[mindcrime]

I don't want my stream filled with junk the way my Facebook stream is.

Me either, but - IMO - the answer to this isn't to leave out the "write API," but rather the answer is to give users better filters and the ability to define their stream to meet their own desires. Of course doing so with a simple and intuitive UI is the tricky part, but hey... those G+ engineers make a lot of money, they should be able to come up with something, no?

[ed209]

> they should be able to come up with something, no?

no! :)

Probably the hardest thing in social media is noise filtering via a UI (it's hard for Google, it's hard for Facebook, it's just a hard problem).

For my own (app) selfish reasons, I'd like to see a write api for comments first as that doesn't cause issues on the same scale as writing posts to the stream. Could be a nice place to start.

[Jach]

Noise filtering is just another form of spam filtering, Google has a lot more in-house experience with that than Facebook I imagine. I think even something as simple as a personalized Naive Bayes filter would go far.

Ad hoc groupings of people in your Lists and ad hoc set operations on your current groups (school intersection friends, union of all groups filtered where profession = X or Y, etc.) is something a write api would let a budding entrepreneur develop an app around and potentially make a lot of money.

[ed209]

> Noise filtering is just another form of spam filtering

Logically you might think so, but noise is not always noise, sometimes it's what you're interested in reading about, some days it isn't. Spam, however, is always spam.

Google+ have tried to solve this "noise" problem through the ui with a "volume" slider. But it's like fine tuning a recording desk if you have to do that to 50 circles to get the mix just right.

I'm just not sure it's solvable through the UI and that we may have to be quite harsh on what "noise" we let through and rely on "if it's important enough, it will find me". There's just too much data to consume otherwise.

[mindcrime]

Google+ have tried to solve this "noise" problem through the ui with a "volume" slider. But it's like fine tuning a recording desk if you have to do that to 50 circles to get the mix just right.

The problem is, that's not even the right approach. It's built on a faulty assumption - that "who" is the sole determinant of whether or not I'm interested in something. That's not even close to true, unfortunately. I have plenty of social-network connections for whom I value their posts on certain topics and have less-than-zero interest in what they have to say about other things. For example:

"I follow Bob, and I want to see Bob's posts on technical topics, but NOT Bob's posts on religion."

Or more generally:

"I follow all these people, but regardless of who posts what, I never want to see pictures of cats."

I'm just not sure it's solvable through the UI and that we may have to be quite harsh on what "noise" we let through and rely on "if it's important enough, it will find me". There's just too much data to consume otherwise.

I don't know, I think a solution that was 80% effective would be incredibly valuable. I don't see this as something that's "all or nothing."

And my personal take is that, in the case of G+, the Circles thing is a good start, but they need a way to specify "excludes." Whether those excludes should be explicitly defined by the user, or whether they can use machine learning to figure out what I don't want to see is an open question, granted. But I'd take some UI for specifying "exclude this topic" (while acknowledging there is some inherent fuzziness in this) in the meantime.

[kooshball]

> Logically you might think so, but noise is not always noise, sometimes it's what you're interested in reading about, some days it isn't. Spam, however, is always spam.

Spam is not always spam. There is a subset of spam that are obvious spams such as ones that tries to sell you viagra. Outside of that, there are mails that are harder to classify since different people will respond differently to them. It might seem like that's just me being pedantic but the result is that user end up with mail they do not want to receive, and mark it as spam. This is similar to how I respond to news feeds except I dont have an easy "mark as spam" button without blocking that person forever.

G+ should implement some sort of rating system to rank items that I liked and push those ones similar to it higher on the feed.

[mokus]

As I understood it from the context, the point of the quoted comment was that, even for the _same person_, whether something is noise or not varies wildly depending on their state of mind but for spam that is much less true.

Different people might have different opinions about what is and is not spam, but one person will be fairly consistent about it. For example, it doesn't matter much what mood I'm in - I'll almost always classify the same things as spam. But the same cannot be said of classifying things as "interesting" or "not interesting".

Your idea about a preferential ranking system would probably be a good way to go. Definitely better than trying to train a strictly go/no-go filter to recognize "interesting". The way Google+ posts seem to jump around every time I open the page (with no obvious correlation with things like number of comments or +1's), I actually wouldn't be surprised if they are already doing something like that.

[rys]

Why should the UI or API get involved with filtering what's inherently the human's problem?