Experts Matter. Find Yours.
Connect for media, speaking, professional opportunities & more.
Survival analysis: Forecasting lifespans of patients and products
How long will you live? Should you spring for that AppleCare+ warranty for your iPhone? When will your buddy pay you back for that lunch? For centuries, soothsayers have striven to understand the lifespan of things – be they patient longevity, product lifecycles, or even time to loan default. Nowadays, scientists have turned away from reading tea leaves and toward survival analysis – a complex data science method for predicting not only whether an event will happen (the death of a patient, the failure of a product or machine, default on a payment, and so on) but when this event is likely to occur. But it’s problematic. Until now, the tools of survival analysis have only been applicable in certain settings. This is due to the inherent heterogeneity of what is being analyzed: differences in patient lifestyles, demographics, product usage patterns, and so on. New research by Goizueta Business School’s Donald Lee, associate professor of information systems and operations management and of biostatistics and bioinformatics, has yielded a new tool that greatly extends survival analysis to broader use cases. “Historically, scientists have used classic survival analysis tools to predict the lifespan of different things in different fields, from products to patients,” Lee said. “Since the 1950s, the Kaplan-Meier estimator has been the benchmark for analyzing lifetime data, particularly in clinical trials. The next breakthrough came in the 1970s when the Cox proportional hazards model was introduced, which allows researchers to incorporate variables that can affect the predictability of things like patient mortality.” The problem with the existing survival analysis tools, Lee said, is that they make certain assumptions that can skew the predictions if the assumptions are not met. “There are very few existing tools that can incorporate variables without imposing assumptions on how they affect survival, let alone when there are a lot of variables that can also change over time. For example, two iPhones will have different lifespans depending on the temperature at which they are stored, amongst many other factors. But it’s unlikely that storing your phone at 30 degrees will halve its lifespan compared to storing it at 60 degrees. This sort of linear relationship is commonly assumed by existing tools.” Lee’s team developed a new survival methodology based on something called gradient boosting: a machine learning technique that combines decision trees to yield predictions. The method, Lee said, is totally assumption-free (or nonparametric in technical parlance) and can deal with a large number of variables that can change continuously over time, making it significantly more general than existing methods. Nothing like it has been seen until now, he noted. “Calculating the survival rate of anything is super complex because of the variables. Say you want to create an app for a smart watch that monitors the wearer’s vitals and use this information to create a real-time warning indicator for stroke. Doing this accurately is difficult for two reasons,” Lee explained. “First, a large number of variables may be relevant to stroke risk, and the variables can interact in ways that break the assumptions central to existing survival analysis methods. And second, variables like blood pressure vary over time, and it is the recent measurements that are most informative. This introduces an additional time dimension that further complicates things.” The software implementation of Lee’s method, BoXHED, overcomes both issues and allows scientists to develop real-time predictive models for conditions like stroke. The trained model can then be ported to a watch app to tell its wearer if and when they’re likely to have a stroke, a process known as inferencing in machine learning lingo. The implications, Lee said, are huge. “BoXHED now opens the door for modern applications of survival analysis. In previous research, I have looked at the design of early warning mortality indicators for patients with advanced cancer and also for patients in the ICU. These use other methods to make predictions at fixed points in time, but now they can be transformed into real-time warning indicators using BoXHED.” He cited the case of end-stage cancer patients who are often better served by hospice care than by aggressive therapy. “Accurate predictions of survival are absolutely critical for care planning. In previous analyses, we have seen that using existing predictive models to inform end-of-life care planning can potentially avert $1.9 million in medical costs and 1,600 days of unnecessary inpatient care per 1,000 patient visits in the United States. BoXHED is likely to lead to even better results.” Lee’s research paper is forthcoming in the Annals of Statistics. He has also created an open-source software implementation of BoXHED, which can radically improve the accuracy of survival analysis across a breadth of applications. The paper describing BoXHED was published in the International Conference on Machine Learning, and the latest version of the BoXHED software can be found online. If you are a journalist or looking to speak with Donald Lee – simply click on his icon now to arrange an interview or appointment today.
Covering Eating Disorders Week? Let our experts explain how COVID-19 can affect eating disorders
COVID-19 is presenting many different issues across all spectrums of society and life. The experts at Michigan State University took questions and provided answers in order to assist those looking to know more about how COVID-19 can affect eating disorders. Eating disorders can often stem from trauma or stress. Kelly L. Klump, professor in the Department of Psychology and fellow in the Academy for Eating Disorders, answers questions on eating disorders and how the pandemic may trigger or exacerbate this disorder. Q: Is there any evidence that the pandemic triggers eating disorder behaviors among teenagers? We have emerging data on risk for eating disorders during COVID-19. Although data are in the early stage, we are seeing increased weight-shape concerns, increased binge eating and, potentially, increased dietary restriction during COVID-19. These symptoms seem to be increasing in the general population, but results are more consistent in showing exacerbation of these symptoms in individuals with anorexia nervosa (increased restriction and potentially exercise) and bulimia nervosa (increased binge eating and purging). Reasons for these increases aren’t entirely clear, but theories focus on increased stress, increased isolation and, for individuals in recovery, decreased access to care during the pandemic. There are also fears of weight gain due to less activity overall that may fuel concerns about weight/shape and later, eating disorder symptoms. Limitations in access to food during the pandemic also seem to be related to these symptoms. Although, how they are related may vary across eating disorder symptoms. Q: What are some signs parents should be aware of that might indicate eating disorder behaviors or warning signs? These signs would be similar to those that we watch for during non-pandemic times. Decreased food intake, increased exercise and increased discussion of weight concerns are early signs. In addition, if food that was present (particularly high fat/high sugar foods) comes up missing frequently, this could be a sign of binge eating. Because eating disorders are highly comorbid with depression and anxiety, increased signs of these conditions (e.g., sad mood, withdrawal, increased anxiety about a range of concerns) could be early signs, particularly if in combination with the weight/shape/binge eating early signs mentioned above. Q: What should a parent who is concerned their child is exhibiting eating disorder behaviors do to address the issue? The first step is to talk with your teen and listen. Check in on how they are doing generally, but then also let them know about the signs you are seeing and your concerns. Empathic listening is key in these conversations and letting them know that you would like to do whatever is needed to help. They may not be willing to talk the first time they are approached. It might take multiple conversations for them to open up and/or admit that they need help. Q: What resources are available to parents looking to get help for their kids right now? There are some websites that can help parents identify eating disorder specialists in their area, including: • Academy for Eating Disorders. Find an Expert page • National Eating Disorders Association Q: Are families facing obstacles in getting preteens and teenagers help for eating disorder behaviors because of COVID-19 measures? A potential decrease in treatment resources appears to be present for eating disorders and other psychiatric illnesses. Treatment that is available may be in the form of telehealth, which some individuals may find very helpful, while others may feel is not enough. We are still collecting data on treatment availability during COVID-19, so we don’t have great data on availability. But early theories are that treatment access may be decreased. Q: What advice do you have for parents who feel like they are seeing their teenagers’ past eating disorders either reappear or become more severe in light of COVID-19? Seek help and do so early. Catching an increase or exacerbation of symptoms early in the process will increase the chances that you can catch the symptoms before they become more severe. Your teen may need “booster” sessions with treaters that can help them get back on track and help them cope with current stressors. If you are a journalist looking to know more or interview Dr. Klump, then let us help - simply click on her icon now to arrange an interview today.
Study of auto recalls shows carmakers delay announcements until they can 'hide in the herd'
BLOOMINGTON, Ind. - Automotive recalls are occurring at record levels, but seem to be announced after inexplicable delays. A research study of 48 years of auto recalls announced in the United States finds carmakers frequently wait to make their announcements until after a competitor issues a recall - even if it is unrelated to similar defects. This suggests that recall announcements may not be triggered solely by individual firms' product quality defect awareness or concern for the public interest, but may also be influenced by competitor recalls, a phenomenon that no prior research had investigated. Researchers analyzed 3,117 auto recalls over a 48-year period -- from 1966 to 2013 -- using a model to investigate recall clustering and categorized recalls as leading or following within a cluster. They found that 73 percent of recalls occurred in clusters that lasted 34 days and had 7.6 following recalls on average. On average, a cluster formed after a 16-day gap in which no recalls were announced. They found 266 such clusters over the period studied. "The implication is that auto firms are either consciously or unconsciously delaying recall announcements until they are able to hide in the herd," said George Ball, assistant professor of operations and decision technologies and Weimer Faculty Fellow at the Indiana University Kelley School of Business. "By doing this, they experience a significantly reduced stock penalty from their recall." Ball is co-author of the study, "Hiding in the Herd: The Product Recall Clustering Phenomenon," recently published online in Manufacturing and Service Operations Management, along with faculty at the University of Illinois, the University of Notre Dame, the University of Minnesota and Michigan State University. Researchers found as much as a 67 percent stock market penalty difference between leading recalls, which initiate the cluster, and following recalls, who follow recalls and hide in the herd to experience a lower stock penalty. This indicates a "meaningful financial incentive for auto firms to cluster following recalls behind a leading recall announcement," researchers said. "This stock market penalty difference dissipates over time within a cluster. Additionally, across clusters, the stock market penalty faced by the leading recall amplifies as the time since the last cluster increases." The authors also found that firms with the highest quality reputation, in particular Toyota, triggered the most recall followers. "Even though Toyota announces some of the fewest recalls, when they do announce a recall, 31 percent of their recalls trigger a cluster and leads to many other following recalls," Ball said. "This number is between 5 and 9 percent for all other firms. This means that firms are likely to hide in the herd when the leading recall is announced by a firm with a stellar quality reputation such as Toyota. "A key recommendation of the study is for the National Highway Traffic Safety Administration (NHTSA) to require auto firms to report the specific defect awareness date for each recall, and to make this defect awareness date a searchable and publicly available data field in the auto recall dataset NHTSA provides online," Ball added. "This defect awareness date is required and made available by other federal regulators that oversee recalls in the U.S., such as the Food and Drug Administration. Making this defect awareness date a transparent, searchable and publicly available data field may discourage firms from hiding in the herd and prompt them to make more timely and transparent recall decisions." Co-authors of the study were Ujjal Mukherjee, assistant professor of business administration at the Gies College of Business at the University of Illinois who was the lead author; Kaitlin Wowak, assistant professor of IT, analytics, and operations at the Mendoza College of Business at the University of Notre Dame; Karthik Natarajan, assistant professor of supply chain and operations at the Carlson School of Management at the University of Minnesota; and Jason Miller, associate professor of supply chain management at the Broad College of Business at Michigan State University.
Eliminating The Barriers To Telehealth & Patient Retention
During the ongoing national pandemic, healthcare is in a period of rapid evolution, bringing telehealth to the forefront of patient care. Telehealth is a proven strategy to improve health outcomes, but it’s gated behind socioeconomic privilege and leaves behind many of our community’s most vulnerable patients. One such disparity is the inability of many Americans to access digital health care. This silent epidemic affects lives daily. Many patients, especially those in rural communities, face obstacles when trying to get the care they need. From access to reliable transportation and affordable child care to financial instability and lack of culturally competent providers, there is no shortage of hurdles standing in the way of disadvantaged populations accessing quality care. Well-implemented telehealth services can offer a clear path through these common barriers to care while improving health outcomes and boosting patient retention. “We know that mobile health intervention is an effective tool for retaining patients in care, but it’s only as effective as it is accessible,” said Richard Walsh, our CEO. “It would be negligent to assume that every individual has access to the devices, internet, or knowledge necessary to engage in telemedicine.” Like other leaders in the industry, we know telehealth is a privilege, but at Continuud, we believe it should be a right.” As Nathan Walsh, our CXO, said, “During a public health crisis such as this, we have to be proactive in ensuring that underserved communities have access to the care that they need in every way possible.” Through our research and conversations with community health leaders, we have identified 4 common barriers to telehealth success: access to video-ready phones or tablets, access to a reliable & affordable internet connection, an understanding of how to use the device to access services, and trust in technology being used for health services. Our solution is to create a platform that not only solves these problems but also enhances the patient experience and drives the best possible outcome of telehealth intervention. Our platform, Access, provides 8-inch tablets with an unlimited data connection to patients. Each device ships with a secured environment and limited functionality customized by the health care provider to include the tools that patients need to access care. We have created a simple deployment and warehousing solution to make it easy for organizations to get started quickly. Our end-to-end deployment and recall services handle every aspect of the platform so organizations can remain focused on serving their patients. The platform supports patient-by-patient interface customizations, so each patient’s experience is tailored to their unique treatment plan. We have device insurance and same-day replacement built into the program to account for loss, theft, and damaged devices, so organizations will always have access to the inventory they need to serve their clients. At Continuud, we offer an integrated ecosystem designed from the ground up to enable health care providers to work more efficiently toward a common goal of driving positive health outcomes in their communities. Continuud is known throughout Indiana for our innovative approach to connecting high-risk populations to care and implementing strategic technology to help retain and learn from patients so providers can evolve with the needs of their patients. To learn more about our platform, click here to visit our homepage. If you would like to schedule a demo with our team to talk about the platform in greater detail, click here.
As the pandemic enters its second year, household expenses remain top of mind for consumers, as does the cost of auto insurance. It’s one of the reasons why CAA Insurance has seen a dramatic increase over the past ten months in the number of drivers who have signed up for Canada’s only pay-as-you-go auto insurance payment program, CAA MyPace™. New CAA Insurance data reveals, that the number of new CAA MyPace policies between April and December 2020 increased by almost 300 per cent compared to the same period in 2019. “While the growth is remarkable, it reinforces what consumers are telling us, that a one-size-fits-all solution to auto insurance isn’t working for them, especially during these challenging times where many vehicles are not being used as much,” says Matthew Turack, president of CAA Insurance. “CAA MyPace is a one-of-a-kind payment program that lets customers take control of their car insurance costs by giving them the freedom to pay for the distance they drive.” On average, people who switch to CAA MyPace are saving 50 per cent on their auto insurance costs, compared to a traditional policy. The pay-as-you-go program was launched in 2018, and benefits motorists who drive less than 9,000 kilometres per year. In 2020, CAA Insurance provided over $60 million in relief to support policyholders in managing their expenses during the pandemic. The Financial Services Regulatory Authority of Ontario (FSRA) identified CAA Insurance as the leading insurer, providing the highest percentage of rate relief to its policyholders. Nearly one year ago, CAA Insurance led the insurance industry by providing a 10 per cent rate reduction for a 12-month term to all active CAA Auto and Property Insurance policyholders. CAA Insurance will continue to apply this rate reduction in 2021 for its active policyholders. No action is necessary by policyholders to receive this reduction.
Why customers hold the key to a company’s true valuation
When determining a fair valuation for a company—especially in anticipation of an initial public offering (IPO)—investors often rely heavily on “top down” approaches focusing primarily on traditional financial measures to do so. But what if this approach doesn’t paint the full picture? Daniel McCarthy, assistant professor of marketing at Emory’s Goizueta Business School, is building the case that augmenting traditional data sources with customer behavior data gives investors a more accurate company valuation. For the past several years, McCarthy and Peter Fader, professor of marketing at the Wharton School of the University of Pennsylvania, have worked to refine a customer-driven investment methodology they created. “Customer-based corporate valuation (CBCV) simply brings more focus to how individual customer behavior drives the top line,” they explained in “How to Value a Company by Analyzing Its Customers,” an article published in the Harvard Business Review (HBR) earlier this year. “This approach is driving a meaningful shift away from the common but dangerous mindset of ‘growth at all costs,’ towards revenue durability and unit economics—and bringing a much higher degree of precision, accountability, and diagnostic value to the new loyalty economy.” Fader, McCarthy’s PhD advisor while he was at Wharton, had done some of the seminal work on forecasting customer shopping/purchasing behaviors. This helped build baseline expertise for how one could go about the customer-level modeling. McCarthy recognized that this behavioral modeling could be put to good use in a financial setting, if done the right way. “There was this untapped source of intellectual property that’s been accumulating within marketing over the last 30 years,” McCarthy said. While other academics have done some conceptual work in the area, none, McCarthy noted, had done so in a way that was consistent with how financial professionals go about performing corporate valuation. McCarthy and Fader merged these well-validated customer-level models with standard corporate valuation methods, then put their resulting valuation tool head-to-head with alternative approaches. They found that their CBCV model subsequently outperformed. A full article on this subject is attached, within it, you will find key CBCV highlights such as: Using unit economics to more accurately predict revenue forecasts Gaining access to the right data The CBCV model is also good for managers and for customers Working to have publicly traded companies adopt CBCV McCarthy’s work on the CBCV methodology has earned him a number of awards, including the MSI Alden G. Clayton, American Statistical Association, INFORMS, and the Shankar-Spiegel dissertation awards. If you are a journalist covering this topic or if you want to learn more about this work or customer-based corporate valuation – then let our experts help. Daniel McCarthy is an Assistant Professor of Marketing at Emory University's Goizueta School of Business where his research specialty is the application of leading-edge statistical methodology to contemporary empirical marketing problems. If you are looking to contact Daniel – simply click on his icon now to arrange an interview today.
Why are U.S. corporate boards under-diversified?
Research tells us that firms with diverse workforces generally outperform those that do not. And in recent years, corporate America has taken significant strides towards greater heterogeneity in the employee base. But a problem remains at the top. U.S. boardrooms remain overwhelmingly Anglo Saxon and male. No less than 81 percent of the Standard & Poor (S&P) 1500 Index directors in America today are white men. White women account for 11 percent, while ethnic minority men make up 6 percent. Meanwhile, female minority board members account for just 2 percent of the total. For businesses, this is becoming problematic, not least because institutional investors and regulators like the Securities and Exchange Commission have started asking firms to open up about their processes in selecting board members. Where diversity is a criterion, firms are required to be transparent about specifications and frameworks. Shedding light on this issue is new research from Grace Pownall, professor of accounting, and Justin Short, assistant professor of accounting, at Emory University’s Goizueta Business School. Together with Zawadi Lemayian of Washington University, they parsed 12 years of data on gender, ethnicity, and salaries from the S&P 1500 to build a composite picture of who’s who and who’s paid what in U.S. boardrooms. What they found points to a systemic shortage of female and minority executives making it onto shortlists for board appointments. But that’s not all. Once women and minority men do make it onto the board, there’s another roadblock waiting for them: they are not getting promoted at the same rate as their white, male counterparts. There seem to be two complex dynamics at play, said Short: a glass ceiling effect hampering the upward trajectory of Black, female, and other minority executives, and what he and his co-authors call “myopic” bias on the part of corporate America. “We developed two hypotheses that might explain what’s behind the lack of diversity on boards,” explained Short. “The glass ceiling hypothesis comes from what we see as a shortfall of women and ethnic minorities in the workforce relative to white men—so the theory here is that these groups just aren’t getting promoted to the point where they would be considered for board positions.” “The alternative hypothesis we worked on was that there might actually be a plentiful supply, but that companies just don’t see directors from different backgrounds as being as valuable in the same way,” he said. “And we would put this down to some kind of institutional myopia or bias at the very highest echelons of business.” To put these hypotheses to the test, Short and his colleagues first collected demographical data on American board members from a database compiled by Institutional Shareholders Services. Here they were able to determine the gender and ethnicity of individuals. They also ran a simple statistical regression on salaries using data from S&P. Then they compared the two. “Economic theory tells us that if there’s a high demand for diverse directors—women and ethnic minorities—and there’s a low supply of them, then these directors will be able to command higher salaries than others,” said Short. “It’s a simple case of supply and demand, and minorities will come at a greater premium.” Looking at the S&P 1500 data, they found that female and minority directors were indeed getting paid more on average than white male counterparts in other companies. And when they analyzed this more closely, Short and his co-authors found that these salaries were in general being paid by larger, more successful firms. “We can see that women and minorities are commanding higher compensation than the average white male director across the S&P universe of 1500 companies, and it’s the bigger, better paying firms that are hiring them,” Short said. “So that tells us that the top companies are proactively trying to build diversity in their boardrooms. At the same time, it shows there is a deficit of supply in this talent pool—the so-called glass ceiling dynamic.” To understand whether bias or institutional myopia might also be limiting the prospects of Black, female, and ethnic directors, Short et al. also looked at differences in compensation within the same company, and here they found something striking. While they made more on average than the typical white male director in U.S. firms, minority directors were being paid around 3 percent less than their direct counterparts – the white male directors on the same board. All this scrutiny begs the questions: What is going on in the American boardroom? And why is there still such a stark lack of diversity in the upper echelons of business in the U.S. today? “This tells us something important,” said Short. “Once these directors make it to the board, for most of them that’s it. They don’t advance or achieve promotion at the same rate.” This could be due to bias or what Short calls a Rolodex effect: “Maybe it’s because they didn’t go to the same school as the chairman of the board, or weren’t connected socially in the same way, so they don’t appear in the Rolodex of candidates with right or familiar credentials to get promoted within the board,” he said. “We know it’s not about hard skills or aptitudes because the data shows us that women and minority directors typically hold more qualifications than their counterparts. But for whatever reason, once they are on the board, they fail to advance in the same way as white men.” Interestingly, Short and his colleagues found that there was a very small number of women and minority directors sitting on the boards of multiple companies in the U.S. “Pulling it all together, we see that there’s a generalized shortage of women and ethnic group candidates in the U.S.,” Short said. “Successful companies are proactively on the lookout for them and offer higher compensation to attract them. “But there seems to be a glass ceiling effect acting as a bottle neck for talent. We also see that minority directors become a bit stuck once they’re on a board. The upward momentum tails off relative to their white, male colleagues. This could be due to bias or myopic thinking.” All of this should provide rich food for thought for the most senior decision-makers in U.S. enterprises, according to Short and his co-authors. With the pressure on to drive board-level diversity in corporate American, leaders need to be cognizant of the roadblocks or cut-off points to tie to ethnicity and gender. “Diversity is something we urgently need to enable and nurture in the United States. Without diversity, creativity and innovation can stall, and in business you run the risk of deferring to group think—sourcing ideas and perspectives from the same small pool of shared experience or expertise,” said Short. “It’s encouraging to see that diversity has increased over time and the largest companies are proactive. But there are still vast gaps of representation on the board compared to the workforce. There’s still work to be done because diversity in American business should be commonplace.” If you are a journalist looking to cover this research or to learn more about the diversification of corporate boards in America, then let our experts help. Grace Pownall, professor of accounting, and Justin Short, assistant professor of accounting, at Emory University’s Goizueta Business School are both available for interviews; simply click on either expert's icon to arrange a time today.
Big Data Offers New Insights Into Biological Components of Autism Spectrum Disorder
In recently published research, blood sample analysis showed that mothers of children with autism spectrum disorder (ASD) had several significantly different metabolite levels two to five years after they gave birth when compared to mothers of typically developing children. The research team behind this finding was Juergen Hahn, the head of the Department of Biomedical Engineering at Rensselaer Polytechnic Institute, a pioneer in the use of big data to investigate biological components of ASD. Hahn is available to discuss the findings of his recent research, as well as his overall approach to studying autism. Previously, Hahn discovered patterns with certain metabolites in the blood of children with autism that can be used to successfully predict diagnosis. He has since successfully applied his big data approach to evaluating potential ASD treatments. For the recent paper, which Hahn co-authored, his team analyzed blood samples from 30 mothers whose young children had been diagnosed with ASD and 29 mothers of typically developing children. They found differences in several metabolite levels between the two groups of mothers. While the samples analyzed were taken two to five years after pregnancy, these research findings raise the question of whether or not the differences in metabolites may have been present during pregnancy as well. In addition to his specific findings, Hahn is available to discuss the use of big data in improving society's understanding of the biological mechanisms at work in ASD.
Alternative Data Can Offer Insight into GameStop Action on Wall Street
Thomas Shohfi, an assistant professor in the Lally School of Management at Rensselaer, says that looking at alternative data can offer important insights into the turbocharged trading of GameStop. “Volatility in options on GameStop, Nokia, and AMC is exploding,” says Shohfi. “Short sellers are taking massive losses, covering their positions and pushing the stocks even higher. WallStreetBetters keep looking for soft spots in the hedge fund short portfolios to ‘blow up’ their value again and again. It’s a mob-fund mentality.” Working with a former student, Shohfi has analyzed trends from the WallStreetBets subreddit and other social media platforms. He can provide an astute understanding as to how investors have reached this moment, why this has happened, and what the impact that this event will have in the future. “Practitioners, investors, and analysts are always looking for an advantage,” according to Shohfi. “The biggest advantage that we see today is through alternative data. It’s just a rich environment to observe human behavior, incentives, and gender and disclosure effects that affect capital markets like we have seen with GameStop.”
On January 06, America watched with shock as a mob of protesters stormed the gates in Washington, D.C. and invaded the Capitol buildings. For hours, the rioters looted and occupied America’s halls of power and though some were apprehended, many found a way to get out and get back home avoiding arrest. However, media coverage was substantial and some of the protesters were even bold enough to be caught posing for social media. Slowly, authorities are tracking them down, and Dr. Derek Riley, an expert at Milwaukee School of Engineering (MSOE) in the areas of computer science and deep learning, has been explaining how artificial intelligence (AI) technology that’s taught at MSOE is capable of enabling law enforcement's efforts to identify individuals from pictures. "With these AI systems, we’ll show it example photos and we’ll say, 'OK, this is a nose, this is an ear, this is Billy, this is Susie,'" Riley said. "And over lots and lots of examples and a kind of understanding if they guess right or wrong, the algorithm actually tunes itself to get better and better at recognizing certain things." Dr. Riley says this takes huge amounts of data and often needs a supercomputer—like MSOE's "Rosie"— to process it. To get a computer or software to recognize a specific person takes more fine-tuning, Riley says. He says your smartphone may already do this. "If you have a fingerprint scan or facial recognition to open up your phone, that’s exactly what’s happening," Riley said. "So, they’ve already trained a really large model to do all the basic recognition, and then you provide a device with a fingerprint scanning or pictures of your face at the end to be able to fine-tune that model to recognize exactly who you are." Riley says this technology isn't foolproof—he says human intelligence is needed at every step. He added we might be contributing to the data sources some of the technology needs by posting our pictures to social media. "Folks are uploading their own images constantly and that often is the source of the data that is used to train these really, really large systems," Riley said. January 14 – WTMJ, Ch. 4, NBC News The concept of facial recognition and the use of this technology in law enforcement (and several other applications) is an emerging topic – and if you are a reporter looking to cover this topic or speak with an expert, then let us help. Dr. Derek Riley is an expert in big data, artificial intelligence, computer modeling and simulation, and mobile computing/programming. He’s available to speak with media about facial recognition technology and its many uses. Simply click on his icon now to arrange an interview today.
