I'll be doing a presentation on topics relating to future location, which is of course our focus with whereyougonnabe. As we've got further into this, I've really come to realize that there are some hard problems to solve in the space of future location which have a lot of value, and many ways that the core technology we're developing for our initial social networking application can be applied. I thought I'd just post a brief preview of some of the topics I'll be talking about.
Broadly speaking, I think that the hard problems fall into three main areas:
Modeling and performance. How you model future space-time has a huge impact on what you can do in your application - for example Dopplr is a nice application in many ways, but its underlying model of space-time has some serious limitations which really constrain the functionality they can provide compared to the approach we are using. I'll talk about coarse-grained versus fine-grained modeling of space-time, and using discrete "places" versus a continuous coordinate system. Implementing a fine-grained system with the ability to handle variable search radii in both space and time is pretty challenging from a performance and scalability perspective with large data volumes, and I'll discuss some aspects of how we've approached this. I'll also talk about various interesting modeling challenges including handling overlapping activities, uncertainty (possible / probable activities), alternative plans (may be either here or there), flexible plans (I'm intending to go to the shopping mall sometime tomorrow afternoon), linear activities (like road or train trips).
Determining what's "interesting". Since we are focused on fine-grained future activities, which may be close to home or far away, users may be "close" to their friends frequently, and what constitutes "close" may vary significantly depending on the context. If I am in England for two weeks I am interested in notifications happening between me and friends who are within a relatively large distance of me (a hundred miles or so, say), whereas if I'm going out for a drink in downtown Denver, where I live, my notification radius should be relatively small. If I have friends who live close to the bar I'm going to, should I notify them about my plans or not - they may well be interested but we don't want to overwhelm people with too much information. I'll talk about factors involved in determining what's interesting, including the distance from home of each person, normal distance between the people, how often you see them, length and nature of the activity, and more.
Gathering data on future activities. Of course one of the big challenges with future location compared to current location is creating the raw data about where people will be and when. Obviously making it easy for people to create activities is one aspect of this, but integration with other data sources is also key, including calendars, online booking systems for travel, restaurant reservations, concert or sporting tickets, etc etc. As we pull data from multiple sources, merging those together in the optimal way can be challenging. I'll discuss a variety of ideas and plans we have in this area.
I'm talking from 4:30 to 5pm on the Wednesday, July 23, so if you'll be at the conference please stop by!