Wild Ambitions for AI

MEREDITH GORE was eight-and-a-half months pregnant when a federal agent called with an unexpected request: Could she take in 900 other babies—the kind with shells? Customs agents at the Canada-Michigan border had just caught a man with “irregularly shaped bulges” in his sweatpants trying to smuggle what would amount to a small fortune in young terrapins to sell as illegal pets. Most were stuffed in cereal boxes in his luggage, but some were taped to his legs and groin.

“Just no. I’m a social scientist. I’m a dog person. I wouldn’t know what to do with 900 turtles,” Gore told the U.S. Fish and Wildlife Service agent.

But she had cared for confiscated reptiles at the authorities’ request before, so the then-Michigan State professor and expert on the illegal wildlife trade finally agreed to foster a solitary African spur tortoise that would serve as evidence to finally convict the smuggler.

The incident represented just one thread pulled out of the vast fabric of the global illegal wildlife trade, with an estimated annual U.S. value exceeding $1 billion; the worldwide total is anyone’s guess, Gore says. For every obvious criminal like the man with turtles in his pants, perhaps 100 more-creative ones slip through—hummingbirds nestled inside wigs, scorpions labeled as chocolates, seahorses buried in boxes of chili peppers, even iguanas stuffed into a prosthetic leg.

“We do a terrible job reducing the scope and scale of the illegal trade with seizures,” says Gore, today a human-focused geographer at the University of Maryland. One hurdle is the ceaseless demand for animals in traditional medicines and dishes in various cultures and their diaspora communities. Others are inconsistent data collection and reporting, and the herculean nature of organizing international law enforcement operations.

The environmental and social costs of this illicit trafficking are extreme: habitat destruction, loss of species, damage to human communities reliant on healthy ecosystems, even the potential spread of diseases and pandemics. “This is a massive driver of accelerating biodiversity loss on our planet—one that’s basically hidden,” Gore says.

But the environmental law enforcement challenges that seemed nearly insurmountable before look less so today, as artificial intelligence (AI) continues to grow in sophistication. The advanced computing methods can sift through vast piles of information, quickly find subtle connections that human researchers might be slow to spot—or never see—and suggest new approaches to solving some of the world’s greatest challenges.

Gore is collaborating with computer science experts at the University of Southern California (USC) to use machine learning and AI to leverage the “big data” available from airport crime and travel records. While the information can be gap-toothed, the team sees potential and has begun to demonstrate how the system could identify hotspots or smuggling routes, giving traffickers a run for their illegally gotten money in a kind of futuristic “CSI: Wildlife.”

GORE GREW UP chasing snakes and finding injured birds to nurse in the Central Massachusetts woods with her dog, Abby, at her heels. She became a scientist to help solve problems for wildlife, but today it’s the motivations of people that move her the most.

As a student intern tasked with filing at the nonprofit Defenders of Wildlife, she discovered studies on human behaviors relating to wildlife conservation and wondered, “Why aren’t decision makers using the social science data in these studies? I wanted to change that,” she recalls. Later, while studying human-black bear interactions for her doctorate, she became frustrated by people’s continual feeding of the animals despite clear evidence—and laws—against it. She began seeing how criminology could be the teeth of her wildlife work.

“In a way, I’m an ‘anti-disciplinary’ scientist, meaning I’m not one thing or the other,” she says. “I work in the spaces in between.”

Helping to develop and then run the Conservation Criminology program at Michigan State solidified Gore’s devotion to that in-between. Then, after more than a decade building a scientific understanding of wildlife crime, she brought her layered expertise to UMD’s geographical sciences department, where she continues to lead multidisciplinary projects studying the social drivers of global environmental change and seeking solutions.

Gore is no deskbound scientist: She’ll sit on a wooden crate under a tree in Cameroon with villagers as they draw maps showing where scaly anteaters called pangolins are poached, or squat on a dock in Mexico talking to fishermen about incentives to protect sea cucumbers from illegal fishing. She’s interviewed rhino guards about their perceptions of crime prevention from the back of a pickup truck in Zambia and led focus groups about local poaching methods with park managers in fire towers overlooking vast forests in Vietnam.

Her students are similarly hands-on, whether researching wildlife poaching in Namibia, tiger trafficking in Indonesia, illegal logging in Madagascar or overfishing in Brazil’s Pantanal wetland.

Time and again, she’s seen the “why” of illegal wildlife trade. It often starts with rural people trying to provide for their families through activities like hunting and trapping, similar to those their forebears engaged in legally, she says. “What a terrible calculus to have to choose to break the law to get food or medication for your child.”

On the demand side, buyers are looking to cure illness or boost sexual prowess with folk medicine, or to impress peers with luxury foods or exotic pets, never comprehending the ecological harm they contribute to.

With so many stakeholders and facets to the problem, a single knockout blow to broadly solve wildlife smuggling is an unrealistic goal. “The aim is disruption at as many points as possible,” Gore says. “So we are tackling the problem with evidence from different angles and all along the chain.”

Gore’s ability to embrace complexity—the way wildlife crime often nests with drug smuggling or human trafficking, for example, or the need to unite different academic disciplines to comprehend the challenge—is one of her superpowers, colleagues say.

“She knows how to collaborate, how to talk to all sorts of stakeholders, and how to tackle things across different dimensions,” says research collaborator Bistra Dilkina, a USC professor of computer science.

REDUCING THE ILLEGAL TRADE requires focus on another “in between,” Gore realized: It’s not just about the buyers or sellers themselves, but how they connect. After all, if you can’t get the dried seahorses from the boat dock to the curio shop two continents away, the trade withers and dies.

Finding those paths and links essentially meant making the invisible visible, and that’s where the AI-based methods used by Dilkina and her Ph.D. student, Hannah Murray, became central. They teamed up with Gore to identify the global flight networks smugglers prefer, using AI to build models and find patterns that might predict trading hubs not yet on officials’ radar.

“We wanted to know, why do traffickers select the airports they do—what are the drivers? And where are the undetected hotspots of trafficking activity?” says Murray. “Which routes are they most likely to use?”

The lack of standardized reporting means airport trafficking data is inconsistent and sorely incomplete, creating an obvious bias toward detected incidents. But AI and other computational strategies can suggest answers to these questions by parsing out the hidden interactions in data.

For a study published last year in Nature Communications Earth & Environment, the team used AI to create a model that considered reams of historical trafficking data from some 2,000 airports worldwide: arrest locations, what was snatched, and smugglers’ origins and destinations. It also assessed 1,300 airport features influencing their airport choice, from size and proximity to natural resources to whether it’s located in a country that has joined the Convention in International Trade in Endangered Species of Wild Fauna and Flora (CITES).

“Why would a smuggler choose Rome instead of Milan, Denver over Detroit?” asks Gore. “By looking at the feature analysis, we can understand better what influences the social network.”

The model used machine learning to obtain an unprecedented kind of tipoff: It offered the best predictors of an airport’s likely involvement in wildlife crimes, a list topped by centrality in the global airline network, history of unlawful plant trafficking (which often goes hand in hand with animal smuggling) and level of anti-trafficking measures at each location.

Then the model highlighted 307 airports as potential “hidden” trade hubs (where illegal wildlife commerce hadn’t yet been detected), circling 11 of those as likely hotspots. Those included Dallas Fort Worth International and Denver International, not previously flagged in global trafficking databases based on seizures. The list also points out airports in Indonesia, the Philippines, China, Mexico and Italy, suggesting they’re worth watching more closely.

“The technology isn’t just predictive; it can also help in decision making and resource allocation,” Dilkina says. The model’s revelations of the negatives—places smuggling would be expected but hasn’t been recorded—suggest where to boost anti-interventions such as putting more law enforcement personnel at airports, providing more training for workers and posting signs in bathrooms asking travelers to report suspicious behavior.

FOR ANIMALS CAUGHT up in the trafficking world, AI is proving the master of “speed data-ing.”

Case in point: pangolins, the world’s most trafficked mammal. The nocturnal cat-size animals, native to parts of Asia and Africa, have long been poached for their meat and scales, the latter of which is used mostly in traditional Asian medicine. As a result, CITES has listed all eight pangolin species as being threatened with extinction and banned commercial trade in the animals and their parts.

Gore and Dilkina joined conservation ecologist Matt Shirley of Florida International University (FIU) for Operation Pangolin (OP), a multidisciplinary effort using AI to analyze information combined from wildlife crime, wildlife population monitoring and socio-ecological systems in areas pangolins live.

One aim, says Shirley, OP’s lead researcher, is to predict where crimes will occur—like where animals are likely to be captured, slaughtered or sold to a smuggler—and shape effective responses and policy recommendations to prevent them from happening in the first place. (In Gore’s past projects, prevention has ranged from involving local law enforcement to working with churches or community organizers. Arresting people is not always the aim, she says.)

Pairing AI with other technologies using a range of sensors and tracking systems is central to OP’s work. “For example, imagine the time needed to look through 100,000 images or videos trying to pick out the pangolins, then having to answer, black-bellied or white-bellied?” Gore says. “AI can do this kind of work for us at speed and scale.”

Now the OP team is setting up machine learning systems to automatically process data as soon as it arrives, Shirley says. “So, now, if a field biologist finds a trapped animal
we can get a text with an instant ID, and we know right away if a sound he or she reports hearing was a tree fall or a gunshot, which has real-time implications for how we respond.”

Land animals aren’t the only ones swept up for illegal sale. Sharks are hot commodities globally as food; currently CITES lists 149 threatened species. Gore’s shark work has focused in part on understanding public perception of the animals; their sinister image can be an obstacle to protecting them from illegal fishing and trade. Her FIU collaborator, Diego Cardeñosa, uses molecular and forensic tools to fight shark fin trafficking. He developed an inexpensive DNA test that takes just two hours to identify the species in a sample taken from a suspicious seafood shipment, instead of the months that sorting out DNA can normally take.

“Often if you don’t provide evidence that a shipment of shark fins is illegal within the first 24 hours, you have to let the container go, and illegal behavior goes unpunished,” he says.

The team’s goal is to generate an increasing flow of data on where containers are coming from, where they’re going, which species are being traded and who is involved, and use AI to categorize the seemingly chaotic permutations arising from all vessels, routes and species entangled in the global illicit trade. It’s a powerful way to funnel information into CITES and other authorities to trigger investigations, enforcement and punishment.

A SUCCESSFUL DAY in the field for Gore looks something like one that took place in August 2024 in a vast transnational “peace park” spanning South Africa, Mozambique and Zimbabwe.

She and her team were seeking links between elephant poaching, poisoning of vultures to prevent them from leading law enforcement to crime scenes, trade in vulture parts for traditional use, and disease transmission risks from elephants to vultures to people. Operating from an empty classroom in a school in a remote Mozambican village, she met with a succession of local people, including elders, teenagers, and working men and women, all willing to help with “participatory mapping” of, literally, where the wild things are, and where they’re being killed.

Once the team explained the disease linkages—and residents got past their surprise that an institution like UMD would send people thousands of miles just to learn what they had to say—the interviews provided a torrent of valuable information. “The local people have so much knowledge and understand the trade perfectly,” she says. “They are the best teachers.”

“Successful” doesn’t always mean easy: Gore and her team faced the logistics of tagging vultures with ethical and safe “backpacks,” moving a team of 15 from camp several hours away to the village via front-wheel-drive trucks on dirt roads and buying and preparing chickens or a goat to feed everyone. The demands of feeding a voracious AI algorithm are even greater, however, and the interviews required voluminous and precise data gathering that left the team “cognitively exhausted,” she says.

Maybe they wouldn’t need AI if the depth of cooperation they found in this village existed everywhere, or they’d had permission to work on the project in Zimbabwe—or if her U.S. Defense Threat Reduction Agency grant hadn’t fallen victim to and sweeping cuts to federally funded science, keeping her and her team out of South Africa.

The result is “data deserts” that exist in parts of the park, or even whole countries, inaccessible for political, logistical or funding reasons. “The AI helps us stitch all these things together, giving us the ability to extrapolate data in a rigorous, replicable way that previously was not possible,” she says.

Dilkina notes that while continuing to improve the airport model, “we want to apply this method to other transportation modes, as a lot of illegal wildlife trade is happening by maritime transport.”

Across the projects, Gore says, AI and machine learning aren’t just a kind of super calculator; they give her and her collaborators a chance to leapfrog barriers to knowledge with a tool that combines human intuition and cutting-edge computing. With AI, they’re slicing through the mystery around a trade its practitioners are working furiously to hide.

“With so many people in different disciplines working to reduce illegal trade, if there were an easy solution—or even a moderately hard solution—we would have found it already,” Gore says. “These computational approaches are the force multiplier for data integration that conservation has been waiting for.” TERP