About This Project
All species on earth abide by the same rules of behavior. We eat, sleep, build homes, and build community, all in the service of evolutionary fitness. For over a century, the field of ethology has endeavored to understand and quantify the behaviors we produce through careful observation, yielding great insights into the study of behavior. In this project, we develop a platform utilizing recent advances in machine learning to delve deeper into the organization of animal behavior.
Ask the Scientists
Join The DiscussionWhat is the context of this research?
The recent surge in machine learning techniques has sparked a revolution in studying animal behaviors, leading to the burgeoning field of "computational ethology" in which computational tools are used to interpret and understand animal behaviors in ways that are often not accessible to human experts. For example, discovering novel structure in the way a fish hunts or how a bird’s song is passed down through the generations.
Although most behavioral research in the lab remains focused on the study of a small handful of model organisms, such as lab mice, fruit flies, and zebrafish, these data-driven techniques show promise in elucidating the diverse forms that behavior takes.
What is the significance of this project?
Animals produce behavior to serve biological needs. Each animal’s behavior serves its own ecological niche. Yet, many of our behaviors are conserved from species to species. By creating a general framework for computational ethology that can be applied to a broader range of non-model species, we can explore how variations across taxa in ecological niches, body morphology, and behavior relate. Classic ethograms describe behavior in human-understandable terms, for example, when an animal is eating, building shelter, or affiliating with others. These broad-strokes approaches miss key details in behavior, often imperceptible to humans, that capture how subtle variations in behavior enable vast differences in the habitats, social structures, and climates an animal can inhabit.
What are the goals of the project?
Our primary goal is to establish an open-source, modular platform capable of recording intricate behaviors across a wide range of animal species. The aim is to create a versatile tool that can adapt to different experimental setups and data types, thereby enabling researchers to study a broader spectrum of behaviors without the constraint of proprietary systems.
With that goal, we then intend to leverage the established platform to gather behavioral data in various formats to understand how different species structure their behaviors in the service of different contexts. For example, how do social hierarchies dictate how we act around conspecifics, and how do non-verbal communication signals interplay with vocal communication?
Budget
High speed and high resolution machine vision cameras are critical to monitoring the complex skeletal dynamics of an animal in action. By recording simultaneously from an array of synchronized cameras, we can use machine vision tools to triangulate each animals skeleton as it moves about the world.
A reproducible 8020 arena is needed to produce a well-controlled environment for recording from animals. By controlling the lighting and housing of each animal, we can focus solely on the structure of behavior.
These budget items cover the main expenses of this project which also requires a high-power computer and custom circuitboards and lighting.
Endorsed by
Project Timeline
Our goal in this project is to develop a behavioral recording arena, record a large scale dataset, analyze that data and publish our results.
Large datasets like this, and the development of new technologies, takes time. We already have one prototype for this recording arena, the funding you provide will enable further development and the high-throughput needed to see this project through.
Dec 01, 2023
Build a behavioral arena and set up the videography rig.
Dec 07, 2023
Project Launched
Apr 01, 2024
Conclude data collection for a diversity of rodent species.
Jul 01, 2024
Prepare and publish a detailed manuscript outlining our findings and detailing the developed apparatus.
Meet the Team
Tim Sainburg
Tim Sainburg is a Postdoctoral Scholar and Schmidt Science Fellow at Harvard University studying the syntax of natural behavior. He has a background in animal behavior, having worked with chimps, monkeys, songbirds, and now rodents.
Lab Notes
Nothing posted yet.
Project Backers
- 2Backers
- 100%Funded
- $5,005Total Donations
- $2,502.50Average Donation