4 ms·

Twitter Has A (Secret) Reputation Score For Every User

9 points by obilgic 16y ago

abraham 16y ago
Really TechCrunch? You are surprised by this?
- arfrank 16y ago
  That was my response too. It's like when Apple revealed their iPhone testing labs. Obviously these companies have internal tools/stats that people have no need to know about, but definitely make their company better able to perform.
  - protomyth 16y ago
    The iPhone testing lab story at least had some cool pictures.
- feral 16y ago
  Completely agree. There's a whole science to the study of networks, looking at things like Centrality, local topological features, and even ranking schemes like PageRank. Of course twitter are going to generate scores for their users as they do recommendation; even if it wasn't explicit, it would have to be implicit in any recommender system. TC would do well to have technical people look over their articles before publication. http://en.wikipedia.org/wiki/Centrality http://en.wikipedia.org/wiki/Centrality http://en.wikipedia.org/wiki/Social_network#Social_network_analysis http://en.wikipedia.org/wiki/Social_network#Social_network_a...
  - ojbyrne 16y ago
    That second to last paragraph made me laugh. TC (nor any other large tech blog) can't even get proofreaders to look over their articles before publication.
- wyclif 16y ago
  It was a slow news day at TechCrunch, there were no announcements from Apple today. This was just a fluff piece. Nothing to see here, move along.
Xuzz 16y ago
I'm also guessing this is used in the Top Tweets algorithm, which seems to favor lesser known users in the amount of interest needed to get chosen, but I have no way to back that up.
dotBen 16y ago
When I went to the launch/welcome party for Klout in the building for Twitter HQ, I invited a Twitter employee friend to "come down from the top floor" to check it out. He walked in and said "yeah we have this internally already, these guys are fucked". I think the idea of a reputation score around your twitter account makes total sense, and is definitely a marketable product. And the Klout team are a great bunch of guys. But when Twitter offers it there is no real point to a 3rd party option which is going to be inherently less accurate. Yes, Klout now calculates Facebook authority too but unless you have created an account and linked your profiles its meaningless (and what %age of folks have done that?). Klout dodged a bullet today in that Twitter didn't announce this info going into the API, but I don't think that's far out...
- urbanjunkie 16y ago
  "inherently less accurate"? I can see that there's potential for Twitter's offering to be good, but no guarantee that it's perfect. 'Reputation' is a nebulous concept, and Twitter certainly don't have a lock on how it's defined and calculated. Who knows which centrality measure they've chosen, how they're valuing and determining influence, or any number of other measures. The main advantage they have is of being closer to the raw data (and potentially better access to the entire data set, although a rigorous sampling methodology could take care of that)
  - dotBen 16y ago
    "The main advantage they have is of being closer to the raw data" That's my point. Both companies have the ability to come up with the most amazing formula for working out reputation but applied equally, Klout is going to be "inherently less accurate" because they have to trickle feed the data via the API and many interesting data points that may be stored in the Twitter database are not exposed on the API. Twitter has direct access to all data points.
    - urbanjunkie 16y ago
      Like what though - if you have enough information to generate a social graph, and you have access to the firehose, then you're going to be have what you need. Social scientists will often use a sample set as a very acceptable proxy for full data set analysis (http://www.insna.org/PDF/Connections/v18/1995_I-1-9.pdf http://www.insna.org/PDF/Connections/v18/1995_I-1-9.pdf) There's little evidence that the Twitter guys have come up with an outstandingly amazing reputation algorithm and that this niche has been filled. The opaqueness of Klout doesn't fill me with confidence that their methodology is valid or truly meaningful.
joshu 16y ago
one assumes that this is just the converse of a spam score...
bkudria 16y ago
If this news is surprising to you, you should rethink your mental model of, well, the business of tech startups.

9 points by obilgic 16y ago

abraham 16y ago

Really TechCrunch? You are surprised by this?

arfrank 16y ago
That was my response too. It's like when Apple revealed their iPhone testing labs. Obviously these companies have internal tools/stats that people have no need to know about, but definitely make their company better able to perform.
- protomyth 16y ago
  The iPhone testing lab story at least had some cool pictures.
feral 16y ago
Completely agree. There's a whole science to the study of networks, looking at things like Centrality, local topological features, and even ranking schemes like PageRank. Of course twitter are going to generate scores for their users as they do recommendation; even if it wasn't explicit, it would have to be implicit in any recommender system. TC would do well to have technical people look over their articles before publication. http://en.wikipedia.org/wiki/Centrality http://en.wikipedia.org/wiki/Centrality http://en.wikipedia.org/wiki/Social_network#Social_network_analysis http://en.wikipedia.org/wiki/Social_network#Social_network_a...
- ojbyrne 16y ago
  That second to last paragraph made me laugh. TC (nor any other large tech blog) can't even get proofreaders to look over their articles before publication.
wyclif 16y ago
It was a slow news day at TechCrunch, there were no announcements from Apple today. This was just a fluff piece. Nothing to see here, move along.

Xuzz 16y ago

I'm also guessing this is used in the Top Tweets algorithm, which seems to favor lesser known users in the amount of interest needed to get chosen, but I have no way to back that up.

dotBen 16y ago

When I went to the launch/welcome party for Klout in the building for Twitter HQ, I invited a Twitter employee friend to "come down from the top floor" to check it out. He walked in and said "yeah we have this internally already, these guys are fucked". I think the idea of a reputation score around your twitter account makes total sense, and is definitely a marketable product. And the Klout team are a great bunch of guys. But when Twitter offers it there is no real point to a 3rd party option which is going to be inherently less accurate. Yes, Klout now calculates Facebook authority too but unless you have created an account and linked your profiles its meaningless (and what %age of folks have done that?). Klout dodged a bullet today in that Twitter didn't announce this info going into the API, but I don't think that's far out...

urbanjunkie 16y ago
"inherently less accurate"? I can see that there's potential for Twitter's offering to be good, but no guarantee that it's perfect. 'Reputation' is a nebulous concept, and Twitter certainly don't have a lock on how it's defined and calculated. Who knows which centrality measure they've chosen, how they're valuing and determining influence, or any number of other measures. The main advantage they have is of being closer to the raw data (and potentially better access to the entire data set, although a rigorous sampling methodology could take care of that)
- dotBen 16y ago
  "The main advantage they have is of being closer to the raw data" That's my point. Both companies have the ability to come up with the most amazing formula for working out reputation but applied equally, Klout is going to be "inherently less accurate" because they have to trickle feed the data via the API and many interesting data points that may be stored in the Twitter database are not exposed on the API. Twitter has direct access to all data points.
  - urbanjunkie 16y ago
    Like what though - if you have enough information to generate a social graph, and you have access to the firehose, then you're going to be have what you need. Social scientists will often use a sample set as a very acceptable proxy for full data set analysis (http://www.insna.org/PDF/Connections/v18/1995_I-1-9.pdf http://www.insna.org/PDF/Connections/v18/1995_I-1-9.pdf) There's little evidence that the Twitter guys have come up with an outstandingly amazing reputation algorithm and that this niche has been filled. The opaqueness of Klout doesn't fill me with confidence that their methodology is valid or truly meaningful.

joshu 16y ago

one assumes that this is just the converse of a spam score...

bkudria 16y ago

If this news is surprising to you, you should rethink your mental model of, well, the business of tech startups.