Yesterday, I watched this video about the airline industry’s outdated logistics software. Southwest Airlines uses this tool called SkySolver to route their flights and work around delays and cancellations, and it’s old & rickety and caused this major cancellation-chain meltdown over the holiday season last year.
While watching, I found myself thinking, could there be a system that allowed humans to do a task like this efficiently? The obstacle to overcome would obviously be the need to coordinate and process more information than a single human is capable of handling. When rerouting a plane, you have to be aware of all the places it could go, the schedules of the crews involved, the realtime needs of other cancelled routes, and much more. But if given the right tools, I think there could be a massive advantage in leveraging the intelligence and creativity of human thinkers.
This all reminded me of a note to myself I recorded last April: “Are there hivemind-facilitating tools?” What I had in mind was a big canvas, with a realtime collaboration like Google docs, but built for maybe 1000 people at a time, for example. Everyone can view and move around the canvas freely, providing a large shared space for everyone to work, whether together or separately.
People could break down large problems into their components and tackle tasks at various levels of abstraction. For this, you’d maybe need people acting in different roles. Some would handle the low-level data processing tasks (possibly with machine help), while others would work on pulling everything together and looking for insights. Others might simply facilitate communication between those working at different levels of abstraction.
There would probably also be a voice chat, but you wouldn’t want everyone talking over each other. You’d probably need something like proximity chat in video games, where people can only hear one another if they are close to one another. With modern audio transcription being so fast, you could even convert people’s words to writing automatically, making it possible to drop into conversations with the last couple minutes of context easily available.
This whole idea would probably only work for a specific problems, and would probably benefit from augmentations specific to those problems. For example, if it were to be used by an airline for logistics planning, it would probably need to be integrated with realtime flight data and some kind of request system.
It’s possible this whole idea is completely unworkable. Problems like this might still be too complex for humans to reason about well no matter how much help they had. Keeping that many people coordinated at once might be more messy than it’s worth. But I’m curious to see what a group plugged into a well-tuned “hivemind” system could do, once they mastered it.
Do you know of any software that approximates this approach?
Do you think a software-facilitated “human hivemind” would be good for anything?
Do you want to try to build a prototype?
Let me know your thoughts at my Ctrl-C email:
gome @ ctrl-c.club.