Question for datawankers about the most highly favorited posts. June 18, 2012 10:46 AM Subscribe
Question for datawankers about the most highly favorited posts.
I have this theory and I was wondering if any datawankers have been curious about this before. It seems like people are more likely to favorite a post or comment that already has a lot of favorites.
So, when a good post/comment is first made, it might get 20 favorites. If it gets linked a few months later, it might get another 20. If it gets linked a third time, months later, then it might get 40. A fourth time, it might now get 80.
It seems like a higher percentage of people reading a post - the same post - will favorite it when it has more faves already than when it has less.
Has anyone else gotten curious about this?
I have this theory and I was wondering if any datawankers have been curious about this before. It seems like people are more likely to favorite a post or comment that already has a lot of favorites.
So, when a good post/comment is first made, it might get 20 favorites. If it gets linked a few months later, it might get another 20. If it gets linked a third time, months later, then it might get 40. A fourth time, it might now get 80.
It seems like a higher percentage of people reading a post - the same post - will favorite it when it has more faves already than when it has less.
Has anyone else gotten curious about this?
When you attempt this analysis remember that not everyone has the "show favorite counts" option set in their profile.
posted by ceribus peribus at 10:58 AM on June 18, 2012
posted by ceribus peribus at 10:58 AM on June 18, 2012
I feel like at some point in the past we looked at things that got a "favorites boost" once they hit 12 favorites which is the point at which they show up on the sidebar and other people would see them and possibly then favorite them. I have favorites turned off so I don't even have anecdata for this.
posted by jessamyn (staff) at 11:02 AM on June 18, 2012 [1 favorite]
posted by jessamyn (staff) at 11:02 AM on June 18, 2012 [1 favorite]
It's an interesting question, and it's also a complicated one because there's a lot of things that could be going on with how many people and who are seeing an item at any given time. You might need to answer questions like:
- What's the level/kind of visibility of each revisiting event? (mefi sidebar? BestOf post? mention in a Metatalk post? mention in a comment in a thread? link on a different site?)
- Is the new revisiting event different in kind from previous ones? You might see a bigger pile of new favorites is an askme post shows up in meta for the first time vs. being referenced again in an otherwise thematically similar askme, etc.
- How long since the original post or previous revisiting event? Bigger gaps might suggest more disjoint groups of potential favoriters as the population changes over time.
- How has the overall favoriting rate of the site changed since the original post or previous revisiting event? If favoriting rate in general grows over time, then it may be reasonable to expect more favorites to appear at any given kind of revisiting event at a later date than a similar event at an earlier date even if all else is held equal.
- Is the current favorite count on something contributing to its visibility via Popular Favorites, the contact activity sidebar, or people playing around with data via the Infodump in Metatalk or over on the Infodumpster?
I'd say that generally speaking, something having a lot of favorites puts it in the position of being (a) likely to be recalled in futures situations and (b) more likely to show up in datawankery type analyses as an outlier, and between that and the (last I remember looking) trend of growth in total favorites given out per year, it's likely that lots of favorites begets a better chance of lots more favorites, but I'd be curious about a more in-depth analysis.
I'd also wonder whether there's that much in common between how whatever multiplying effects might play out for highly-favorited stuff and how things play out for things with more moderate numbers of favorites. The latter group is by far the majority of cases, so if the effects aren't same it could be easy to draw wrong conclusions about how favoriting feedback effects behave almost all of the time vs. how they behave for outlier cases.
posted by cortex (staff) at 11:07 AM on June 18, 2012 [1 favorite]
- What's the level/kind of visibility of each revisiting event? (mefi sidebar? BestOf post? mention in a Metatalk post? mention in a comment in a thread? link on a different site?)
- Is the new revisiting event different in kind from previous ones? You might see a bigger pile of new favorites is an askme post shows up in meta for the first time vs. being referenced again in an otherwise thematically similar askme, etc.
- How long since the original post or previous revisiting event? Bigger gaps might suggest more disjoint groups of potential favoriters as the population changes over time.
- How has the overall favoriting rate of the site changed since the original post or previous revisiting event? If favoriting rate in general grows over time, then it may be reasonable to expect more favorites to appear at any given kind of revisiting event at a later date than a similar event at an earlier date even if all else is held equal.
- Is the current favorite count on something contributing to its visibility via Popular Favorites, the contact activity sidebar, or people playing around with data via the Infodump in Metatalk or over on the Infodumpster?
I'd say that generally speaking, something having a lot of favorites puts it in the position of being (a) likely to be recalled in futures situations and (b) more likely to show up in datawankery type analyses as an outlier, and between that and the (last I remember looking) trend of growth in total favorites given out per year, it's likely that lots of favorites begets a better chance of lots more favorites, but I'd be curious about a more in-depth analysis.
I'd also wonder whether there's that much in common between how whatever multiplying effects might play out for highly-favorited stuff and how things play out for things with more moderate numbers of favorites. The latter group is by far the majority of cases, so if the effects aren't same it could be easy to draw wrong conclusions about how favoriting feedback effects behave almost all of the time vs. how they behave for outlier cases.
posted by cortex (staff) at 11:07 AM on June 18, 2012 [1 favorite]
It would be cool if there was a sidebar thingy that told you when your contacts posts/comments make it to the "Popular" page... e.g. "Jessamyn's post "Librarians Read Books" made it to the Popular Page" or "Blake's Comment on "Librarians Read Books" made it to the popular page"
posted by Blake at 11:07 AM on June 18, 2012
posted by Blake at 11:07 AM on June 18, 2012
I've long suspected, but do not know, that there's probably some self-reinforcing going on of favorites. For example, it feels like it sometimes takes hours for a comment on Ask.Me to get one favorite, and then will get a couple more really quickly -- like it was that first favorite that prompted other people to also favorite it.
posted by jacquilynne at 11:10 AM on June 18, 2012
posted by jacquilynne at 11:10 AM on June 18, 2012
On the other hand sometimes I will not favorite a heavily-favorited comment out of a misguided attempt to maintain my independence and renegade spirit.
posted by shakespeherian at 11:14 AM on June 18, 2012 [10 favorites]
posted by shakespeherian at 11:14 AM on June 18, 2012 [10 favorites]
The contact activity sidebar thing is a little further complicated by the fact that different people have different numbers of contacts, so someone who has a lot of people who've contacted them will have a correspondingly larger potential favorite bump on an item than someone who few people have contacted.
But also, someone who has contacted a lot of people will have a busier sidebar than someone who has contacted only a few, so the chances of seeing a favorite bump from a given contactee will potentially vary on that front too.
I am almost certain that someone did a bit of playing around with this stuff previously, yeah. I'll see if I can find something on the wiki.
posted by cortex (staff) at 11:17 AM on June 18, 2012
But also, someone who has contacted a lot of people will have a busier sidebar than someone who has contacted only a few, so the chances of seeing a favorite bump from a given contactee will potentially vary on that front too.
I am almost certain that someone did a bit of playing around with this stuff previously, yeah. I'll see if I can find something on the wiki.
posted by cortex (staff) at 11:17 AM on June 18, 2012
I notice this, too. I feel like I have very few comments with only 6-12 favorites. Once something hits a threshold of about 5 seed favorites, the count tends to pretty consistently shoot up to around 20 or 30.
posted by phunniemee at 11:20 AM on June 18, 2012
posted by phunniemee at 11:20 AM on June 18, 2012
By the way, this is the (fantastic) post that got me thinking about this. It was relinked a few times in the space of a few months (once by me), and just on Ask. It's interesting to see what happened every time it was linked.
posted by cairdeas at 11:21 AM on June 18, 2012
posted by cairdeas at 11:21 AM on June 18, 2012
On the other hand sometimes I will not favorite a heavily-favorited comment out of a misguided attempt to maintain my independence and renegade spirit.
I will admit to having a somewhat repulsive little frisson of self-satisfied delight upon having been the very first person to favourite a later-heavily-favourited comment that initially seemed to have gone unnoticed.
posted by elizardbits at 11:21 AM on June 18, 2012 [3 favorites]
I will admit to having a somewhat repulsive little frisson of self-satisfied delight upon having been the very first person to favourite a later-heavily-favourited comment that initially seemed to have gone unnoticed.
posted by elizardbits at 11:21 AM on June 18, 2012 [3 favorites]
oh fuck i am a comment hipster
posted by elizardbits at 11:22 AM on June 18, 2012 [11 favorites]
posted by elizardbits at 11:22 AM on June 18, 2012 [11 favorites]
I would like to claim elizardbits as my favorite.
posted by shakespeherian at 11:23 AM on June 18, 2012 [4 favorites]
posted by shakespeherian at 11:23 AM on June 18, 2012 [4 favorites]
I have this theory and I was wondering if any datawankers have been curious about this before...
For centuries. This is the bandwagon effect.
posted by weapons-grade pandemonium at 11:28 AM on June 18, 2012
For centuries. This is the bandwagon effect.
posted by weapons-grade pandemonium at 11:28 AM on June 18, 2012
You're not a true comment hipster until you un-favorite it because it made the sidebar and you can no longer take the awful opinions those people have about it now that it's broken through, but then quickly re-favorite it just so that, down the line, no one can ever claim you weren't there when it happened.
Wait, hang on, am I talking about favorites or Joanna Newsom?
posted by griphus at 11:45 AM on June 18, 2012
Wait, hang on, am I talking about favorites or Joanna Newsom?
posted by griphus at 11:45 AM on June 18, 2012
Real comment hipsters print out their favorite comments and paste them into their moleskine and mark the best ones with vintage paper bookmarks.
posted by cortex (staff) at 11:47 AM on June 18, 2012 [5 favorites]
posted by cortex (staff) at 11:47 AM on June 18, 2012 [5 favorites]
Real comment hipsters favorited all of Tao Lin's comments before he was banned.
posted by shakespeherian at 11:49 AM on June 18, 2012 [1 favorite]
posted by shakespeherian at 11:49 AM on June 18, 2012 [1 favorite]
Soundtrack for those bemoaning the bandwagon effect: Bandwagon Blues (note: NOT ACTUAL BLUES). Soundtrack for those bemoaning the jump-up style heard in Bandwagon Blues: Rinse It Out Propa (NSFW words at moments), aka Blandwagon Poos.
posted by filthy light thief at 11:51 AM on June 18, 2012
posted by filthy light thief at 11:51 AM on June 18, 2012
oh fuck i am a comment hipster
My [+] is a fixie.
posted by arcticseal at 12:20 PM on June 18, 2012 [1 favorite]
My [+] is a fixie.
posted by arcticseal at 12:20 PM on June 18, 2012 [1 favorite]
Datawankers?
Datawanker, n.
One who engages in the practice of datawankery; a data nerd; a wonk.
posted by cortex (staff) at 12:26 PM on June 18, 2012
Datawanker, n.
One who engages in the practice of datawankery; a data nerd; a wonk.
posted by cortex (staff) at 12:26 PM on June 18, 2012
Shaaaaakes.
My dawg.
My main man.
My ace in the hole.
posted by SpiffyRob at 12:49 PM on June 18, 2012
My dawg.
My main man.
My ace in the hole.
posted by SpiffyRob at 12:49 PM on June 18, 2012
Finally, a post on metatalk that confirms a connection between wankers and favorites.
posted by crunchland at 12:50 PM on June 18, 2012 [2 favorites]
posted by crunchland at 12:50 PM on June 18, 2012 [2 favorites]
I sort of wanted to favorite some of the above comments but now I feel like I'd be making some sort of inadvertent meta-statement by doing so...
posted by mlle valentine at 12:51 PM on June 18, 2012 [1 favorite]
posted by mlle valentine at 12:51 PM on June 18, 2012 [1 favorite]
[sidehug]
posted by shakespeherian at 12:52 PM on June 18, 2012 [1 favorite]
posted by shakespeherian at 12:52 PM on June 18, 2012 [1 favorite]
If that were true you would expect favorites to follow a preferential attachment model -- so more power-law than Poisson, right? That seems pretty testable. But there's probably a better, more explicit way of formulating the question statistically.
posted by en forme de poire at 2:36 PM on June 18, 2012 [1 favorite]
posted by en forme de poire at 2:36 PM on June 18, 2012 [1 favorite]
You won't see this jessamyn, but I favorited your post because I had never heard of "anecdata" before.
posted by cgk at 3:35 PM on June 18, 2012
posted by cgk at 3:35 PM on June 18, 2012
cortex: "Datawankers?
Datawanker, n.
One who engages in the practice of datawankery; a data nerd; a wonk."
I understood the context. I just didn't know this was a thing. First time I ever saw the word. I must lead a sheltered life.
posted by Splunge at 4:48 PM on June 18, 2012
Datawanker, n.
One who engages in the practice of datawankery; a data nerd; a wonk."
I understood the context. I just didn't know this was a thing. First time I ever saw the word. I must lead a sheltered life.
posted by Splunge at 4:48 PM on June 18, 2012
oh man .. someone please post about Microsoft's new tablet. I can't post I am too busy throwing money and credit cards at my monitor. Why won't Balmer just take my money and send me one.
posted by Ad hominem at 5:30 PM on June 18, 2012
posted by Ad hominem at 5:30 PM on June 18, 2012
I thought wanker was an insult so it surprises me that this would be an okay word. Then again, geek used to be an insult too.
posted by IndigoRain at 5:32 PM on June 18, 2012
posted by IndigoRain at 5:32 PM on June 18, 2012
Ballmer is talking about how it's a desktop replacement and they don't mention price or a ship date. Yep. They copied every aspect of an Apple new product presentation except the two most incredibly important parts: WHEN, and HOW MUCH. I mean, yeah, it was really cool, and I can see a lot of nice ideas and promises, but ultimately it was just a sad little wank session ending in friction burn and frustration.
posted by seanmpuckett at 6:04 PM on June 18, 2012
posted by seanmpuckett at 6:04 PM on June 18, 2012
I thought wanker was an insult so it surprises me that this would be an okay word. Then again, geek used to be an insult too.
Well, wanker is an insult, more commonly so in the UK than in the US, but wanking is more of a hobby, and the metaphorical notion of wanking around with data is more a sort of self-deprecating term for the sake of a laugh than anything. I wouldn't go calling random strangers datawankers in a null context or anything, but I'll accuse myself of datawankery without pause.
posted by cortex (staff) at 6:38 PM on June 18, 2012
Well, wanker is an insult, more commonly so in the UK than in the US, but wanking is more of a hobby, and the metaphorical notion of wanking around with data is more a sort of self-deprecating term for the sake of a laugh than anything. I wouldn't go calling random strangers datawankers in a null context or anything, but I'll accuse myself of datawankery without pause.
posted by cortex (staff) at 6:38 PM on June 18, 2012
cortex: "I thought wanker was an insult so it surprises me that this would be an okay word. Then again, geek used to be an insult too.
Well, wanker is an insult, more commonly so in the UK than in the US, but wanking is more of a hobby, and the metaphorical notion of wanking around with data is more a sort of self-deprecating term for the sake of a laugh than anything. I wouldn't go calling random strangers datawankers in a null context or anything, but I'll accuse myself of datawankery without pause."
Hobby? Hobby! Wanking is a way of life, man. I'd say more but... BRB.
posted by Splunge at 7:00 PM on June 18, 2012
Well, wanker is an insult, more commonly so in the UK than in the US, but wanking is more of a hobby, and the metaphorical notion of wanking around with data is more a sort of self-deprecating term for the sake of a laugh than anything. I wouldn't go calling random strangers datawankers in a null context or anything, but I'll accuse myself of datawankery without pause."
Hobby? Hobby! Wanking is a way of life, man. I'd say more but... BRB.
posted by Splunge at 7:00 PM on June 18, 2012
I am almost certain that someone did a bit of playing around with this stuff previously, yeah.
I know I've looked at this a bit, and maybe others have too. In this comment, I looked to see if there was any change in favoriting "speed" once a comment has enough favorites to appear on the contact activity side bar. There seems to be, and only after the contact activity sidebar feature came along, but the effect (if real) looks pretty small.
posted by FishBike at 7:00 PM on June 18, 2012
I know I've looked at this a bit, and maybe others have too. In this comment, I looked to see if there was any change in favoriting "speed" once a comment has enough favorites to appear on the contact activity side bar. There seems to be, and only after the contact activity sidebar feature came along, but the effect (if real) looks pretty small.
posted by FishBike at 7:00 PM on June 18, 2012
I'm more of a Lorefrigger.
posted by fleacircus at 7:55 PM on June 18, 2012
posted by fleacircus at 7:55 PM on June 18, 2012
BTW, I think swipe to fav breaks the old method of favoriting. I turned it on for my iPad and can't fav FPPs anymore.
posted by Ad hominem at 7:58 PM on June 18, 2012
posted by Ad hominem at 7:58 PM on June 18, 2012
I suspect that this, like so many things in social media, is subject to a power law curve.
posted by Chocolate Pickle at 9:31 PM on June 18, 2012
posted by Chocolate Pickle at 9:31 PM on June 18, 2012
SWIPE TO FAV
976-FAVS
$7.99 for the first minute. $0.99 each add'l minute.
FAV ME TONIGHT!!!
[+]
posted by not_on_display at 9:36 PM on June 18, 2012
976-FAVS
$7.99 for the first minute. $0.99 each add'l minute.
FAV ME TONIGHT!!!
[+]
posted by not_on_display at 9:36 PM on June 18, 2012
Obligatory response to power laws: So You Think You Have a Power Law — Well Isn't That Special?
posted by stebulus at 9:46 PM on June 18, 2012 [3 favorites]
posted by stebulus at 9:46 PM on June 18, 2012 [3 favorites]
Anecdotally: I took a quick look at my "favorited by others" pages. My impressions:
1) I have two high-favorite outliers at 96 and 126 favorites
2) The main body of favorites seems to otherwise have a fairly arithmetic progression, starting at around 65. By this I mean that as you go down the list (sorted by popularity) you see a pretty steady progression (65 favs, 64 favs, 63 favs, etc) with each quantity of favorites having a pretty similar number of posts.
3) It starts to shift as you get lower in favorites, such that more and more posts have that number of favorites. This really begins to get noticeable around 10-11 favorites.
4) Posts with one favorite are by far more numerous than any other number of favorites.
I want to make a graph that illustrates this, but the graphical editing tools I have access to are really clunky. =|
posted by kavasa at 10:35 PM on June 18, 2012
1) I have two high-favorite outliers at 96 and 126 favorites
2) The main body of favorites seems to otherwise have a fairly arithmetic progression, starting at around 65. By this I mean that as you go down the list (sorted by popularity) you see a pretty steady progression (65 favs, 64 favs, 63 favs, etc) with each quantity of favorites having a pretty similar number of posts.
3) It starts to shift as you get lower in favorites, such that more and more posts have that number of favorites. This really begins to get noticeable around 10-11 favorites.
4) Posts with one favorite are by far more numerous than any other number of favorites.
I want to make a graph that illustrates this, but the graphical editing tools I have access to are really clunky. =|
posted by kavasa at 10:35 PM on June 18, 2012
Let me amend that: the posts in the 60s are outliers as well, there's a big jump from those to posts with favs in the 40s.
posted by kavasa at 10:37 PM on June 18, 2012
posted by kavasa at 10:37 PM on June 18, 2012
I'll accuse myself of datawankery without pause.
I don't think it's wank-negative of me to point out that perhaps letting a little more time pass between datawank sessions might be beneficial.
posted by Rhomboid at 1:40 AM on June 19, 2012
I don't think it's wank-negative of me to point out that perhaps letting a little more time pass between datawank sessions might be beneficial.
posted by Rhomboid at 1:40 AM on June 19, 2012
I think you would need better tools than we have available to answer this one. When my answer about public speaking jumped from, like, 3 favorites to 40, a bunch of stuff had happened all at once (the sidebar, the "best of" blog, etc.) such that I don't think it'd be possible to sort out which favorites were the result of what action/stimulus. I still occasionally get a new favorite, and I'd assume I'd been linked somewhere if that became four or five new favorites at once. But you'd still have to sort out what role the existing favorites (as opposed to the content of the answer) really played.
I'm quite sure "having more favorites" is correlated with "obtaining even more favorites," but it'd be beyond tricky to sort out the causation side.
posted by SMPA at 6:16 AM on June 19, 2012
I'm quite sure "having more favorites" is correlated with "obtaining even more favorites," but it'd be beyond tricky to sort out the causation side.
posted by SMPA at 6:16 AM on June 19, 2012
Ha, yeah, stebulus, I know... People in my field are obsessed with pointing out power laws for things, often without much evidence beyond that log-log plot, so I'm pretty allergic to them at this point as well. Still, if the preferential attachment thing really applies to favorites, it seems like you should be able to do some kind of likelihood test (e.g. Vuong) using a Poisson as the null and a Poisson-with-attachment (Zipf?) as the alternative. (Not sure what the other long tailed discrete distributions are.)
posted by en forme de poire at 8:15 AM on June 19, 2012 [1 favorite]
posted by en forme de poire at 8:15 AM on June 19, 2012 [1 favorite]
About 90 people have added me as a contact, but most of them haven't favorited me more than a few times, from what I can tell. My most frequent favoriters aren't contacts at all, so I'm skeptical about how much impact the sidebar has, tbh.
posted by empath at 8:32 AM on June 19, 2012
posted by empath at 8:32 AM on June 19, 2012
(the obvious ones, I mean - like lognormal for the continuous case)
posted by en forme de poire at 8:32 AM on June 19, 2012
posted by en forme de poire at 8:32 AM on June 19, 2012
I thought wanker was an insult so it surprises me that this would be an okay word-- IndigoRain
I think it's safe to say that most British men would be insulted if you called them a wanker, and most British women puzzled. It's a fairly strongly gendered insult roughly equivalent, from what I can tell, to asshole or jackass.
posted by ambrosen at 10:41 AM on June 19, 2012
I think it's safe to say that most British men would be insulted if you called them a wanker, and most British women puzzled. It's a fairly strongly gendered insult roughly equivalent, from what I can tell, to asshole or jackass.
posted by ambrosen at 10:41 AM on June 19, 2012
Real comment hipsters print out their favorite comments and paste them into their moleskine and mark the best ones with vintage paper bookmarks.
Print? Noooo. They would sit on their rescued art deco couch covered with a sustainaby-woven organic cotton throw made by artisans in India and infused with the cat hair from two adopted feline mutts, Moleskine in hand, and carefully recopy favorite comments into it with a fountain pen. And really, who needs bookmarks when you have your great-grandmother's leftover ribbons.
posted by fraula at 11:58 AM on June 19, 2012
Print? Noooo. They would sit on their rescued art deco couch covered with a sustainaby-woven organic cotton throw made by artisans in India and infused with the cat hair from two adopted feline mutts, Moleskine in hand, and carefully recopy favorite comments into it with a fountain pen. And really, who needs bookmarks when you have your great-grandmother's leftover ribbons.
posted by fraula at 11:58 AM on June 19, 2012
fraula, I hardly need mention that when cortex said print, he did at least mean artisan letterpress. That goes without saying.
posted by ambrosen at 1:48 PM on June 19, 2012 [1 favorite]
posted by ambrosen at 1:48 PM on June 19, 2012 [1 favorite]
carefully distressed antique composing sticks loaded with handcarved ironic Papyrus and Comic Sans sorts.
posted by elizardbits at 5:29 AM on June 20, 2012
posted by elizardbits at 5:29 AM on June 20, 2012
I like to think through careful favoriting, over a number of years, this site will contain my personality as a whole solely in my favorite records. Aliens/future archaeologists feel free to use this information to emulate me back to life
posted by MangyCarface at 1:04 PM on June 20, 2012 [1 favorite]
posted by MangyCarface at 1:04 PM on June 20, 2012 [1 favorite]
I've noticed this, and my theory would be that some of it comes from people who are scanning through responses. Once you see a response that has collected a few favorites, I think you're more likely to read it in greater detail, and you're primed with the impression that it's good advice, so you're more likely to add it as a favorite.
Maybe you could test the power law hypothesis and see if there's a change in the derivative of the curve that would indicate that there's a threshold above which more favorites tend to accrue?
posted by alphanerd at 11:37 AM on June 22, 2012
Maybe you could test the power law hypothesis and see if there's a change in the derivative of the curve that would indicate that there's a threshold above which more favorites tend to accrue?
posted by alphanerd at 11:37 AM on June 22, 2012
Hah... Obviously it applies not just to "advice" but to all comments. I sort of geared my response to what happens in the human relations section of AskMe because I read it so much.
posted by alphanerd at 11:41 AM on June 22, 2012
posted by alphanerd at 11:41 AM on June 22, 2012
You are not logged in, either login or create an account to post comments
So more favorites gives a post more exposure through the "popular" sortings, I think, and then those members who are using that method as a "MeFi Lite" (if you will) are probably going to favorite from that limited pool - therefore, more favorites will beget more favorites. Yeah?
posted by flex at 10:58 AM on June 18, 2012 [2 favorites]