From Natural Disasters to Deception Detection, Faculty Honored for Research Excellence

On January 15, 2022, an underwater volcano erupted in the South Pacific — an explosive event NASA researchers characterize as 500 times more powerful than the atom bomb dropped on Hiroshima during World War II — devastating the nearby island kingdom of Tonga. Though experts had noted a recent uptick in volcanic activity, the timing and severity of the event caught the Tongan government, and volcanologists around the world, by surprise.  

Natural disasters are inherently arbitrary, but researchers are increasingly turning to mathematical models to better understand and anticipate potentially cataclysmic events. One of those researchers is Moinak Bhaduri, and in a 2020 paper in the multidisciplinary journal Scientific Reports, the assistant professor of Mathematical Sciences outlines an innovative method he believes can help “predict the unpredictable.” 

For his research, Bhaduri focused on two geographically and categorically distinct phenomena: volcanic activity in Hawaii (specifically, Mount Kilauea and its neighbor, Mauna Loa) and hurricanes in the Western Atlantic. Using historical data collected by the U.S. Geological Survey and the National Oceanic and Atmospheric Administration, respectively, he created comprehensive data sets reflecting the duration and intensity of each. 

To parse the data, Bhaduri began with a hidden Markov model (HMM), a traditional mathematical formula used to determine the probability of random events happening over a specified time period. Instead of including the entire series of data points in his calculations, however, he created smaller sets based on shared characteristics — for example, grouping “weak” hurricanes (categories 1 and 2) in one series, and “strong” hurricanes (categories 3, 4 and 5) in another.

The key difference here, Bhaduri explains, is relevance. “In elementary statistics, we’re told that the larger the data set, the better the results. But in a forecasting context, auto-correlation — grouping data based on the degree of similarity — yields more precise outcomes.” 

Bhaduri used the smaller data sets to create what he calls an empirical recurrence rates ratio (ERRR), a value he then used to modify the traditional Markov formula. The result? “The ERRR facilitates the transfer of forecasting-related information from one point process to its possibly related counterpart — a volcano and its neighbor, for example, or two different types of hurricanes — more efficiently,” he says, which makes it easier to predict the likelihood of catastrophic events. Simply put, “this formula works better,” offering a “vision of surer forecasts” that can be used to better monitor — and mitigate — natural disasters. 

Read the Research

For Gopal Krishnan, 2022 marks a major academic milestone: The publication of his 75th research paper. An impressive feat by any definition, it’s particularly noteworthy given the rigors of his field. As the Trustee Professor of Accounting notes, the peer-review process for accounting-related research papers routinely takes two to three years, and “many manuscripts are rejected in the first round.” 

Given such context, it’s no surprise that the most successful research offers insights that advance the discipline. That’s certainly the case with Krishnan’s exploration of the interplay among customer concentration, client business risk and audit complexity. Published in the American Accounting Association’s Journal of Management of Accounting Research in 2018, his paper has been downloaded more than 1,200 times via the Social Science Research Network (SSRN), an esteemed online repository for academic research. 

“Although cross-industry relationships along the supply chain are important in assessing firm performance and valuation,” Krishnan explains, “there’s little research exploring the implications of major customer relationships with external auditors.” He sought to fill that void with a deeper understanding of how a supplier’s customer concentration — that is, the degree to which a firm’s revenue is dependent on individual customers — affects audit complexity and quality. 

Conventional wisdom cautions companies against over-reliance; according to Forbes, if even 10% of revenue is attributable to a single client, it poses a significant business risk. But Krishnan found that when it comes to audits, the opposite is true.  

“Suppliers with more concentrated customer bases spend less on audit fees, particularly when the supplier and client use the same auditor,” he says, noting that audit-related costs decreased by 6.73% per standard deviation increase in customer concentration. “This evidence is consistent with reduced audit effort due to efficiency gains in the audit process, most likely achieved through collaboration and information-sharing.” 

Although higher audit fees are typically associated with higher audit quality, Krishan found that supplier audits are not adversely affected by these cost savings. Using restatements — that is, revised documents issued to correct previous audit errors — as a proxy for overall quality, he found the likelihood of a restatement decreased by 11.83% per standard deviation, suggesting that “audit quality increases with customer concentration.” 

With few other studies exploring the impact of supply-chain relationships on audit complexity and quality, Krishnan’s research “provides new managerial insights on the costs and benefits of major customer relationships for supplier firms — a topic of growing interest among both academics and practitioners.” 

Read the Research

At first glance, it seems unlikely that a teenager’s test scores could have geopolitical significance. But in today’s increasingly globalized and knowledge-based economy, a well-educated workforce is seen as essential in developing and maintaining a nation’s influence and prosperity.  

As a result, policymakers are paying closer attention to standardized tests that measure student achievement on a global scale. One such test — the Program for International Student Assessment (PISA), which is administered every three years to fifteen-year-olds in nearly 80 countries — has proven particularly influential, with America’s lackluster performance a growing cause of concern: After the U.S. ranked 31st in math on the 2009 test, then-Secretary of Education Arne Duncan urged the nation to “face the brutal truth that we’re being out-educated.” 

Jeff Livingston, however, thinks the handwringing over PISA rankings is overblown. Indeed, the Gibbons Research Professor of Economics believes standardized tests aren’t an accurate index of student ability, and in a 2019 paper in the journal American Economics Review: Insights, he explains why: “Students in different countries have different levels of intrinsic motivation to perform well.”  

PISA tests, and others like it, Livingston explains, are “low stakes” events, as “the results of these tests have no direct bearing on students’ lives.” For U.S. students, he says, this often translates into indifference, as “the American mentality is ‘What’s in it for me?’” In other countries, however, test results are regarded as a source of national pride, creating significant cultural pressure for students’ success. China, for example, which highly values academic achievement, has historically outperformed the U.S. on standardized tests. 

To better explore the interplay between effort and ability, Livingston conducted an experiment with tenth grade students at two schools in the U.S. and four schools in Shanghai. All students were given 25 minutes to complete a 25-question math test. Immediately before starting the exam, randomly selected groups were offered a financial incentive. “U.S. students were given an envelope with $25 in one-dollar bills, and were told that the money was theirs, but that we would take away one dollar for each question left unanswered,” Livingston explains. In Shanghai, students received identical instructions, and the equivalent amount in renminbi (RMB).  

The outcomes, he says, were “striking.” Compared with control groups, scores for American students who received money “increased substantially,” while those of Chinese students remained the same. This suggests that “U.S. students are more responsive than Shanghai students to incentives for effort because they are less motivated at baseline.” In other words, American students have fewer cultural incentives to achieve. 

Inherent differences in attitude make it impossible to accurately compare aptitude, Livingston cautions, and U.S. policymakers should refrain from putting too much stock in international rankings. “Policy reforms that ignore the role of intrinsic motivation to perform well on standardized tests may be misguided and have unintended consequences.” 

Read the Research

Can a computer mouse double as a lie detector? For Jeffrey Proudfoot and his co-researchers, the answer is a resounding yes. 

In research recognized by the Journal of the Association for Information Systems as its 2019 Best Paper of the Year, the associate professor of Information and Process Management introduces a novel approach to detecting deception in the workplace: Using mouse-cursor movements to determine if an employee is telling the truth.  

According to Proudfoot, individuals who engage in adverse business behaviors — whether deliberately (e.g., corporate espionage) or inadvertently, such as improperly storing proprietary data on personal devices — constitute a “substantial risk” to their organizations. “Security incidents facilitated by information concealment are prevalent,” he says, costing businesses “hundreds of thousands of dollars on average per incident.” 

Yet, identifying individuals whose actions could jeopardize their employers is a problematic process. “The detection of concealed information is difficult, expensive, error-prone and time-consuming,” he explains. Traditional methods, such as monitoring a person’s actions and environment (including computer use and social media activity) and direct questioning via polygraph, are “plagued with high numbers of false positives.”  

This research, however, suggests a more accurate alternative: Tracking mouse-cursor movements of individuals as they respond to an online questionnaire. This survey, adapted from the polygraph-based Concealed Information Test (CIT), is specifically designed to reveal an individual’s knowledge of a particular crime. As “hand movements can show powerful traces of internal cognitive processes,” the respondent’s guilt or innocence is indicated by their overall response time and mouse-cursor trajectories. 

Proudfoot and his co-researchers administered their questionnaire to 66 subjects, 30 of whom had committed a mock theft by downloading “classified” credit card information. They found that guilty respondents took almost twice as long to answer as their truthful counterparts. “People exert more cognitive effort when concealing information,” he says, which makes less working memory available for other tasks, including motor response.  

The mouse-cursor trajectories of guilty respondents reflect this cognitive dissonance. “When people see a question regarding an adverse behavior they’re guilty of committing,” Proudfoot explains, “their attention briefly turns to the truthful answer.” In other words, if asked “Have you stolen any classified information?” the mouse-cursor trajectories of guilty respondents orient initially towards “yes” before ultimately choosing “no.” 

In addition to uncovering a “theoretically sound and mass-deployable technique” for protecting businesses from employee malfeasance, the study strengthened Proudfoot’s existing portfolio of cybersecurity research and led to an opportunity to join MIT’s Cybersecurity at MIT Sloan (CAMS) group, a consortium dedicated to exploring critical cybersecurity issues. As a CAMS research affiliate, Proudfoot is proudly “working with some of the top security scholars and industry experts in the world.”  

Read the Research