THE DATAFIED ANIMAL: USING TECHNOLOGY TO UNDERSTAND OUR NATURAL WORLD
From left: Emily Wanderer, Justin Kitzes
Humans are part of nature and have always affected the natural world. That reality is behind the concept of the anthropocene, the proposed name of a geological epoch in which human activity—more than any other force on Earth—is the main driver of change from the micro to the planetary level. The term was rejected as an official designation by the International Union of Geological Sciences but has entered common usage, especially among environmental scientists.
If humans inevitably affect the natural world, the question becomes—in what ways should humans affect the natural world?
Emily Wanderer is an associate professor in the Department of Anthropology, Kenneth P. Dietrich School of Arts and Sciences, and author of an influential book “The Life of a Pest,” a study of the politics of nature in Mexico. She recently received an award from the National Science Foundation for her study, “The Datafied Animal: Biologging, Machine Learning and Wildlife Conservation.”
There have been many studies on the effects of datafication on human lives. However, what happens when these technologies are used to study animals? Wanderer calls this “the datafied animal.” The study of datafied animals is at the intersection of animal life and technology but also takes into account the humans who develop and use this technology. “The questions become what kind of data we are gathering,” says Wanderer. “What are some of the things that get left out?”
“It's interesting to collaborate with scientists using these new technologies and to understand what things don't make it into the data collection. As an anthropologist, I argue that data is made by people,” says Wanderer. “It is important to think about the cultural, social and technological aspects of producing data. As a cultural anthropologist, I am also interested in the symbolic and cultural meanings of animals.”
Wanderer describes a project accompanying scientists in Sicily in which dogs and goats wore collars to study if the animals could help predict when the volcano Mt. Etna would erupt. The researchers could track the animals on their phones, but the tracking apps do not record the interactions that humans traditionally have with dogs and goats.
“Only certain kinds of information are useful to science,” Wanderer explains. “The embodied and affective connections people have with animals are not translated into the datafied animal.”
Associate professor of biological sciences in the Dietrich School Justin Kitzes works to expand the possibilities and value of data about wildlife, specifically using acoustic recorders and AI to identify, in ever greater detail, lives of birds, frogs and other animals by identifying the sounds they make. In the course of seven years of recording and refining the algorithm that analyzes the recordings, Kitzes and his team have recorded more than one million hours of audio.
“Our audio analysis software, OpenSoundscape, is now published and out in use,” Kitzes says. And we have got big collaborations and long-term relationships with funders and partners.”
Part of the funding for the research came from the National Science Foundation and other foundations, including the National Fish and Wildlife Foundation. Kitzes’ team is looking at the success of large-scale forest restoration across Pennsylvania by searching their recordings for the presence of three bird species—goldwing warbler, cerulean warbler and wood thrush—using them as indicators of healthy forest habitat.
Kitzes describes the acoustic recording as akin to museum specimens.
“They are permanent records of a particular time and place, and we can go back to that record years later. Let's say, hypothetically, there was an insect apocalypse-type event or a major change in the community—we can go back and ask what did this place look like five years ago, 10 years ago, 20 years ago? When did we stop hearing these species?”
In the past, he points out, biologists jarring specimens in ethanol and collecting skins filling up natural history museums did not know that DNA existed. The research value of the specimens could not have been imagined when naturalists originally collected them.
“I like to think that we're doing something similar,” says Kitzes. “We're building this corpus of data showing how the world looks right now, in part so that future scientists can come up with ideas and answer questions that we're not even thinking about.”
Kitzes says there are ongoing challenges to creating a full picture of the life and health of a habitat. A bird song is usually a male advertising for a mate or defending a territory. The song says nothing about the presence of females or successful nesting. He hopes to learn more from other sounds, such as the sound of a juvenile bird begging for food.
A perennial challenge is finding rare species and sounds.
“If there aren't many of the species, there aren't many sounds from them,” Kitzes explains. “We are sifting through an enormous haystack, looking for tiny needles like the juvenile begging sound. That juvenile may have only passed by the recorder for one minute in an entire season. It's with those rare events where the technology really shines.”
As an anthropologist, Wanderer is enthusiastic about the possibilities of the datafied animal that the audio data and AI methods produce.
“I get to do the cultural and social analysis of the work that Justin does in his lab to think about what it means for people and their relationship to wildlife,” she says. “It is exciting—and complementary work.”