As with any innovative technology, the latest developments in artificial intelligence (AI) and machine learning (ML) are only valuable if one knows how and where to use them. Academic research can play a vital role in finding and validating business use cases that help to justify the large investments of money and resources that AI entails.
In recent months, Costello College of Business faculty have produced scholarship documenting innovative and impactful uses for artificial intelligence, in fields as disparate as nonprofit fundraising, investment advising and corporate audit regulation. This effort aligns with George Mason University’s deep commitment to tapping the vast potential of this technology while maintaining ethical guardrails, thus balancing grit with audacity.
What follows are just a few examples of Costello’s contribution to this institutional initiative—by no means a complete list.
WHAT AI KNOWS ABOUT YOUR COMPANY (THAT YOU LIKELY DON’T)
Yi Cao, assistant professor of accounting, is using AI to ferret out hidden company information in unlikely places, such as LinkedIn job postings. His research finds, for example, that companies planning to engage in shady accounting behavior are likely to telegraph that intention in their job ads. In this case, AI was used to help identify “rulebender” and “rule-follower” language in job postings, and the predictions turned out to be accurate.
Job postings can serve as early indicators of a company’s intent, offering insight into its values without needing insider access. By analyzing public-facing language used in hiring, the study offers a new way for regulators and investors to identify financial manipulation, i.e. detect risk not just through numbers, but through words.
“We utilize large language models (LLMs) as a tool that is capable of gauging verbal or textual variation—using this powerful tool to find information that can be indicative of financial misbehaviors,” Cao says. “This is beneficial to everyone in the financial market: policy makers, shareholders, tax agencies, and other stakeholders. So, we hope this study can bring new insight to a hot topic that people care about.”
In addition, Cao and Long Chen, area chair for accounting, have coauthored research suggesting that large language models such as ChatGPT could rival investment professionals in at least one area essential to firm valuation: generating lists of relevant “peer firms” against which to benchmark corporate performance.
Using “Gemini,” a Google-developed LLM, Cao and Chen could generate about seven peer firms for a focal firm, a number that is similar to the SEC recommendations on how firms should disclose their segments.
The researchers then compared the LLM’s performance to the lists generated by three human experts for a set of 40 leading computer software companies. The average overlap was a little over 40 percent, greater than expected.
But it wasn’t a clear-cut case of LLM supremacy, since standard systems outperformed AI for mid-sized firms.
“We need to embrace this new technology, but we must recognize that it is not yet in a perfect state. Competition to improve the technology is fierce. Our findings might just represent the lower bound of the effectiveness of the technology.”
— Long Chen, area chair for accounting at the Costello College of Business at George Mason University
Chen concludes, “We need to embrace this new technology, but we must recognize that it is not yet in a perfect state. Competition to improve the technology is fierce. Our findings might just represent the lower bound of the effectiveness of the technology.”
ARTIFICIAL INTELLIGENCE’S SOCIAL IMPACT
Further, Costello faculty are discovering how AI can help missiondriven organizations do more to further their worthy causes. Amid robust debate about Big Tech’s impact on society, these use cases indicate that there’s more to AI than a profit engine for businesses. Indeed, AI can help promote societal virtues such as inclusivity.
Jiyeon Hong, assistant professor of marketing, developed an algorithm to help improve the effectiveness of donor solicitations for DonorsChoose.org, a digital fundraising platform for teachers looking to finance classroom projects.
After training with more than 60,000 DonorsChoose projects from the period 2009-2017, the algorithm identified which sentences within the DonorsChoose proposals were the most pivotal to the outcome (i.e. getting funded or not). The effectiveness of the algorithm was then verified in an experiment, in which teachers revised the essays and a group of undergraduates rated the revised work. Overall, the essays revised with the help of the algorithm were judged 4.5 percent more likely to be funded, a difference amounting to nearly $10 million in additional funding.
The takeaways from this experiment may also apply to fundraising language in general, not just for DonorsChoose. Noting that more complicated language was often more persuasive with donors than a more simplistic presentation, Hong says “We conclude that the most successful appeals for help will not be those that make the simplest and tightest arguments. Instead, they will a) expose the reader to a modest amount of desirable difficulty; and b) put forth a detailed case that is low on emotional coloration.”
Similarly, Siddharth Bhattacharya and Pallab Sanyal, two professors in the information systems and operations management area, are partnering with real nonprofits to perfect and refine their chatbots and thus increase donations to these deserving organizations.
“We completed a project in Minneapolis and are working with other organizations, in Boston, Massachusetts, New Jersey and elsewhere, but the focus is always the same,” Sanyal says. “How can we leverage AI to enhance efficiency, reduce costs, and improve service quality in nonprofit organizations?”
Their field experiment with a Minneapolis-based women’s health organization involved designing a custom chatbot to interact with prospective patrons through the organization’s Facebook Messenger app. The bot was programmed to adjust, at random, its responses to be more or less emotional, as well as more or less anthropomorphic (human-like).
Over the course of hundreds of real Facebook interactions, the moderately emotional chatbot achieved deepest user engagement, as defined by a completed conversation. But when the emotional level went from moderate to extreme, more users bailed out on the interaction.
When human-like features were layered on top of emotionalism, completion rates fell further, reducing the organization’s ability to use emotional engagement as a motivational tool.
The takeaway may be that “there is a sweet spot where some emotion is important, but beyond that emotions can be bad,” as Bhattacharya explains. Furthermore, he observes that “in the retail space, studies have shown anthropomorphism to be useful. But in a nonprofit context, it’s totally empathy-driven and less transactional. If that is the case, maybe these human cues coming from a bot make people feel creepy, and they back off.”
GRAND CHALLENGE INITIATIVE
As the above examples make clear, faculty at the Donald G. Costello College of Business are producing world-class research in line with George Mason University’s core belief that innovative technologies can be instrumental in bringing about the future we want. Helping to drive this work across the institution, the university’s recently-launched Grand Challenge Initiative is dedicated to developing transformative solutions through strategic investments in talent, projects, programs and/or infrastructure.