Loonnet https://loonnet.humanitarians.ai/

A common loon's year depends on a short list of decisions, and one bad choice at any of them can lose the whole season. This video introduces LoonNet, a Humanitarians AI project that turns that year into something people can play, and it explains the field work that sits behind the games.

Why a loon's year is so fragile

The video opens in Minnesota, where common loons return every spring to the Whitefish Chain. It picks out three moments in the year that matter most. First, loons nest a few feet from the water, because a loon can barely walk. Second, they dive for the fish they eat, and they have to see those fish, so clear water matters. Third, every fall a loon needs about a quarter mile of open lake to get airborne.

The point the video makes is that these are not forgiving moments. One bad choice at any of them, and the year is lost. That is the idea the games are built to let people feel for themselves.

Four games, about a minute each

LoonNet is made of four short browser games, each lasting around a minute. The video describes each one in a sentence.

• Nest Site. You pick where the nest goes. Put it too high and the bird cannot reach it. Put it too low and a boat wake swamps it.
• Deep Dive. You feed chicks while the water clouds over, with lead tackle as the hazard.
• Chick Taxi. Aimed at the youngest players, who carry chicks on a parent's back.
• Runway. You build speed across open water for the one takeoff that matters.

Each game maps to one of the pressure points described earlier: the nest on the shore, the dive for fish in clear water, the care of chicks, and the long run across the lake before a fall departure.

The field work behind the games

The games are only the visible part of the project. The video explains that Northeastern graduate volunteers fly drones over the lake and turn the footage into an annotated dataset. From that dataset they build tools for spotting and counting loons, and those tools are built so that the researchers can keep and use them. The games and the field work therefore serve one project: the play side teaches the stakes, while the data side produces something a research team can actually use.

Built by non-experts, corrected by biologists

The video is candid about who is making this. Its line is that the team are not loon experts, and that this is the point. The process it describes is simple: the games get built, the biologists correct them, and the corrections make the next version truer. In other words, the games are treated as drafts that experts are invited to fix, not as finished authority on loon biology.

How to play

The games run in a web browser, and the video stresses that there is nothing to install. They live at the LoonNet address on the humanitarians.ai domain, which the video repeats at the end.

Key takeaways

• Common loons face three make-or-break moments each year: nesting near the water, diving in clear water, and getting airborne in the fall.
• LoonNet turns those moments into four browser games of about a minute each.
• The four games are Nest Site, Deep Dive, Chick Taxi and Runway.
• Nest Site shows how a nest placed too high or too low fails; Deep Dive adds lead tackle as a hazard.
• Chick Taxi is designed for the youngest players.
• Northeastern graduate volunteers fly drones over the lake and build an annotated dataset and loon spotting and counting tools for the researchers to keep.
• The team states plainly that it is not expert in loons, and relies on biologists to correct each version.

Try it yourself

Open LoonNet in your browser and play all four games in a row, which takes only a few minutes. After each one, ask what the single decision in the game says about the real bird's year, and notice anything that looks wrong to you. The project is built on the idea that corrections make the next version truer, so a careful observation is exactly the kind of feedback it is designed to use.