Humans make nearly 35,000 decisions every day, from whether it’s safe to cross the road to what to have for lunch. Every decision involves weighing the options, remembering similar past scenarios, and feeling reasonably confident about the right choice. What may seem like a snap decision actually comes from gathering evidence from the surrounding environment. And often the same person makes different decisions in the same scenarios at different times.

Neural networks do the opposite, making the same decisions each time. Now, Georgia Tech researchers in Associate Professor Dobromir Rahnev’s lab are training them to make decisions more like humans. This science of human decision-making is only just being applied to machine learning, but developing a neural network even closer to the actual human brain may make it more reliable, according to the researchers.

In a paper in Nature Human Behaviour, “The Neural Network RTNet Exhibits the Signatures of Human Perceptual Decision-Making,” a team from the School of Psychology reveals a new neural network trained to make decisions similar to humans.

Decoding Decision

“Neural networks make a decision without telling you whether or not they are confident about their decision,” said Farshad Rafiei, who earned his Ph.D. in psychology at Georgia Tech. “This is one of the essential differences from how people make decisions.” 

Large language models (LLM), for example, are prone to hallucinations. When an LLM is asked a question it doesn’t know the answer to, it will make up something without acknowledging the artifice. By contrast, most humans in the same situation will admit they don’t know the answer. Building a more human-like neural network can prevent this duplicity and lead to more accurate answers.

Making the Model

The team trained their neural network on handwritten digits from a famous computer science dataset called MNIST and asked it to decipher each number. To determine the model’s accuracy, they ran it with the original dataset and then added noise to the digits to make it harder for humans to discern. To compare the model performance against humans, they trained their model (as well as three other models: CNet, BLNet, and MSDNet) on the original MNIST dataset without noise, but tested them on the noisy version used in the experiments and compared results from the two datasets. 

The researchers’ model relied on two key components: a Bayesian neural network (BNN), which uses probability to make decisions, and an evidence accumulation process that keeps track of the evidence for each choice. The BNN produces responses that are slightly different each time. As it gathers more evidence, the accumulation process can sometimes favor one choice and sometimes another. Once there is enough evidence to decide, the RTNet stops the accumulation process and makes a decision. 

The researchers also timed the model’s decision-making speed to see whether it follows a psychological phenomenon called the “speed-accuracy trade-off” that dictates that humans are less accurate when they must make decisions quickly. 

Once they had the model’s results, they compared them to humans’ results. Sixty Georgia Tech students viewed the same dataset and shared their confidence in their decisions, and the researchers found the accuracy rate, response time, and confidence patterns were similar between the humans and the neural network.

“Generally speaking, we don't have enough human data in existing computer science literature, so we don't know how people will behave when they are exposed to these images. This limitation hinders the development of models that accurately replicate human decision-making,” Rafiei said. “This work provides one of the biggest datasets of humans responding to MNIST.” 

Not only did the team’s model outperform all rival deterministic models, but it also was more accurate in higher-speed scenarios due to another fundamental element of human psychology: RTNet behaves like humans. As an example, people feel more confident when they make correct decisions. Without even having to train the model specifically to favor confidence, the model automatically applied it, Rafiei noted. 

“If we try to make our models closer to the human brain, it will show in the behavior itself without fine-tuning,” he said.

The research team hopes to train the neural network on more varied datasets to test its potential. They also expect to apply this BNN model to other neural networks to enable them to rationalize more like humans. Eventually, algorithms won’t just be able to emulate our decision-making abilities, but could even help offload some of the cognitive burden of those 35,000 decisions we make daily.

Dietra Trent, executive director for the White House Initiative on HBCUs, attended the first day’s sessions. Frances Williams, vice president for Research and Sponsored Programs at CAU and one of the organizers and co-leaders of the HBCU CHIPS Network, welcomed the attendees and outlined the meeting agenda.  

Motherhood in the U.S. can be dangerous. The nation spends more on healthcare than any other high-income country. But women giving birth here — particularly Black women, and particularly in Georgia — are more likely to die in childbirth. A big reason for this maternal mortality crisis is a lack of access to obstetric care.

“Georgia has a problem with access to care — the whole country does,” said Meghan Meredith, a fourth-year Ph.D. student in the H. Milton Stewart School of Industrial and Systems Engineering (ISyE) who has spent much of her academic career studying the problem, which is particularly acute in rural, lower-income places.

Many of these places have been designated “maternity care deserts” by the March of Dimes. If a county doesn’t have any obstetric care or providers, it’s considered a desert. Another commonly used measure is whether a pregnant woman lives within 50 miles of critical care obstetrics (CCO). 

These measures are often referred to in academic literature and popular media to highlight a lack of healthcare access, and by public policy leaders trying to address the issue. But it’s become evident to Georgia Tech researchers that they just don’t add up.

“These measures don’t capture the complete picture,” said Meredith. “They aren’t an accurate representation of access to care.”

And that’s what concerns Meredith and her faculty advisor, ISyE Assistant Professor Lauren Steimle.

“We’ve been interested in access to maternal care for a long time, and in countless news stories, the maternity care desert measure is reported on,” Meredith said. “We recognized the limitations, so we thought, ‘Let’s write a paper that explains how this measure is not a complete representation of access.’”

They published their work recently in the journal BMC Health Services Research.

Modeling the Landscape

To study these measures of access, Meredith and Steimle used the same kind of computer-based mathematical model that helps companies decide where to place a new distribution center, retail outlet, or even electric car charging stations: a facility location model.

“This model helps us determine where to place facilities, so demand is sufficiently covered with the fewest number of facilities,” said Steimle. “There are tons of potential applications for this model, but we’re using it for healthcare.” For this study, they used the model to identify where Georgia would need to expand healthcare facilities to improve access under the commonly used measures. 

Here’s some of what the researchers found:

• Of the 1,910,308 reproductive-age women in Georgia, 104,158 (5.5%) live in maternity care deserts, while 150,563 (7.9%) live more than 50 miles from CCO services; 38,202 live in both situations.

• Fifty-six counties in Georgia meet current “maternity care desert” measures, which means eliminating these deserts would require 56 new obstetrics hospitals. That would increase the number of obstetric hospitals statewide from 83 to 139 (a 67% increase). 

• Strategically expanding 16 hospitals (a 19% increase) would reduce the number of reproductive-age women living in deserts by half.

• 82% of reproductive-age women designated as living in deserts live within 25 miles from an obstetric hospital.

The researchers conclude that policymakers should be warned: Using the maternity care desert measure alone as a basis for where and how to invest in healthcare resources isn’t a great idea.

“If we really want to improve pregnancy outcomes, our measures of access should promote risk-appropriate and regionalized care systems,” Steimle said.

Turns out, Georgia is already headed in that direction.

Counting Counties: One Size Doesn’t Fit All

To illustrate the problems with the maternity care desert measure, Steimle compared Georgia with a very different state on the opposite side of the U.S.: Nevada.

“A major problem with the maternity care desert measure is its emphasis on county-by-county infrastructure,” she said. “It’s a one-size-fits-all approach that doesn’t tell the whole story about access to care.”

For example, Georgia has 159 counties and more than three times the population of Nevada. Meanwhile, Nevada has twice the square mileage of Georgia — and 16 very large counties. 

At 18,147 square miles, Nye County is Nevada’s largest, and it’s been labeled a maternity care desert. There’s also lots of actual desert in Nye, which is larger than nine U.S. states. So, it’s difficult to accurately compare a vast jurisdiction like Nye with, say, central Georgia’s Lamar County. Lamar, also labeled a desert, is a mere 185 square miles in size. It's also surrounded by counties that are veritable oases of care.

“A lot of people in Georgia may be falsely labeled as not having access, at least geographically speaking, when in fact they have services nearby,” noted Steimle. “Meanwhile, in a state like Nevada, some women may be labeled as having access, but might be very far from obstetric hospitals in their county.”

Steimle also point out that measuring access on a county-by-county basis ignores efforts to coordinate care across the whole state. “The maternity care desert model doesn’t hold up. And it doesn't reflect Georgia’s approach to a regionalization system.”

Since 2009, the Georgia Department of Public Health has organized the state into six geographic perinatal regions (the perinatal period covers pregnancy, childbirth, and early postpartum). The idea is to coordinate the delivery of health services to ensure people in all regions have access to risk-appropriate maternal care.

Build a Better Model

Each of Georgia’s perinatal regions has a “hub” — a major care center serving as an administrative unit to enable the coordination and delivery of maternal care services. For example, The Emory Perinatal Regional Center at Emory University Hospital is the coordinating center for the 39-county metro Atlanta region. 

This regionalization strategy also tries to address the problem of hospital closures, a troubling trend that leads to more deserts. In Georgia, 12 hospitals have closed since 2013; 18 rural hospitals are currently at risk of closure. And this new Georgia Tech study indicates that Georgia would somehow need to add 56 new facilities to eliminate the state’s maternity care deserts — at least by the standards used by the March of Dimes.

“Eliminating maternity care deserts in Georgia would mean adding a larger number of obstetrics facilities to make sure every county has an obstetric hospital,” Steimle said. “But this is likely unrealistic with the current economic forces pushing hospitals to close their obstetric units. With that many facilities in Georgia, some facilities would have a very small number of deliveries, which is not economically sustainable.”

In other words, eliminating maternity care deserts in Georgia wouldn’t sufficiently address the larger problems related to access to care. Instead, Steimle and Meredith advocate for approaches that simultaneously consider the different dimensions of an ideal maternal healthcare system, not just access alone.

For this initial study, Steimle and Meredith just focused on spatial access. They haven’t yet addressed the complex issues of racial disparities, insurance access, or other hurdles facing reproductive-age women in Georgia. That may be coming.

“This is a start,” Steimle said. “Our future work entails thinking about how to come at this with the goal of maximizing or improving outcomes for women.”

And as policy leaders across the country begin to address the maternal mortality crisis, Steimle believes her team’s approach using more sophisticated tools can be helpful. So far, they’ve shared their results with the Centers for Disease Control and Prevention, and members of the Georgia, Iowa, and Nevada departments of public health.

“How do we make measurements that point us toward our end goals? Our tools as mathematical modelers can really help us think through the system holistically and think through strategies before trying them in the real world,” Steimle said. “Think of it as a policy sandbox.”

CITATION: Meghan Meredith, Lauren Steimle, and Stephanie Radke. “The implications of using maternity care deserts to measure progress in access to obstetric care: a mixed-integer optimization analysis.” BMC Health Services Research (June 2024)

doi.org/10.1186/s12913-024-11135-4

New research from Georgia Tech is giving scientists more control options over generative artificial intelligence (AI) models in their studies. Greater customization from this research can lead to discovery of new drugs, materials, and other applications tailor-made for consumers.

The Tech group dubbed its method PRODIGY (PROjected DIffusion for controlled Graph Generation). PRODIGY enables diffusion models to generate 3D images of complex structures, such as molecules from chemical formulas. 

Scientists in pharmacology, materials science, social network analysis, and other fields can use PRODIGY to simulate large-scale networks. By generating 3D molecules from multiple graph datasets, the group proved that PRODIGY could handle complex structures.

In keeping with its name, PRODIGY is the first plug-and-play machine learning (ML) approach to controllable graph generation in diffusion models. This method overcomes a known limitation inhibiting diffusion models from broad use in science and engineering.

“We hope PRODIGY enables drug designers and scientists to generate structures that meet their precise needs,” said Kartik Sharma, lead researcher on the project. “It should also inspire future innovations to precisely control modern generative models across domains.” 

PRODIGY works on diffusion models, a generative AI model for computer vision tasks. While suitable for image creation and denoising, diffusion methods are limited because they cannot accurately generate graph representations of custom parameters a user provides.

PRODIGY empowers any pre-trained diffusion model for graph generation to produce graphs that meet specific, user-given constraints. This capability means, as an example, that a drug designer could use any diffusion model to design a molecule with a specific number of atoms and bonds.

The group tested PRODIGY on two molecular and five generic datasets to generate custom 2D and 3D structures. This approach ensured the method could create such complex structures, accounting for the atoms, bonds, structures, and other properties at play in molecules. 

Molecular generation experiments with PRODIGY directly impact chemistry, biology, pharmacology, materials science, and other fields. The researchers say PRODIGY has potential in other fields using large networks and datasets, such as social sciences and telecommunications.

These features led to PRODIGY’s acceptance for presentation at the upcoming International Conference on Machine Learning (ICML 2024). ICML 2024 is the leading international academic conference on ML. The conference is taking place July 21-27 in Vienna.

Assistant Professor Srijan Kumar is Sharma’s advisor and paper co-author. They worked with Tech alumnus Rakshit Trivedi (Ph.D. CS 2020), a Massachusetts Institute of Technology postdoctoral associate.

Twenty-four Georgia Tech faculty from the Colleges of Computing and Engineering will present 40 papers at ICML 2024. Kumar is one of six faculty representing the School of Computational Science and Engineering (CSE) at the conference.

Sharma is a fourth-year Ph.D. student studying computer science. He researches ML models for structured data that are reliable and easily controlled by users. While preparing for ICML, Sharma has been interning this summer at Microsoft Research in the Research for Industry lab.

“ICML is the pioneering conference for machine learning,” said Kumar. “A strong presence at ICML from Georgia Tech illustrates the ground-breaking research conducted by our students and faculty, including those in my research group.”

Visit https://sites.gatech.edu/research/icml-2024 for news and coverage of Georgia Tech research presented at ICML 2024.

A new machine learning (ML) model created at Georgia Tech is helping neuroscientists better understand communications between brain regions. Insights from the model could lead to personalized medicine, better brain-computer interfaces, and advances in neurotechnology.

The Georgia Tech group combined two current ML methods into their hybrid model called MRM-GP (Multi-Region Markovian Gaussian Process). 

Neuroscientists who use MRM-GP learn more about communications and interactions within the brain. This in turn improves understanding of brain functions and disorders.

“Clinically, MRM-GP could enhance diagnostic tools and treatment monitoring by identifying and analyzing neural activity patterns linked to various brain disorders,” said Weihan Li, the study’s lead researcher. 

“Neuroscientists can leverage MRM-GP for its robust modeling capabilities and efficiency in handling large-scale brain data.” 

MRM-GP reveals where and how communication travels across brain regions. 

The group tested MRM-GP using spike trains and local field potential recordings, two kinds of measurements of brain activity. These tests produced representations that illustrated directional flow of communication among brain regions. 

Experiments also disentangled brainwaves, called oscillatory interactions, into organized frequency bands. MRM-GP’s hybrid configuration allows it to model frequencies and phase delays within the latent space of neural recordings.

MRM-GP combines the strengths of two existing methods: the Gaussian process (GP) and linear dynamical systems (LDS). The researchers say that MRM-GP is essentially an LDS that mirrors a GP.

LDS is a computationally efficient and cost-effective method, but it lacks the power to produce representations of the brain. GP-based approaches boost LDS's power, facilitating the discovery of variables in frequency bands and communication directions in the brain.

Converting GP outputs into an LDS is a difficult task in ML. The group overcame this challenge by instilling separability in the model’s multi-region kernel. Separability establishes a connection between the kernel and LDS while modeling communication between brain regions.

Through this approach, MRM-GP overcomes two challenges facing both neuroscience and ML fields. The model helps solve the mystery of intraregional brain communication. It does so by bridging a gap between GP and LDS, a feat not previously accomplished in ML.

“The introduction of MRM-GP provides a useful tool to model and understand complex brain region communications,” said Li, a Ph.D. student in the School of Computational Science and Engineering (CSE). 

“This marks a significant advancement in both neuroscience and machine learning.”

Fellow doctoral students Chengrui Li and Yule Wang co-authored the paper with Li. School of CSE Assistant Professor Anqi Wu advises the group. 

Each MRM-GP student pursues a different Ph.D. degree offered by the School of CSE. W. Li studies computer science, C. Li studies computational science and engineering, and Wang studies machine learning. The school also offers Ph.D. degrees in bioinformatics and bioengineering.

Wu is a 2023 recipient of the Sloan Research Fellowship for neuroscience research. Her work straddles two of the School’s five research areas: machine learning and computational bioscience. 

MRM-GP will be featured at the world’s top conference on ML and artificial intelligence. The group will share their work at the International Conference on Machine Learning (ICML 2024), which will be held July 21-27 in Vienna. 

ICML 2024 also accepted for presentation a second paper from Wu’s group intersecting neuroscience and ML. The same authors will present A Differentiable Partially Observable Generalized Linear Model with Forward-Backward Message Passing.

Twenty-four Georgia Tech faculty from the Colleges of Computing and Engineering will present 40 papers at ICML 2024. Wu is one of six faculty representing the School of CSE who will present eight total papers.

The group’s ICML 2024 presentations exemplify Georgia Tech’s focus on neuroscience research as a strategic initiative.  

Wu is an affiliated faculty member with the Neuro Next Initiative, a new interdisciplinary program at Georgia Tech that will lead research in neuroscience, neurotechnology, and society. The University System of Georgia Board of Regents recently approved a new neuroscience and neurotechnology Ph.D. program at Georgia Tech. 

“Presenting papers at international conferences like ICML is crucial for our group to gain recognition and visibility, facilitates networking with other researchers and industry professionals, and offers valuable feedback for improving our work,” Wu said. 

“It allows us to share our findings, stay updated on the latest developments in the field, and enhance our professional development and public speaking skills.”

Visit https://sites.gatech.edu/research/icml-2024 for news and coverage of Georgia Tech research presented at ICML 2024.

OrthoPreserve, a startup founded by Georgia Tech alumnus Jonathan Schwartz, is striving to make debilitating meniscus injuries a thing of the past and to address the long-term complications associated with meniscus tears, a common issue among athletes and aging adults.

The meniscus is a C-shaped structure that acts as a shock absorber and stabilizer in the knee, distributing impact and protecting bone cartilage from deteriorating. Meniscus injuries are frequent in sports — aggressive movements from running or rapidly adjusting leg positions can cause the meniscus to overextend and tear. The risk of injury also increases with age. Over time, degeneration in the knee can wear down the meniscus, making it weaker and easier to tear, even during normal daily activities. 

“Right now, the main treatment for meniscus injuries is surgery to cut out the damaged part of the meniscus to relieve pain and impairment, but pain often returns within a few years due to degradation,” Schwartz said. “Once the meniscus is cut, the only treatments are pain medication, injections, physical therapy, and even knee replacement.” 

According to Schwartz, over half of meniscus surgery patients, regardless of age, get early-onset arthritis because a severed meniscus can no longer cushion the knee as effectively.

“Patients don’t like hearing that the only treatment available is to cut out their meniscus — which will accelerate arthritis development,” Schwartz said. “Our mission is to use our meniscus implant to help people return to activity quickly and avoid the long-term consequences of surgery.” 

The implant functions like a natural meniscus in the knee. Credit: Jonathan Schwartz

From Academia to Industry

Schwartz’s journey to entrepreneurship began at Georgia Tech, where he enrolled as a bioengineering Ph.D. student and joined the Biofluids and Medical Device Research Group, led by David Ku, Regents’ Professor and Lawrence P. Huang Endowed Chair for Engineering Entrepreneurship. There, Schwartz and Ku discussed potential areas of need in the medical device field. Growing up in a family that experienced numerous knee and meniscus issues and having a longstanding interest in orthopedics, Schwartz decided to design an implant that could mimic the properties of the natural meniscus. 

After two years in the program developing a prototype, Schwartz left Georgia Tech with a master’s degree to work in the biopharmaceutical industry. In 2021, he co-founded OrthoPreserve with orthopedic clinicians Cyrus Kump, M.D., and Max Guillot, PA-C.

While working full-time, Schwartz both improved upon his prototype and developed his company on the side. But when he secured a National Institutes of Health research grant for his implant in 2023, he left his job to pursue OrthoPreserve exclusively.

An Innovative Implant 

Made from a biocompatible hydrogel material reinforced with high-strength fibers, the implant mimics the shape, structure, and biomechanical properties of the meniscus. It is designed to restore normal joint mechanics and provide long-term protection and stability to the knee joint.

The company has also developed a minimally invasive surgical technique, allowing patients to recover quickly after a short surgery entailing just four or five small incisions. During surgery, the natural meniscus is removed, and the meniscus implant is attached to the knee joint similarly to the natural meniscus. 

OrthoPreserve recently completed its first animal study in sheep, with promising results. 

"All the sheep were able to walk normally within two or three weeks, and the implants held up without breaking down,” Schwartz said. “The cartilage was protected at the same level as the natural meniscus on the sheep’s other knees.”

This year, Schwartz met with the FDA and laid out a testing plan that will allow the company to start trials in humans within two years. The next animal study will last six to 12 months and will assess the long-term protective capabilities of the implant. After testing, the next steps will be to start human clinical trials and refine manufacturing techniques.

A Campus Community 

OrthoPreserve’s home base is BioSpark Labs, a life sciences incubator space in Georgia Tech’s Science Square district. It is a collaborative startup environment that provides office space, shared laboratories, and equipment.

“We built BioSpark Labs because there was a need for more wet lab space and cleanroom space in the Atlanta area, and we wanted to keep biotech research coming out of Atlanta, and especially Georgia Tech,” said Noriko Walker, associate director of Portfolio Management in Georgia Tech’s Real Estate Office. “We kept hearing researchers say they have companies in Boston or San Francisco, and we wanted to provide a place where they could stay, do their research, and grow their companies here instead.”

For Schwartz, one of the most valuable benefits of BioSpark is access to Georgia Tech’s core facilities. 

“It is a great atmosphere that comes with the vast resources available for research at Georgia Tech,” he said. “We also plan to take advantage of our location and look to Georgia Tech talent when we start hiring.”

Schwartz has several goals for OrthoPreserve. He wants to address the significant clinical need to help meniscus injury patients recover quickly, remain symptom-free, and avoid getting arthritis or a knee replacement. He wants to reduce the high costs associated with ongoing pain treatment and invasive surgeries. Improving patients’ quality of life is also on the list. 

“With serious meniscus injuries, you can't get the joy you used to from being active, and often you just have to sit on the sidelines and watch. We hope our implant can help these patients resume their normal activities and do the things they love.”

A new data visualization tool designed by a Georgia Tech Ph.D. student is helping a team of microbial ecologists, geobiologists, and oceanographers gain more insight into how deep-sea microorganisms interact within their environment.

What began as an internship at NASA turned into a unique opportunity for fourth-year Ph.D. student Adam Coscia. Coscia worked under the supervision of an interdisciplinary team of collaborative researchers from Caltech, the Jet Propulsion Laboratory (JPL) Caltech manages for NASA and the ArtCenter College of Design.

Coscia’s mentors recommended him to a Caltech research team led by Victoria Orphan, a renowned microbial ecologist who studies microbial communities in the ocean and how they function within habitats in deep seafloor sediments. 

Orphan and her team, the Orphan Lab at Caltech, have conducted their research since 2004. They recently decided to take a data visualization approach to record their findings and plan future expeditions.

“Historically, our data sets have been discrete and have lived in separate Excel spreadsheets,” Orphan said. “Maybe at the end, we’ll do some statistical analysis to find correlations in that data. Then we compare those to our maps. We didn’t have a way of consolidating everything under one umbrella that allows us to learn more about these ecosystems.”

Orphan said her team typically takes one or two research expeditions off the California coast annually. They spend three weeks using remotely operated vehicles (ROVs) to collect sediment samples from the ocean floor. Because time is at a premium, identifying the locations of the best samples is crucial.

Orphan is also an adjunct scientist at the Monterey Bay Aquarium Research Institute (MBARI) and works with the Seafloor Mapping Lab. The lab uses an ROV-mounted low-altitude survey system to produce detailed maps of seafloor topography. 

To help the Orphan Lab work effectively with topographic and photographic data, Coscia designed DeepSee, an interactive web browser that can annotate and chart data using 3D visualization models and environmental maps.

“The idea is once you have the samples, and you’re interested in a specific area with prior samples, you can go in and annotate on the map where to collect samples next with our drawing tool,” Coscia said.

“We focused on the exploration and notetaking process with maps and data and having new ways of visualizing it. Scientists can draw and map out all their samples in real time. They can reference specific data much easier and determine where the team should go to get the best samples.”

The Orphan Lab has taken DeepSee live onboard its ship for its two most recent expeditions. Orphan has noticed an increased efficiency in expedition planning.

“The infrastructure put in place by Adam will make this an enabling tool not only for my group but for other oceanographers and scientists in other fields — anywhere there is a spatial distribution of information you want to connect to other metadata,” she said.

Orphan brings new researchers into her lab at Caltech every year, and DeepSee has accelerated the process of getting newcomers up to speed.

“We can onboard them much easier and give them a sense of what data is available and where we’ve collected information in a way that’s much clearer than having them refer to an Excel spreadsheet,” she said.

DeepSee also creates 3D data models under the sea floor using data interpolation, which estimates new data points based on the range of a set of known data points. Using the known data points, DeepSee fills in the blanks of the estimated data quality the researchers may find in nearby locations or further underneath the surface where samples were collected.

“You would never see anything visually below the sea floor,” Coscia said. “You’d have to go dig. But our 3D models show you that you might have data suggesting a hotspot just a few feet below the floor. That tells you where to sample next.”

Coscia aims to incorporate machine learning (ML) models into a future version of DeepSee that will use collected data to predict future sites for sampling. However, ML model accuracy requires significantly more data.

Coscia hopes the current version of the tool catches on so researchers can more easily incorporate machine learning into their work.

For now, the current version has plenty of uses, he said.

“Being able to organize and see your data, especially with maps, is always valuable,” he said. “My passion is helping researchers and scientists see their data in new and valuable ways.”

Coscia authored a paper on developing DeepSee, which he presented in May at the Conference on Human Factors in Computing Systems (CHI) in Honolulu, Hawaii.

Chaouki Abdallah, Georgia Tech's executive vice president for Research (EVPR), has been named the new president of the Lebanese American University in Beirut.  

Abdallah, MSECE 1982, Ph.D. ECE 1988, has served as EVPR since 2018; in this role, he led extraordinary growth in Georgia Tech's research enterprise. Through the work of the Georgia Tech Research Institute, 10 interdisciplinary research institutes (IRIs), and a broad portfolio of faculty research, Georgia Tech now stands at No. 17 in the nation in research expenditures — and No. 1 among institutions without a medical school.  

Additionally, Abdallah has also overseen Tech's economic development activities through the Enterprise Innovation Institute and such groundbreaking entrepreneurship programs as CREATE-X, VentureLab, and the Advanced Technology Development Center. 

Under Abdallah's strategic, thoughtful leadership, Georgia Tech strengthened its research partnerships with historically Black colleges and universities, launched the New York Climate Exchange with a focus on accelerating climate change solutions, established an AI Hub to boost research and commercialization in artificial intelligence, advanced biomedical research (including three research awards from ARPA-H), and elevated the Institute's annual impact on Georgia's economy to a record $4.5 billion.  

Prior to Georgia Tech, Abdallah served as the 22nd president of the University of New Mexico (UNM), where he also had been provost, executive vice president of academic affairs, and chair of the electrical and computer engineering department. At UNM, he oversaw long-range academic planning, student success initiatives, and improvements in retention and graduation rates. 

A national search will be conducted for Abdallah's replacement. In the coming weeks, President Ángel Cabrera will name an interim EVPR. 

The Children's Healthcare of Atlanta Pediatric Technology Center at Georgia Tech (PTC) is excited to announce that Gian-Gabriel Garcia will serve as its Pillar 1 Co-Lead. Pillar 1 focuses on data science, machine learning, and artificial intelligence. In his new role, Garcia’s responsibilities will include setting the pillar’s strategy and vision, selecting and managing projects, overseeing various pillar activities, and working collaboratively across research groups and institutions. He will also identify cutting-edge technology and engineering solutions to implement priority projects while balancing the pragmatism and feasibility of these approaches.

The PTC brings clinical experts together with Georgia Tech scientists and engineers to develop technological solutions to problems in the health and care of children. The Center provides extraordinary opportunities for interdisciplinary collaboration in pediatrics, creating breakthrough discoveries that often can only be found at the intersection of multiple disciplines. 

Garcia will work under the leadership of PTC Co-Directors Dr. Stanislav Emelianov (Georgia Tech) and Dr. Wilbur Lam (Children’s) of Georgia Tech’s Wallace H. Coulter Department of Biomedical Engineering at Georgia Tech and Emory University. Dr. Naveen Muthu of Children’s Physician Group will be Garcia’s counterpart in leading Pillar 1. 

Since 2021, Garcia has served as an assistant professor in Georgia Tech’s H. Milton Stewart School of Industrial and Systems Engineering. His research group has published numerous journal and conference papers, and book chapters related to data-driven machine learning and optimization in healthcare, including various applications in diagnosis and disease management of concussion, opioids, cardiovascular disease, glaucoma, and maternal health. He has received federal funding as a primary investigator from both the National Institutes for Health and the Agency for Healthcare Research and Quality. He and his research group have received several national and international recognitions for their work. 

Garcia also teaches graduate-level courses in machine learning and optimization for healthcare. He received his Ph.D. in industrial and operations engineering at the University of Michigan and was a postdoctoral fellow at the MGH Institute for Technology Assessment.

Over 90% of the U.S. population has internet access. 

However, many households, particularly those of low socioeconomic status, are “smartphone-dependent,” meaning they rely purely on their smartphone for internet access. As a result, their connection may be unstable or slow, and they may be constrained by data caps that limit how much they can use the internet. This puts them at a disadvantage compared to households with internet access through smartphones and other broadband connections at home and work, perpetuating digital inequality between disadvantaged and advantaged households. 

The smartphone dependence of many disadvantaged households begs the question: If mobile internet service was better – e.g. if it was faster, more reliable, and/or didn’t come with data constraints – could that reduce digital inequality and level the playing field? Researchers from the Georgia Tech Scheller College of Business and Southern Methodist University Cox School of Business studied this question and found the answer is “yes.”

Karthik Kannan, assistant professor of IT and Operations Management at the Cox School of Business and Georgia Tech Ph.D. graduate, led the project. “I was interested in the effect of data caps. For example, when you have 10GB of data per month and use more, you are charged extra, or your connection is throttled,” said Kannan. “So, I partnered with a large telecommunications provider to study what happens when their subscribers switched from capped to unlimited data plans. I was particularly interested in differences between high-income and low-income households.”

Kannan, along with Eric Overby, Catherine and Edwin Wahlen Professor of Information Technology Management, and Sri Narasimhan, Gregory J. Owens Professor of Information Technology Management, at the Scheller College of Business, found that while all households increased their data use after switching to an unlimited plan, the increase was significantly larger for families of low socioeconomic status.

“That was our initial finding: that improving mobile internet service by removing the data cap had disproportionately large benefits for disadvantaged households,” said Overby. “But that didn’t mean much in and of itself. If those households weren’t using the additional data for ‘enriching’ purposes like accessing educational, health care, or career-related data, the additional data consumption wouldn’t translate into positive social benefits. Indeed, years of research on digital inequality have consistently shown a ‘usage gap’ in which advantaged households take fuller advantage of internet access improvements than disadvantaged households. The result is that internet improvements often exacerbate inequality. So, we dug deeper.”

Specifically, the researchers leveraged the telecommunication provider’s data categorization system to study changes in the consumption of educational data. They found that disadvantaged households experienced disproportionate increases in education data consumption (as well as in overall data consumption) after switching to unlimited mobile data. Although advantaged households increased their education data consumption by approximately 15MB (or about three digital textbooks) per month after switching to unlimited data, disadvantaged households increased their education data consumption by approximately 24MB (or about five digital textbooks) per month.

 “We can’t be sure that these disproportionate increases in education data consumption will help disadvantaged households narrow gaps in educational outcomes. However, this is clearly a step in the right direction,” said Kannan. 

 The research is directly relevant to the Federal Communications Commission’s 2023 inquiry into the effects of data caps on disadvantaged households. Narasimhan explains, “Let’s say that based on their inquiry, the FCC decides to limit the use of data caps. A logical question is: will that do any good? In other words, will disadvantaged households take advantage of their improved mobile internet service in a way that can reduce digital inequality? Prior to our research, we didn’t really know. But based on our research, the answer is yes.”

 The research paper is forthcoming in Management Science and available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4173558.