Findory Blogory
June 17, 2004 9:12 PM Subscribe
Findory Bloggery is an an aggregator with a difference--this one (supposedly) learns which types of stories you enjoy reading, automatically suggesets more, and doesn't require a login.
Is that optional?
Looks like it. From the login page: "You don't have to login to use Findory Blogory and see personalized news stories"
posted by punishinglemur at 9:34 PM on June 17, 2004
Looks like it. From the login page: "You don't have to login to use Findory Blogory and see personalized news stories"
posted by punishinglemur at 9:34 PM on June 17, 2004
I assume that it does simply set a cookie (or perhaps track your IP?) to enable the login-lessness to work. I just discovered this service about ten minutes before I posted it here, so I'm not entirely sure yet how well it works, but I thought the idea was cool. There's been a couple of people experimenting with Bayesian filters in desktop RSS aggregators, as well: the Growlmurdurr aggregator, the read4me project, and probably others, but this is the first CentralizedShinyWidget[*] version I've seen.
apologies as necessary for stealing quonsar's sayings. smock smock smock.
posted by arto at 10:10 PM on June 17, 2004
apologies as necessary for stealing quonsar's sayings. smock smock smock.
posted by arto at 10:10 PM on June 17, 2004
I helped Findory when it started by seeding their DB off of memigo's database (note that memigo is similar to Findory News, but has different behavior and goals). Arto is correct, Findory works off of client-side cookies. Here's the blog of the Findory owner/coder/sole-proprietor, Greg Linden, who used to work for Amazon's personalization group.
Personally, as much as I like and I am (obviously) interested in this field, I think it's a few months away from Google moving in with Google Blogs and personalization in Google News to kill it off (disclaimer: at least my site pre-dates Google News by a full year+).
posted by costas at 10:28 PM on June 17, 2004
Personally, as much as I like and I am (obviously) interested in this field, I think it's a few months away from Google moving in with Google Blogs and personalization in Google News to kill it off (disclaimer: at least my site pre-dates Google News by a full year+).
posted by costas at 10:28 PM on June 17, 2004
I find the whole idea of auto-customized displays pretty troubling. Where's the serendipity? Most of the new ideas I get come from seeing things I didn't expect to see, making connections I didn't expect to make?
You get the same problem by relying on self-selected news sources, of course, but if you select broadly and scan headlines on them all, you at least have a shot at seeing something new. And sometimes, just seeing a headline is enough. A service like this can't possibly track what you read, what you see - only what you link to.
... And, of course, I worry about the algorithms. Given that it only knows about what I click on, not what I look at or what I think (and what I click on does not tell the system what I think, incidentally), its capability to really display stuff I'd want to see is dubious.
... And in what way is past behavior a reliable indicator of what I want to see, now? Maybe my mood changes; if I have a full palette to look at, maybe I won't stick wiht mauve for five weeks running.
... And who's to say that the data is used "honestly", whatever that means? Let's say I show a preponderance of clicks on articles about John Kerry; Findory might start seeding my results with articles about George Bush. (They'll all mention Kerry, right?) Or it might preferentially target articles
... And the targeting can only be as good as the metadata. If the metadata is supplied with the article, that's a problem, because the publishers get to skew their own placement. If the metadata is derived, a la Google Page Rank, then that's a problem, because we all know how that can be skewed.
I know what the supposed motivators are for a system like this (they generally involve contorted arguments about programmatically achieving serendipity), and I know what the real ones are. And at the end of my browsing day, one thing would remain true: Findory would have mined valuable, resaleable information from me, without much real value having been delivered in return -- but with a lot of fancy words that might lead someone to think they'd gotten value. In fact, I'd argue that you've gotten negative value -- you've degraded the quality of your experience, without even knowing it.
ON PREVIEW: I look at my litany of objections, and I don't believe they'll make a difference to anybody reading this. This is really all about being cool, and doing the bleeding edge thing. We're all first-adopters here, right? What right have I got to be critical? I buy the fancy new gadgets and I use news aggregators. The answer is that living in the territory doesn't absvole me of an obligation to live in it thoughtfully, critically.
posted by lodurr at 10:35 AM on June 18, 2004
You get the same problem by relying on self-selected news sources, of course, but if you select broadly and scan headlines on them all, you at least have a shot at seeing something new. And sometimes, just seeing a headline is enough. A service like this can't possibly track what you read, what you see - only what you link to.
... And, of course, I worry about the algorithms. Given that it only knows about what I click on, not what I look at or what I think (and what I click on does not tell the system what I think, incidentally), its capability to really display stuff I'd want to see is dubious.
... And in what way is past behavior a reliable indicator of what I want to see, now? Maybe my mood changes; if I have a full palette to look at, maybe I won't stick wiht mauve for five weeks running.
... And who's to say that the data is used "honestly", whatever that means? Let's say I show a preponderance of clicks on articles about John Kerry; Findory might start seeding my results with articles about George Bush. (They'll all mention Kerry, right?) Or it might preferentially target articles
... And the targeting can only be as good as the metadata. If the metadata is supplied with the article, that's a problem, because the publishers get to skew their own placement. If the metadata is derived, a la Google Page Rank, then that's a problem, because we all know how that can be skewed.
I know what the supposed motivators are for a system like this (they generally involve contorted arguments about programmatically achieving serendipity), and I know what the real ones are. And at the end of my browsing day, one thing would remain true: Findory would have mined valuable, resaleable information from me, without much real value having been delivered in return -- but with a lot of fancy words that might lead someone to think they'd gotten value. In fact, I'd argue that you've gotten negative value -- you've degraded the quality of your experience, without even knowing it.
ON PREVIEW: I look at my litany of objections, and I don't believe they'll make a difference to anybody reading this. This is really all about being cool, and doing the bleeding edge thing. We're all first-adopters here, right? What right have I got to be critical? I buy the fancy new gadgets and I use news aggregators. The answer is that living in the territory doesn't absvole me of an obligation to live in it thoughtfully, critically.
posted by lodurr at 10:35 AM on June 18, 2004
Lodurr: you either find a service like this valuable, or you don't: I cann't convince you that you will like something if you don't think there's value in it. At any rate, I will speak for myself and memigo and say that I developed it exactly to programmatically achieve serendipity (at least for Web news). Memigo is not how I make a living or intend to make a living (I got another startup to worry about in a completely different space): I wrote it in my spare time so I could learn a few things about stuff I was interested in and have a serendipity agent as well --'cause there wasn't, and there isn't quite anything like it out there.
Now, wrt some of your specific concerns:
Change of mood/interests: interesting data is recent data: memigo ages your ratings, stuff older than a few months is practically thrown away (but not quite).
Past-behavior predicting future behavior: as much as people would like to disagree, past behavior is a good indicator of future behavior, but more importantly any good personalization system is going to use other people's ratings to guess what other new things you may be interested as well (Amazon-like).
Meta-data: metadata is always derived these days after the debacle of the meta tag. True, Google is gamed meta-data wise, but Google is working off an open metadata system (the Web), wheareas personalization systems are usually mostly-closed. You can still game them, but it's much harder.
Trusting the recommendations: you got me there. If you don't trust your recommendations provider, switch.
posted by costas at 11:19 AM on June 18, 2004
Now, wrt some of your specific concerns:
posted by costas at 11:19 AM on June 18, 2004
« Older Crazy | He sure can down a cheeseburger Newer »
This thread has been archived and is closed to new comments
Ironic, though, that the front page is currently filled with Talking Points Memo posts at the exact moment that Josh Marshall has gone on vacation (while teasing that "I and several colleagues have been working on a story that, if and when it comes to fruition --- and I’m confident it shall --- should shuffle the tectonic plates under that capital city where I normally hang my hat. So that’s something to look forward to in the not too distant future.")
Anyway, I just may give this a try. Thanks.
posted by soyjoy at 9:28 PM on June 17, 2004