Hacker News new | comments | show | ask | jobs | submit login
How I would re-design Facebook, Twitter, Brizzly, etc. (philosophistry.com)
25 points by petercooper on June 20, 2010 | hide | past | web | favorite | 15 comments

That's less a redesign than a cry for help against the input overload social networks provide. I'm glad my social surrounding is relative quiet and for Twitter I only follow a handful of people that don't tell me what they just ate.

Or even better, what if a site could combine all my social feeds, and make sure that I only see the most recent update from a person across all their networks.

I haven't used it but it looks like the author would like http://www.friendfeed.com. Maybe not the best implementation, but a start.

The best solution to this problem is to minimise the number of social networks you use and the number of people you follow on them.

Of course, most people don't realise this until they're already suffering from the problem. When I had a Facebook account, I regularly "pruned" people from my friends list. Now I have a twitter account I try to minimise the number of people I follow, and the number of uninteresting posts I make.

For the average internet user, the number of social networks they use and the number of people they follow is monotonically increasing. It's much easier to sign-up for something new than it is to delete your account. And it's much easier to add friends than it is to remove them. It's like asking people to get rid of every item in their house that they don't really need. It just requires too much work to sit and analyze someone and think, "Hmm, do I want to never hear from this person again on this network?"

I'm not sure that the best solution, it's just a solution.

Even interesting people have spammy days where they most too much (like at a conference) or stuff that's uninteresting to you. Permanently unfollowing is not a good solution to that.

I would prefer a sophisticated metric (popularity, karma) over a timestamp. I don't actually want to see last updates from all the people in my social graph, just the ones that are in some way significant. There was talk on filtering out stuff like #lunch on Twitter, but it's hard to really make a CAPTCHA judgement on the semantics of an update.

I believe Facebook already does that, try toggling "Top news" and "Most recent" to see it.

I've been thinking about adding something like this to http://friendbinder.com for a while now, though it's quite hard to get the filtering right to pick out the best stuff (I don't think just filtering the latest update is good enough).

yeah, facebook actually does a pretty good job of this, and so did friendfeed before fb acquired them; I wonder if the friendfeed guys helped improve it.

but as for twitter, I agree completely with the author's assessment

Yeah, somehow Facebook's fuzzy "Top News" system is the best out of the ones I've seen. Unfortunately you still get the loud-friend problem, and muting people requires a lot of thought (and it's not exportable to other clients). Plus, what does "Top News" even mean? As a programmer, I want to know what sorting method is used most of the time, and so it's a slight annoyance to just trust the Facebook to prioritize everything for me.

I'm signing up to friendbinder now, though, and I'd be happy to brainstorm with you.

Some time ago I was thinking about a UI design where "top" and "most recent" news are on the same timeline where you could zoom in to see more and more updates until you see every little update or zoom out where you could only see the most interesting ones. What do you think?

While the author proposes a solution, I don't believe his solution would provide the user with any more value from the pared-down stream. Instead, it would just minimize the deluge of information. Assuming that one is always interested in the most recent message from a user makes a conclusion that most recent = most important. As we all know from experience that isn't the case.

A better solution would systematically rank incoming message types based on prior determinations of importance of related messages (bayesian) and weight can be added to a particular user's messages (positive and negative) to help the cream rise to the top. The addition of collaborative filtering would help those that aren't interested in flagging their own incoming messages (up/down vote) still reap the benefits of the crowd.

Yes, as I was designing this, I thought, okay, 300 updates becomes ~100, fine. But what happens when I'm on even on more networks, and ~1,000 updates becomes ~300? Would then I need another layer on top of that to further reduce the deluge?

Maybe a "pin" and "unpin" feature that would pin these people to the top could work.

So what we have is a timeseries of user-post pairs, and we want a good view for that, a UI that scales well to handle pathological cases. I like the idea presented in the article, but, as you mention in this comment, it is only the first step. Perhaps a library for creating good views for this model would be good, and it should have some good primitives. Then people can develop and share new views for social news. The article's suggestion would be one view, and maybe some of these views would compose (i.e., after the transformation from the article, a timeseries of user-post pairs becomes a timeseries of user-(timeseries of posts) pars, where a (timeseries of posts) is presented as the most recent post and a control with the cardinality of the timeseries as its label and a toggle-hide behavior, and that thing fits in the outer UI the way a "post" would.)

None of the implications of this are explored. Such as that Twitter conversations would no longer be easily (for some values of easily) readable as such. Or that by having "Show More" links next to so many items, the user would constantly be made aware of (and thus anxious about) the fact that they are missing lots of information.

What gave me confidence that this could work, is that Google Buzz already does a little bit of grouping, and it seems elegant enough to me.

But, yes, I didn't explore the details too much about the "show more" feature. Would you list a count of the "unread items" from friends? Or would you need a slider that says "Show updates from the last 48 hours?" Then, would you need to couple this with some algorithm that bubbled up to the top the friends that you clicked "show more" on frequently? I'm not sure about the answers to all these questions, so I'll leave up to developers to figure it out.

I've always thought Twitter should have an auto generated list that shows 1 tweet per day per user. I think I added this to the wishlist for Tweetdeck (Groups) a while back. From glancing over the link, that solution sounds decent too.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact