to Xiong, and there are concerns over whether they affect the retention of high-performance salespersons as they might feel they are being “replaced” by AI.

Based on an analysis of panel data on over 1,000 salespeople with a beauty-product company, this research finds that gen-AI ads may enhance salesperson performance if used properly. The effectiveness of gen-AI ads is conditional on  communication media (mass vs private communication channels) and salesperson experience. Moreover, experienced salespersons who frequently use gen-AI are less (instead of more) prone to leave the company. The findings provide novel implications on an important application of AI in marketing and sales force management.

Like Guiyang, other marketing professors and doctoral students they guide at the Whitman School have been conducting research that advanced academic understanding of AI-centered marketing activities and strategy and their impact on consumers and firms, according to Benaroch. This includes the study of AI-assisted customer purchasing decisions and AI-assisted customer guidance. The research provides practical insights for businesses and highlights broader implications for media AI, particularly in personalized consumer diagnostic and treatment recommendations. The findings guide strategies for leveraging AI innovations across the beauty, wellness and healthcare industries.

Image Misrepresentation Negatively Impacts Selection of Airbnbs

Doctoral marketing student Ali Kozehgaran ’26 Ph.D. is examining how the emergence of digital platforms has transformed commerce, with visual content becoming crucial for consumer decision making. In experiencebased markets like short-term rentals, where pre-purchase evaluation relies heavily on images, the potential for using AI to generate visual misrepresentation poses many challenges.

Ali Kozehgaran

This study, “The Impact of Image Manipulation on Consumer Perceptions of Experience: The Case of Airbnb” (with Yang, L. and Jiang, J.), is in the early stages of how image manipulation affects guest reviews on the Airbnb platform. Unlike traditional e-commerce, these experiences cannot be returned post-consumption. Leveraging a dataset of more than 56,000 properties with more than 221,000 reviews from AirDNA, which collects short-term rental data, the researchers developed a deep (machine) learning model trained on the CASIA2 dataset to detect image manipulation through Error Level Analysis, employed Natural Language Processing to analyze review sentiments, and implemented casual ML models to establish the relationship between manipulated images and guest satisfaction.

According to Kozehgaran, this research represents the first systematic investigation of visual misrepresentation’s impact in the sharing economy. The findings will advance the understanding of information asymmetry in digital marketplaces and inform platform governance policies, particularly in experience-based markets where traditional consumer protection mechanisms may be ineffective.

Taking a Higher View of AI

Developing Consistent AI Policies for INFORMS Journals, Publication

Other Whitman faculty have taken a 10,000- foot view to study the rapid pace at which emerging AI/ML technologies are impacting and transforming the academic and business of other domains.

The Steven R. Becker Professor of Supply Chain Management and the Laura J. and L. Douglas Meredith Professor of Teaching Excellence Burak Kazaz has been chairing the committee in charge of establishing AI policies and guidelines for all Institute for Operations Research and Management Sciences (INFORMS) journals and publications to establish a common language and expectations for all submitting authors and the editorial review team.

Burak Kazaz

According to Kazaz, this has been a huge undertaking as INFORMS is a collection of a diverse set of scholars from engineering, business schools, marketing, finance, information systems, mathematics, analytics and statistics, bringing together over 12,500 researchers. The INFORMS Society publishes 17 journals, and the standards to be established for each is challenging.

“The use of AI in scientific work is inevitable. It’s going to happen, but we have to apply ethical standards. AI will be a benefit, as it does things much faster than we can, but authors have to check their work. We are asking everyone to be more transparent if AI is used, like indicating what prompts were used and how else they integrated AI into their work. If there is an AI error, the authors will be responsible for that,” says Kazaz.

“We are adopting three different guidelines for our editors, because one size does not fit all for the 17 INFORMS journals. We have also created guidelines for reviewers: the use of AI is completely prohibited from the refereeing process because such submissions violates the authors’ intellectual property rights. We’ve been working on this for a year, and we still have work to do, but our goal has been accommodating every journal and focusing without deviating from the idea of new knowledge with the intent of making our society better.” 

Can Machine Learning for Health Revolutionize Health Care?

In “Reproducibility of AI/ML for Health,” Michel Benaroch, along with Edward Raff ’23 MBA, examines the reproducibility crisis facing AI/machine learning (ML) models in the health domain in general.

Michel Benaroch

Machine Learning for Health (ML4H) has a high potential to revolutionize health care through improved medical decisions for diagnosis, prognosis, drug discovery risk assessment, patient empowerment and more, and yet only a relatively small number of ML4H applications made it to clinical use and received FDA approval. According to Benaroch, this is blamed in part on a growing reproducibility problem—an inability to replicate the results of many ML4H studies across different datasets and conditions. If you can’t reproduce someone’s model and results using similar data and slightly different organizational settings, you cannot trust the model and its results. Evidence shows that up to 50% of published ML4H models cannot be replicated due to issues like non-sharing of development data, of the algorithm or code, and of methodology details. Most striking, however, is that of 62 ML/ AI models developed in 2020 for diagnosing COVID-19 from medical scans, none were ready to be deployed clinically because of flaws such as biases in the data, methodology issuues and, most importantly, reproducibility failures.

In the past 10 years, over 100 articles on the ML4H reproducibility problems have been published with over 30 articles in Nature and Science magazines alone. According to Benaroch, the study responds to calls in this literature by identifying eight different aspects of reproducibility based on review of extant literature. Moreover, while extant literature has been focused on reproducibility only at the time a ML4H model is launched, the researchers argue that ML4H models in clinical use are living artifacts that evolve past their launch. Accordingly, the eight aspects of reproducibility they identify span the entire ML4H model lifecycle.

Most importantly, the researchers assess the state of reproducibility of AI/ML4H by reviewing over 900 articles published between 2017 and 2019 and identify which of the eight aspects are addressed by AI/ML4H developers and which aspects represent serious gaps that hold back the potential of ML4H.

The Weakness of AI/ML

Other Whitman professors and doctoral students have been examining areas where AI/ML may show weaknesses.

Who to Trust with Financial Decisions: Human Analysts or AI?

Associate Professor of Finance Si Cheng has a study, “Algorithm (Mis) Appreciation? Investors’ Response to the Mutual Fund Ratings by Human Analysts vs. AI,” providing the first comprehensive analysis of the investor response to two types of mutual fund ratings offered by Morningstar: the Analyst Rating produced by human analysts and the Quantitative Rating generated by AI. The study found that retail investors tend to ignore the recommendations of human analysts and instead chase ratings by AI. However, while human analysts have some success in identifying outperforming funds, IA ratings fail to do so.

Si Cheng

Yet, by constructing a counterfactual AI rating following the Morningstar methodology, Cheng found that human analysts outperform AI mainly through the selective coverage of funds. Furthermore, individual investors ignore the useful soft information in analysts’ reports written by human analysts and incorrectly respond to the tone of the title and summary section instead of the full report. Overall, the findings shed light on the preference of individuals for algorithmic advice versus human advice, suggesting a potential algorithm misappreciation problem and capital misallocation in mutual fund investment. 

Disclosure of AI Use Negatively Impacts Creative Reputation

Third-year doctoral management student Anand Benegal, guided by Professors Lynne Vincent and Joel Carnevale, investigated how the use of AI affects people with a reputation for creativity. They find that disclosing the use of AI has a consistently negative effect on several elements of a person’s creative reputation, such as reputational prestige and creative competence.

Anand Benegal

In “Creative or Contrived? How AI Use Shapes the Social Evaluations of People with Creative Reputations,” Benegal’s work illustrates and unpacks this phenomenon, by focusing on how perceptions of less (or more) AI use ameliorate (or exacerbate) this reputational penalty toward people who use AI, and how these relationships are mediated by perceptions of the person’s creativity as being inauthentic in nature, which goes on to affect their creative reputation negatively.

Alumnus Working in AI

Whitman MBA Grad Using Adversarial Machine Learning for Defense Program

Edward Raff ’23 MBA is director of emerging AI and a distinguished scientist at Booz Allen Hamilton, a company that uses advanced technology to drive speed to outcomes for the federal government. Booz Allen is the largest provider of AI to the federal government, combining AI innovation with cybersecurity, engineering and emerging technology to deliver results for the country’s most critical defense, civil and national security priorities.

Edward Raff

His job title didn’t even exist a few years ago. He joined Booz Allen Hamilton in 2013 to help build biometric fingerprint recognition systems; help with security, malware analysis and detection; and implement the growing techniques of machine learning (ML) and how to use them. Raff has risen through the ranks as a well-respected scientist who is grateful for the ability to shape his career in the direction of the technology that fascinates him.

Raff’s current work involves defense, particularly adversarial ML, like how to circumvent bad actors trying to evade detection, such as scammers trying to avoid being discovered. In health care, he uses using AI to understand data from the National Institutes of Health that measure the amount of pain cancer patients are experiencing between doctors’ visits to determine if they are truly in pain or drug seeking.

Raff’s interest in ML came to him by chance. While studying at the University of East Anglia in the United Kingdom, he accidentally enrolled in a ML class, which sparked a true interest in the subject. Earning a bachelor’s and master’s degree in computer science from Purdue University, Raff noticed that many job opportunities he was interested in required a Ph.D.

He was hired at Booz Allen without one, but several years later earned a Ph.D. in computer science with research in applying ML to malware classifications and clustering from the University of Maryland, Baltimore County—while still working full time.

In 2022, he decided that building business skills was an important complement to his career and the rest of his education. Having just moved to Syracuse, he enrolled in the Whitman’s School residential MBA program, while working at Booz Allen around classes.

“I got more out of the MBA experience than I anticipated, and it provided a framework to be conscious in how I think of things,” Raff says. “I learned from mentorship, from negotiations class where I could apply things I’d been doing on the job, and accounting class answered a lot of questions I had on the financial end of my work. So, getting this kind of formal business training was of real value to me and my career.”

Raff’s advice to those interested in the AI/ML field is not to just try to learn ML but to pick a problem of interest and work on learning the technology through that.

“Build a model about who will lose the next game if you like sports. Examine how traffic works or about revenues or property taxes if you’re interested in city management,” says Raff, who is completing a book on ChatGPT, How Large Language Models Work, to be published by Manning Books. “That’s the motivation you need to have a real understanding of AI and ML, and it’s built into all the standard things you might use. There’s a big difference between being able to click a button and learning how to interpret the answers you got and the mistakes you made. Gaining that level of understanding of machine learning is much easier when you have an idea or an interest nagging at you."

The Future of AI

This is just a slice of the interest in AI that Whitman faculty, students and alumni have undertaken recently. And, while it reflects a vast range of AI-related topics, it is a representation of Whitman’s collective ambition to stay current with, if not in front of, the multifaceted challenges, opportunities and proactive knowledge surrounding AI as a disruptive technology changing the way we live and do business. “I am both intrigued by and proud of the level of work we are doing at Whitman to advance our education and support our faculty and students in becoming future business leaders through their curiosity of AI’s possibilities. No one can exactly predict what AI’s evolution will bring. Yet, by embracing AI and examining the multiple facets of education and discovery, I believe Whitman can make a significant contribution to ensuring that we’re ready for the future.” says McKelvie.

AI Tips from Alumni in the Know

Best Practices for AI Use in Business

Ziyu (Connor) Huang ’21 was a double major in accounting and supply chain management while at Whitman and received a master’s degree in analytics from the USC Viterbi School of Engineering. Today, he is a senior associate business analyst at Amgen, which is using biology and technology to combat diseases by developing, manufacturing and delivering innovative medicines to millions all over the world.

Huang offers some tips about best practices for AI:

1. Prompt engineering is the key to effective AI utilization. A good prompt is always a clear instruction on what you want AI to do by steps. The more detailed information or steps you provide, the more accurate AI will perform, based on your command. Business students should learn how to craft prompts that maximize AI’s potential for productivity, creativity and strategic insights. The normal format he uses for every prompt is a clear instruction on what he wants to achieve, while providing enough background information to help AI understand the current situation. In the meantime, providing feedback on the answer generated by AI can also help increase the accuracy of the answer.

2. Learn how to use AI Application Programming Interfaces (API) for largescale automation. Learning how to write and implement chatbot API calls can help to process large volumes of data, optimize workflows and enhance decision making without needed deep AI expertise.

3. Understand and utilize AI agents. An AI agent can perform tasks, make decisions and interact with users or systems using AI. These agents are often powered by large language models or specialized AI frameworks. Learn to utilize it by giving detailed instruction that will be very beneficial to work efficiency.

Bringing Out the Best of AI

A dual major in entrepreneurship and emerging enterprises at Whitman and political science at the Maxwell School of Citizenship and Public Affairs, Abby Hamilton ’19 (WSM/ MAX) is a product manager for AI science incubation at Microsoft. She has been at Microsoft for five years, working with customers using AI to solve generational issues when it comes to innovation and customers using the technology for scientific purposes.

Hamilton offers the following tips for leveraging AI as a positive technology:

1. Develop a foundational understanding of how AI works. While deep technology knowledge isn’t necessary, understanding how cloud computing and data centers work will set you apart and enable you to engage in informed discussion on how your organization can adopt AI.

2. Champion responsible AI in business decisions because responsible use is not just a technical issue, it’s an imperative. For example, minimizing bias in algorithms is a practice that matters and ensures compliance with regulations that will eventually emerge. Other examples are protecting customer data and being transparent in how AI decision are made.

3. Position yourself as a connector between AI and business strategy. AI processes vast amounts of data, but people come in with creativity and problem solving skills needed for nuanced decision making and problem solving innovation. Creative problem solving and innovative thinking help develop balance and show your strengths.

4. Embrace AI as a catalyst for continuous information. It’s constantly reshaping how businesses run and looking at opportunities for your business, so early adopting will be differentiating across industries and pay off in dividends long term. Others will be left behind if they don’t adopt early and try new things.

5. Look to AI to solve humanity’s greatest challenges, like advancing climate science solutions. Aligning your expertise with these efforts could not only give you the opportunity to be a leader but also to contribute to a positive future.