I actually looked into the technologies for this in quite a bit of detail during my time at CEA. Creating a map for the meetups is quite doable, either with Google Maps or with Mapbox, and doesn’t require that much engineering effort.
The biggest obstacle I see here is both that Facebook is the tool of choice for creating events, and that it is really good at this in a way that we won’t be able to beat. Some kind of Facebook integration might save us here, or maybe fully focusing on reoccuring events is better.
An alternative would be to focus on “communities” instead of meetups. I.e. you can register a community on LessWrong, together with some metadata, such as Facebook groups, community greeter contacts, meetup locations, etc. Those communities are then mapped, and allow nearby rationalists to find those communities and engage with them. This generally reduces the amount of maintenance effort required by the local communities, and makes the system more flexible to integrate with whatever the local group uses as their event coordination tools.
An alternative would be to focus on “communities” instead of meetups.
This is the basic idea; a consideration to keep in mind is that there’s both passive new users (“hey, I just moved to Austin, are there any meetups nearby?”) and exciting events (“hey, we’re hosting a HPMOR wrap party!”), and you want to handle both cases well.
Now, maybe the thing to do here is have something like findable communities (basically, what’s on LW is a geolocation and a link to Facebook/whatever else you use) and then location-based pings (either based on IP or them letting us have their location), which the community owners can create. But I don’t really want LW to be prompting users to allow us access to their location all the time.
We should get cell-tower accurate positioning, which is usually good enough to determine the city and neighborhood you’re in (i.e. we should be able to distinguish between Oakland and Berkeley, but probably not central Berkeley and North Berkeley).
Rough googling suggests about 95% accuracy on city-level detection.
I actually looked into the technologies for this in quite a bit of detail during my time at CEA. Creating a map for the meetups is quite doable, either with Google Maps or with Mapbox, and doesn’t require that much engineering effort.
The biggest obstacle I see here is both that Facebook is the tool of choice for creating events, and that it is really good at this in a way that we won’t be able to beat. Some kind of Facebook integration might save us here, or maybe fully focusing on reoccuring events is better.
An alternative would be to focus on “communities” instead of meetups. I.e. you can register a community on LessWrong, together with some metadata, such as Facebook groups, community greeter contacts, meetup locations, etc. Those communities are then mapped, and allow nearby rationalists to find those communities and engage with them. This generally reduces the amount of maintenance effort required by the local communities, and makes the system more flexible to integrate with whatever the local group uses as their event coordination tools.
This is the basic idea; a consideration to keep in mind is that there’s both passive new users (“hey, I just moved to Austin, are there any meetups nearby?”) and exciting events (“hey, we’re hosting a HPMOR wrap party!”), and you want to handle both cases well.
Now, maybe the thing to do here is have something like findable communities (basically, what’s on LW is a geolocation and a link to Facebook/whatever else you use) and then location-based pings (either based on IP or them letting us have their location), which the community owners can create. But I don’t really want LW to be prompting users to allow us access to their location all the time.
Well, we have location by IP, which is rough but accurate enough for meetups. Only more detailed location requires permission.
Does that work on mobile?
We should get cell-tower accurate positioning, which is usually good enough to determine the city and neighborhood you’re in (i.e. we should be able to distinguish between Oakland and Berkeley, but probably not central Berkeley and North Berkeley).
Rough googling suggests about 95% accuracy on city-level detection.