Thursday, February 11, 2010

What's the Buzz

Like everyone else on the planet, I'd like to talk about Buzz for a second.

I'm not one of the people who thinks Buzz is stupid or off-base: Buzz actually fits rather nicely into Google's framework and policies. But I'm also not someone who thinks Buzz is great.

There are things I like about Buzz that I think we'll see a lot of in future products, and I'd like to hit those and talk about further development on them.

First, Buzz is a prototype for an email replacement system. With just a bit more polish, Buzz could quite literally replace email entirely. I think that's a good thing, and I think we'll see a lot more things with that kind of functionality. I don't think Buzz will replace email, but it is a step towards something that will.

Second, Buzz does threaded conversations. So does Gmail proper, of course, but most "social" solutions don't. Twitter, especially, does threading extremely badly.

Buzz doesn't thread very well, to be honest. There are some nice innovations, but it is still making the assumption that (A) threads are inviolate and (B) I care more about popular threads. Both of these assumptions are wrong.

I expect that threading will become more and more important as time goes on, and in the future I expect to see products that thread much more intelligently. This includes not just Slashdot-style recursive threading, but also "out-threading", connecting to other threads or posts in a method that isn't a manual, clumsy link.

For example, Amy and I are talking about a cool YouTube video about man-eating squirrels. The threading system automatically "out-threads" us to other people who are also talking about that YouTube video. Their comments do not appear "in line" - they're not part of the same thread. But they do appear, and can be tracked.

Similarly, if Bob joins in and links us to another video about a sumo wrestling squirrel, his post is automatically "out-threaded" to link up to other people talking about the sumo wrestling squirrel.

The problem with these kinds of advanced threads is that they can't be done blind. There are a lot of people out there that I just don't care about, and I don't want to hear what they have to say. It's not a matter of just ignoring the little out-threads, either: I want good out-threads, not bad out-threads.

To this end there needs to be an algorithm which tries to determine if someone who I'm not directly connected to is someone I would find interesting. There are some algorithms to do this, but they vary in complexity and some of them are pretty absurd. Normally, I would recommend making them run on the client's machine, because it would be a lot of work to run on the servers. On the other hand, Google (or whoever) will want to abuse it for money, so running on the server might be the only profitable solution.

Anyway, "threading", "context", and a kind of "user terrain" are probably going to be more and more important as the years go by.

I want to follow someone's posts, but I don't want to hear their posts about their shoes or how much weight they've lost. These posts tend to also gain the most comments. This is why new algorithms are required: not only does the algorithm have to tell that I might want to listen to someone, but it also needs to be able to tell when I don't want to listen to someone I follow.

Buzz isn't this solution, but it's starting to point in the right direction.

6 comments:

Patrick said...

What's your take on how this splashes the rest of the ecosystem? Particularly in regards to social game design?

Craig Perko said...

It doesn't. It's going to take some time to see if it takes off much, but I don't think it will. Even if it does, it's not built to support social gaming.

That's actually a strong point, when it comes to replacing email and being context sensitive. Being able to support social gaming means that the framework supports poisonously aggressive replication, which pisses me off. I don't want to know that my friends just grew some beans in FarmIdiot. FarmIdiot's ability to replicate is dominant over the system's ability to thread and build context.

Patrick said...

What if the gameplay in question isn't some massively single-player click gulag but some kind of multiplayer gameplay that is literally social?

Craig Perko said...

Then there's no particular reason it should favor Buzz.

The only game mechanic can see Buzz being unusually good at is a game which revolves around people having fake email addresses and trying to connect to each other. Discovering the other people's addresses would be the big driving factor, as it's not as open as Twitter would be. Still, I can't think of a good way to do it off the top of my head.

brackish meadows said...

there's this stuff my matthew skala on how to create social networks that predict whether strangers are interesting or not:
how to run a conspiracy
http://ansuz.sooke.bc.ca/economics/social-networking/run-a-conspiracy.php
how to run a drama
http://ansuz.sooke.bc.ca/economics/social-networking/run-a-drama.php

Craig Perko said...

Interesting, thanks. I'll read up.