Loonnet https://loonnet.humanitarians.ai/

For a common loon, a good year depends on getting a few things right in the right order, and there is little room for a second try. This video introduces LoonNet, a Humanitarians AI project that packages those decisions as short browser games, and it describes the research work that supports them.

The year the games are built around

The video begins with the return of common loons to the Whitefish Chain in Minnesota every spring. It then walks through why the year is so exposed. Loons nest a few feet from the water because a loon can barely walk. They dive for fish they have to see, so the clarity of the water matters. Every fall they need a quarter mile of open lake to get airborne. At each of those moments, the video says, one bad choice and the year is lost.

LoonNet is presented as a way to turn that whole year into something you can play, instead of something you read about.

What each game asks you to decide

There are four short games, each about a minute long, and each one is built around a single decision.

In Nest Site, you choose where the nest goes. The two failure cases are shown directly: too high and the bird cannot reach it, too low and a boat wake swamps it. In Deep Dive, you feed chicks as the water clouds over, and lead tackle is the hazard in the scene. Chick Taxi is for the youngest players, who carry chicks on a parent's back. Runway has you build speed across open water for the one takeoff that matters.

Because every game is so brief, a player can try all four in a few minutes and see how the stages of a loon's year connect.

Drones, datasets and counting tools

Behind the games is field work. The video says Northeastern graduate volunteers fly drones over the lake, turn the footage into an annotated dataset, and build tools for spotting and counting loons. The tools are described as something the researchers can keep and use, so the work does not end once the games are released. It leaves the research team with tools of their own.

Honest about what the team does not know

One of the clearest statements in the video concerns expertise. The team says it is not made up of loon experts, and that this is the point. The workflow it describes is that the games get built, the biologists correct them, and the corrections make the next version truer. Seen this way, each version of a game is a proposal that the people who study loons are invited to fix, and accuracy improves because of that review rather than in spite of it.

Where to play

The games run in the browser with nothing to install. The video closes by giving the LoonNet web address on the humanitarians.ai domain twice, so viewers know where to go.

Key takeaways

• Common loons return each spring to the Whitefish Chain in Minnesota, and their year hinges on a few critical moments.
• Loons nest close to the water, need clear water to find fish, and need about a quarter mile of open lake to take off in the fall.
• LoonNet consists of four games of roughly one minute: Nest Site, Deep Dive, Chick Taxi and Runway.
• Nest Site teaches that a nest can fail by being too high or too low; Deep Dive includes lead tackle as a hazard.
• Northeastern graduate volunteers collect drone footage, annotate it, and build loon spotting and counting tools for the researchers.
• The team says openly that it is not expert in loons, and uses biologists' corrections to improve each version.
• No installation is needed to play.

Try it yourself

Play the four games in order, then pick the one whose rules felt least realistic to you and write down why. That note is the same kind of correction the project describes: a build, a review, and a truer next version. If you work in conservation or study loons, the video's workflow is also a model for how a volunteer team can ask experts to check its work.