Bird in hand

What can birding teach us about machine learning? And how is AI shaping how we interact with nature? Projects at the intersection of nature observation, citizen science, and machine learning offer useful case studies for examining systems of dataset production, model training and human feedback. They also present an alternative model to the extractive and exploitative “Big Data” approach to training machine learning algorithms, offering many possibilities as well as unique challenges for thinking through how we relate to AI systems.

Platforms such as eBird, Merlin, and iNaturalist draw on long histories of public participation in observing and recording nature as data. Here, we explore how machine learning has come to shape these massive citizen science undertakings—and vice versa—by examining how citizen scientists, professional naturalists, and computer scientists come together with birds and other lifeforms as well as technologies such as computer vision. Our research and writing seeks to understand 1) the unique relationship between labor, expertise, and classification in these “hybrid” systems; 2) how birds and other living organisms become known algorithmically—or resist it; and 3) what implications these projects may have for how we govern and live with AI.


Birds Through A Knowing Glass

Jer Thorp
One midweek morning near the end of September, just after 2am, an angel appeared on the South Shore of Long Island Sound. It grew in size until it was about two miles wide and then it moved down, past Hauppauge and Melville, Garden City and Great Neck, before drifting over the eastern boroughs of New York City. There was a glimmer, just as the first blue light of dawn was appearing, and then it was over. The angel fell earthward and dissolved into hundreds of thousands of winged things. They settled down to the city’s branches and onto its dew-covered lawns.

If Not Them, Us

Hamsini Sridharan & Jer Thorp
A year into the so-called AI revolution, Silicon Valley’s scions are asking us to believe two things at once. First, that this time will be different. More massive than big data, more disruptive than web 2.0, more transformative than social media. That the territory is new and strange, the dragons bigger and more sharply-toothed, the old maps useless. Second, that they are somehow still the right ones to lead the way.

Odd Ducks: The Hybrid Logics of Machine Learning in Fine-Grained Wildlife Classification

Hamsini Sridharan and Jer Thorp
In this article, we study the Visipedia computer vision project, which focuses on collaboration with naturalist communities, as an alternative approach to the extractive and exploitative “Big Data” paradigm of ML datasets. We focus on the processes of classification, labor, and expertise that shape Visipedia’s datasets and tools, including eBird, Merlin, iNaturalist, and NABirds. We argue that logics of hybridity shape Visipedia’s work, showing its possibilities and limitations. We conclude that hybridity is a useful lens for critical studies of ML datasets more broadly.