Even non-lethal exposure to pesticides affects bees, and Dr. Rachel Parkinson’s BEEhaviorLab is using AI-assisted analysis to understand what that means for declining populations in the wild.

Rachel Parkinson’s lab at the University of Oxford holds 30 boxes, each housing about five bees along with a camera, a microphone and sensors rigged to an inexpensive, compact computer. This system tracks subtle shifts in behavior—how the insects move, eat, interact—when exposed to low levels of pesticides.

“You essentially just press go and it records over time,” says Parkinson, a Schmidt AI in Science Fellow. “It analyzes all of the data that’s coming in, and outputs plots describing it.”

Traditionally, this kind of research meant a scientist manually observing one bee for hours. Parkinson’s system, which she calls the BEEhaviorLab, lets a technician run many of these experiments simultaneously while AI handles observation and analysis. She’s now preparing a paper on the invention and talking with the UK Health and Safety Executive about incorporating it into studies that inform regulations governing chemical use.

Pollinators—bees chief among them—are essential to crops from apples to coffee. But their numbers are declining. Pesticides are a major factor, compounding the challenges that climate change, habitat loss and disease pose for these insects.

Current regulatory testing measures how much of a pesticide it takes to kill an insect, typically honeybees and perhaps other pollinator species. That approach misses subtler damage.

“We’re seeing that in the wild, bees are experiencing profound impacts at concentrations far lower than the ones we expected would be toxic,” Parkinson says. Her BEEhaviorLabs capture changes in the pollinators’ behavior to pinpoint the levels at which harm begins.

Parkinson trained as a neuroethologist, studying the neural circuits that guide insect behavior and how environmental factors like pesticides alter their function. Over time, her focus has “zoomed out”—from neurons to the broader fate of pollinators. That shift has pushed her to create AI-based research tools and build cross disciplinary collaborations.

Her focus expanded gradually, as she found herself sitting in rooms of scientists with expertise in ecology, conservation biology and other fields. A Schmidt Sciences-funded AI and ecology workshop in San Diego proved pivotal. 

“I find myself looking at bigger problems and seeing the value in building teams that draw on the strengths of people with skills really different from mine,” she says. “We can have an impact on big problems,” she says. “And, to me, empowering scientists to use AI tools jumps out as the most obvious way to make that possible.”