problem but it's a national level where it's much more of a challenge. >> host: you're going to get another try at this call from california. we missed you earlier. you are on the air now. >> caller: yes. i'm interested in the influence of fred martin said central asia -- >> host: i'm glad you asked the question because i wanted to point out the audience greg martin's and provides the jacket for your book and since it's been published the full 60 minutes controversy so what is his influence as the call wants to know and what are your thoughts on the controversy? >> guest: his had always been not in the cities and this takes place in kabul so i think that his influence in terms of afghans the life that over the years especially at the time spent in kabul was less because that isn't where the central asia institute ever operated. it's where they are mostly.

i can get is an incredibly sad turn of events. i was as surprised as anybody else and i think that what is so important as people realize that the issue of the girls' education is so much more important than the individuals whoever it is, whenever the storyteller is because there are girls in the story with kidnapping and potential acid attacks just for the opportunity to go to school and i think that is the biggest hurdle is keeping most people will never meet those young women fighting every day to go to class. >> host: much has been recovered in the future of afghanistan. as we close out here, what is the one take away that you want people to have about the country's based on your experience in this book? >> guest: what most people want and afghanistan is what most people want in this country, the ability to send their kids to school, the ability to feed their families, and the ability to make sure

that the next generation has a better shot the future that's peaceful and that is what i hear over and over and i could because the coverage shapes so much of what we know about afghanistan a lot of people won't ever meet those people and i hope they get to know some of these unsung heroines to be posed to you had such a success are you already thinking and not the second book? >> guest: ibm because it is sort of mission driven if you've gotten people to pay attention to the women's stories and not so much soft when i think the work the women do is very hard. i want to go to liberia and i met this really compelling entrepreneur who has a very dramatic story and is now running a pretty fascinating business there. >> host: thanks for being with us. again, the dressmaker khair kahn is gayle a lemmon's first book and you can find it easily weather on line as an ebook or in your local booksellllllllllll

>> this was part of the 2011 los angeles times festival of books. to find out more visit events.latimes.com/festival of books. ♪ >> coming up next, booktv presents "after words," an hourlong program where we invite guest hosts to interview authors. this week eli pariser exposes internet secrets in his new book, "the filter bubble: what the internet is hiding from you." in it the president of moveon.org argues that social media and online companies are all manipulating content to fit users, personalizing information in a potentially detrimental way. mr. parisser discusses his concerns and solutions for a loss of privacy with author and professor clay shirk key.

>> host: eli, great to be here with you. >> guest: thank you. >> host: i'm going to start by asking, i think, the most basic question about the thesis. your book is titled: "filter bubble: what the internet is hiding from you." who is it that's doing this hide, and what is it that we're missing when things are being hidden from us? >> guest: well, right. this is a shortcut, but, you know, certainly, increasingly a number of the top web sites and web services, and, you know, even other applications are editing our experience of the web in a personalized way. that means, essentially, they're collecting data from us and using it to carry it to, you know, prioritize pieces of data and show us the things that they think we're going to want to

see. at google this happens in the search results. different people get different search results for the same term based on what they've clicked on in the past, what they, what google knows about what kind of person they are. at yahoo! news it happens on yahoo! news, it's actually the stories themselves. and this is spreading across, you know, it's being baked into more and more services online for a simple reason which is that, you know, if you can provide this sort of, you know, the code word is relevance, then you keep people coming back, and you presumably provide a better experience for them. but the danger here is that because it's very hard to see how your internet differs from anyone else's or on what basis it's being edited or cure rated, you don't know what's being left out. you don't know what you're not seeing that other people might be seeing, and most people don't

even have a sense that this editing is happening at all. >> host: so the internet is turning a lot of the publicly-accessible content into known unknowns, right? >> >> guest: right. >> host: you're not even aware that it's out there. you talked about google knowing what kind of people we are, using that as a way of filtering searches, yahoo! doing the same thing with yahoo! news and so forth. yet this isn't just a matter of being logged in. you're saying google is doing this even if i don't have a google account, even if i'm not logged in as a g-mail user? google is still doing this? >> guest: this is the big shift. google, one engineer told me they collect 57 different variables about people who even if you're not logged in, if you take a new laptop and put it on the desk right here, you can tell what kind of laptop is it, what kind of software is it running, what is the size of the fonts on this laptop, where is

it located? what's the ip address? how long are you lingering before you click on a link? and all of these things can be used to make some guesses about what kind of person you are. there's mac and pc people, big font and small font people. all of this, then, allows you to make these guesses. they may not be, i mean, this may not be a very good portrait of you. it almost certainly isn't at that point. but you don't need that much in order to be able to do this with an increase in on the many iization. and -- opt myization. i talked to the people at hunch, and, you know, they said actually you need very little data in order to start to have a lot of predictive power. so five data points, five particular data points you can then get any other data point with an 80% accuracy. >> host: data points sounds a little vague. give this to me -- what would they need to know about me to

make these kinds of predictions? >> hunch said it was, and i can remember about four of them. so i'll do the four. there's gender, there's introvert/extrovert turns out to be a very important signal -- >> host: oh, that's interesting. >> guest: there's political orientation which is a proxy for a lot of lifestyle things if you say vespa or harley, that turns out to be very -- >> host: recognize the word vespa, you're, obviously, a liberal. >> guest: a coastal -- [laughter] and, you know, then some kind of proxy for your generation, you know, for generational taste, essentially. do you like sort of the hilton or some sort of trendier hotel? sort of like -- >> host: generational tastes rather than age. so hilton versus the w as a hotel choice, is that, is that a

better predicter than just knowing if someone's 25 or -- >> guest: well, i think it's a proxy for age. hunch tries to ask questions that don't directly ask you the question that they're trying to get at a lot of the times. >> host: oh, i see. >> guest: and so, you know, that's -- you could say liberal or key, but harley or vespa is more fun, and, you know -- >> >> host: people don't get weird about the answer in a way that they would feel -- >> guest: that's right. no, it helps get the information out that you need. but with five data point you can then, you know, predict out with 90% accuracy -- 80 percent accuracy almost anything about taste. and not only that, but if you have two people answering those questions and you know that they're friends with a third person and you don't know anything about that third person, you can still actually predict that person's tastes within 80% accuracy having

gathered these -- >> host: so this isn't even a personal issue anymore. the social networks that we're all a part of are exporting information about out -- >> guest: that's right. >> host: even if i'm not? >> guest: it's one of these things i don't -- i think that people who say privacy is over often have an interest in saying that, and i don't totally buy that. but there is this way in which, certainly, what is known about us can increasingly be inferred by what is known about our friends. >> host: right. >> guest: and it's in their hands in some sense. we've crowd sourced it. >> host: there was a paper written out of mit saying we can infer your sexual preference, and in particular we can figure out whether you are g ark y or not -- gay or not. you're saying that is a general principle, that kind of ip for instance because a general principle is part of social network. >> guest: it is, although

there's a really -- you wouldn't want it to make a lot of particular kinds of -- you wouldn't want to use that to make important decisions for a very simple reason which is that, let me rephrase that. if you're a business, you might want to use it to make important decisions. if you're an individual, we really don't want that to be the way decisions are made about us because it's, essentially, guilt by association. not in the case of being gay, obviously, but in the case of -- [inaudible] >> guest: you can guess someone's credit rating based on the credit ratings of their friends. now, should banks not offer people loans because their friends have bad credit ratings? that's not, you know, i don't think we want to get into that world. but we are, you know, but actually because there's so much of this and because it's so pick what's done with it, i think we are more in that world than we think we are. >> host: so we had a scandal here in new york about 15, 20 years ago, the red lining

scandal, where banks would famously not give loans to businesses in harlem because that became, essentially, the metric by which they assessed business rather than looking at each individual business. what you're saying is each of us is in a kind of district of our own, and each of us could be red lined for things that might be true of our network, but not true of ourselves. >> guest: that's right. there's this great book that details how geographically there's all this self-sorting. my neighborhood in this brooklyn has people like me, but that's true across the board. um, and, you know, you would kind of hope that at least when you go online, you would be stepping outside of that a little bit. yes, my neighborhood's all people like me, but at least here i'm going to be exposed to the broader world. increasingly i think, you know, we're seeing that same kind of

situation where there's, you know, where the information that we're exposed to gets self-sorted as well based on what people like us would like. >> host: so you say you would hope, and so that's a normative judgment about what should be the case, and yet people are moving in to neighborhoods of other people like them out of choice. no one -- this is, this is not, now, forced segregation. if it's get toization, it's ghettoization by -- >> guest: well, for at least half those people. the interesting thing is, the big story is about affluent people moving to where people like them live and it's half about those who get left behind, sort of receive sorted by default -- self-sorted by default. basically a detroit or, you know, sort of rust belt city where the people who can't afford the self-sort into some neighborhood in chicago or, you know, new york are with people

like them who are like them because they can't move, you know? >> but it sounds like, it sounds like the danger you're focusing on here is somewhat to the individual. as yo say, you don't -- as you say, you don't want people to make judgments based on guilt by association, but it's also about society as a whole, if i understand you. the danger of this kind of sorting is really, um, about what happens to the larger social group. not just the individual. so what is that danger? what will go wrong if we get better and better at relevance? >> guest: well, there's a couple things. so, you know, one is just this very basic thing of being able to empathize with, understand what's going on in someone else's life or what a different perspective might look like. and if you're, if you're in, you know, if you're exposed to lots of different, diverse pieces of information, if you're seeing stories that are not that

relevant to people like you but are relevant to other people, then you, you know, then, you know, it calls your attention to certain problems. the thing that i think about a lot is homelessness, right? and it is very easy to pretend that that problem opportunity exist at all. poverty in america. you know, if you live in certain neighborhoods, you can imagine that nobody's ever homeless. that's not, you know, that's not a very good situation to be in as a society. >> right. >> guest: and, actually, you do experience this. you know, i was at a sort of conference with a lot of fairly well-off, you know, entrepreneurs and venture capital people, and, you know, the idea that there was this huge crisis where lots of people didn't have jobs in america, that wasn't the crisis for them because they don't know anyone who doesn't have jobs. >> host: right. >> guest: and, frankly, i don't

know that many people who are out of work because of this recession. yet we know there are, you know, there are a lot of them. >> host: but it's just a statistic. 1 is.5% of people -- 11.5% of people, but no one i know personally. >> guest: right. and that's what's concerning because, you know, you don't -- you want to have some ability to track at a societal level here are the things that matter overall even if they're not relevant to my own particular thing. and zuckerberg, i think, i'm very fond of this quote from mac zuckerberg where he says the thing about the -- this is what's great, a squirrel dying -- >> host: the facebook new speed. >> guest: is that a squirrel dying in your front yard may be more relevant to your interests right now than people dying in africa. he's saying this saying, well, let's show people the squirrel. i think it sort of points to the problem which is it wouldn't be good if we were just looking at

our front yard. >> host: so you are going back and forth between, a statement like that, it wouldn't be good if people are looking at their front yards, and saying you want to be in a situation where you're more aware of what other people are saying. but, in fact, i think for a lot of us we actually don't want to be in that situation. we're actually preferring having, having relevance given to us. so is -- how do you get around the kind of eat your peas feel to the argument you're making? >> guest: well, because people do eat peas. [laughter] you know, we do, you know, or you have an obesity epidemic. the, i think the challenge here is the question of want. and the critique you're saying in a way is why shouldn't people get what they want? and what i'm saying is we all actually want a lot of different things. we want to, you know, we want to eat cake, and we want to be thin

and healthy, you know? we want a lot. >> watch 30 hours of tv and be fit. >> guest: no. and, you know, we want to be entertained and just, you know, have fun with our media, and we want to be informed citizens and know about the world. i mean, there are few people who are not going to check that on a survey, yes, i want to be an informed citizen, i want to know about what's going on in the world. and so then the question is, you know, given that internal tension how did the media institution, how do these things serve different wants? and i think what's going on here is that because you can easily track the more short-term kind of compulsive wants -- >> host: i want this right now. >> guest: i'm going to click this -- >> host: that's interesting. >> guest: -- you get media that's biased in that direction.

you get kind of compulsive media. i think this is actually what a lot of facebook is. i think it's very, very rare that i or anyone that i've talked to could walk away from the facebook saying, boy, i just had the most revelatory experience on facebook. i learned about the world, it was such a rich -- you know? that's not what happens. it's not really built to do that. it's built to give you kind of these little compulsive tidbits that give you a little flash of dope mine or -- dopamine or, you know, they feel good, but it doesn't actually, you know, it doesn't, it doesn't change the way you look at the world. and, um, you know, it's almost as if next to the like button you want a the a hard slog at first, but then it changed my life button. [laughter] >> host: it's funny, in the writing it doesn't fit well in an icon. in the essay world there is now a, um, an attempt to do that

kind of precommitment. i would now like to say that i'm going to read that later. >> guest: right. >> host: and there's an app called -- an ipad map called news.me that has recently started returning results that say in the past you said you wanted to use this later, so they're actually trying to stretch out your time commitment. >> guest: yeah. that's exactly the thing. and that's the sort of thing where -- i mean, to be clear i'm not nostalgic for, you know, the good old days when there were editors who just knew their stuff -- >> host: i would not accuse you of that. >> guest: well, but, you know, i do think that there are things that those media knew or learned how to do that have not yet made it into a lot of the algorithmic editing that is currently happening. >> host: that's certainly true, although i spend a lot of time

looking at the media environment, and i think it was actually the scarcity that was around them. this a way they could behave in a certain way because they weren't under quite the same competitive pressure. i remember when walter cronkite died, the incredible outpouring of emotion in part because we were burying the man, but in part because we were burying the role. you can't have someone saying that's the way it is every night without fear of contradiction on the internet. but here, i think, comes the big tension around this, this harnessing of media entities for civic good. each of them is just a commercial be actor in a competitive environment. so you tell the story of going on to facebook and of friending people who are much more conservative than you politically. and facebook's not showing you what they were saying or doing, not surfacing their activity in your news feed because facebook observed, the algorithms

observed you with weren't actuay interacting with those people. but isn't it also people that you're actually spying on them and that facebook isn't in the business of showing you what other people are doing if you're not having a kind of real connection? >> guest: well, you know, this gets to this interesting question of, you know, revealed versus expressed preference, right? so this is the thing, should i trust what you say you want, or should i trust what your behavior candidates that you want -- indicates that you want? >> >> [inaudible] >> guest: in this case i said, i want the hear from these people, and facebook said, yes, but you're not behaving as if you did, and this is one of the interesting sort of power struggles, you know, in this kind of media. netflix does the same thing, actually, now where it used to be all about express.

it used to be i'm clicking on five stars and, therefore, netflix will know i like these kinds of movies. what they found was they didn't like -- so now that they have the streaming, now that so many people are streaming movies on netflix, you can say, well, yes, you watched five minutes of that, but then you flipped over to "fast and furious," so that's actually what you prefer. >> host: right. >> guest: and this, in this a way, gets me back to the thing of, well, what is a preference, and should you trust me when i say that i prefer something, or should you trust my behavior? i would argue that you want -- each though people don't behave in the way that they say that they will, there is a lot of value in taking them at their word and giving them that kind of agency that just as a general principle you want people to

actually make decisions about what they do. you don't want to have to behave in a certain way that then -- >> host: well, businesses don't want that. businesses want people to get what they want in the short term. so, again, the tension between the sort of commercial and civic value, i mean, it strikes me when you say you have to trust me when i say something. you're actually pointing to the idea that there's two different mes. there's the me that's going to do stuff in the next phi seconds, and then there's the me that's going to do stuff in the next five years, and if i commit to the next five years, businesses that only serve me five seconds at a time are undermining this longer-term -- >> guest: right. again, back to the junk food because it's a useful analogy, you know, this is i want to be a thin, healthy person. boy, that burger king burger looks awful good. that's sort of the tension. and you can say, well, it's people's choice. ec choose to have -- they can choose to have burger king or

not. but actually, you know, to change that behavior you both need better signaling from, i would argue you need the companies to say here's what, here's how healthy this is, actually, for you so that you don't, so that you have some guide posts on which to know this. and be then, you know, you want to have, actually, people have options. and then you want to encourage people to use those options well, you know? those are three things you might do to confront that problem. you know, here you don't have any sense of what the other options are necessarily. you don't necessarily know what else is on the menu. and, you know, part of what this is about is, i think, sort of trying to make this actually more pass i of a medium in a way -- passive of a medium in a way. >> this, what do you mean? >> host: making online a more

passive me? >> guest: yeah. essentially what a lot of us -- what i certainly love about the internet is that it is an interactive thing. steve jobs you have to be leaning forward. >> host: right. >> guest: and at the same time that's not actually -- if what you're trying to do is deliver ads or deliver, you know, media that has ads in it or whatever, it's not necessarily what you want because people are going to click over to some other site. so how do you provide a stream of things that's so, that keeps you there, you know? is that keeps you, you know, watching youtube instead of clicking over to facebook or whatever. >> host: uh-huh. >> guest: and that, i think, when you talk about a competitive environment, i think sort of what people have realized -- >> host: you have to rate the activity. >> guest: yeah. how do you do enough of the leg work that people stick with you, you know? finding the next thing. >> host: so we've grown up, everybody in america's grown up in an environment where an

advertiser can get so much information about us coded to our zip code. famously, if you tell me your zip code, i can tell you roughly what you drink and what toothpaste you use. we, nevertheless, lived this that world and were relatively satisfied with it. we may even have preferred it to a world in which advertisers knew nothing about us because of the misfit. how is that different from that, grouping people into category by zip code, how is this different than that, and is this a bigger threat than that? and if so, how much bigger? >> guest: well, so one way in which it's -- i mean, obviously, it's different in that the grouping can be much more disparate. so if i'm an outlier in my zip code, you have the geographical self-sorting and sort of the

personally self-sorting happening, slipping apart. they're not happening in the same -- >> host: right. >> guest: -- and that has some important consequences because there are some values into being, actually, plugged into whatever's going on in your locality. the broader, you know, the broader way that it's different -- so one important way that it's different is it's, you know, not a shared experience with the other people right around, you know, right around you. so if you turn on your local tv, your local tv news, you know, then you're all kind of in the same stories, in the same narratives for better or worse. >> host: right. >> guest: this is, you know, you can live in a very different media universe than the person right next door. >> host: yep. >> guest: but, you know, the bigger problem is, you know, again that you aren't actually,

you don't know -- you can't see the contours of what you're seeing. you know, you don't know that it's local news even though it is local news in a way. >> host: say more about that? >> guest: well, because, you know, local news is news for people -- >> host: who live near you. >> guest: -- that live near you. we're increasingly getting local news for people like us -- >> host: oh, i see you. >> guest: in a different sense, it doesn't say that it's that, you know? it doesn't reveal that it's that. >> host: when i do a google search, it feels like i'm searching google, not that i'm live anything a locality of one. >> guest: right. and, you know, in the same way on google news, you know, you don't, you don't get a sense of how, you know, what is the standard -- how many standard deviations am i off from, you know, from other people in my area, from the average.

you don't get a sense of how different your experience is. and i think that's important because, you know, the process of figuring out what is going on, what is true or real is this process of triangulation a lot of the time where you're comparing, you know, there's a study about liberals listening to rush limbaugh. and it doesn't make them change their mind. you know, it doesn't -- they don't all of a sudden realize that, you know, rush was right. they do start to realize that a lot of the beliefs they held strongly aren't supported by very much information or arguments that they have. that he says something, and they go, i think that's wrong, but i don't know -- >> host: they can't articulate why. >> guest: and it calls attention to the limits of your understanding of the topic, actually.

and i think that's actually a really valuable thing. i mean, i think if you want to know what's happening in the world, if you want to have your attention called to where are the places that i know i know this and where are the places that i think i know it, but i may not, the known/unknown perimeter. >> host: right. i don't know that people think about -- >> guest: right. >> host: so you used an interesting phrase earlier. you called the search for relevance a code word. is this what it's a code word for? when you say it's a code word, what is being -- >> guest: i guess, again, you know, it gets to the -- i mean, i think there's a simplistic idea of what relevance is. so i got into this argument with google, and i said, okay, you say that you want to provide relevance, but if i'm a 9/11 conspiracy theorist and i google

9/11, is the most relevant result the most result i'm -- the result i'm most likely to click, or is it the popular mechanics article that debunks that stuff that, you know -- >> host: is relevant to your beliefs but not in a way that you like? is. >> guest: is relevant in this a way that is uncomfortable. and being a pr guy he was sort of like, oh, you know, i'll get back to you. [laughter] it wasn't, it didn't, they didn't have an answer. but i actually think that is a hard question. >> host: have you posed that question to the search engineers? >> guest: i haven't -- well, this was a -- i did have the conversation with a search engineer, and they said, well, we try to bring some diversity into the mix, but i actually think it's not what -- i think when you hear eric schmidt say what we want google to do is

provide you with one search result, the right one, it doesn't really work in that -- >> host: right. >> guest: it just doesn't make any sense in a way for certain kinds of queries and the way people are using this. there is no right one for 9/11. >> host: right, right. good. well, let's take a break here, and we're going to come back and talk about maybe what we can do with all of this. >> we asked, what are you reading this summer? here's what you had to say.

>> send us a tweet at booktv using hash tag summer reading to let us know what you plan on reading this summer. you can also e-mail us, booktv at c-span.org. >> host: in the stories you tell about google, about facebook, about this filter bubble, um, you talk about the kind of political limitations of, you know, what it is we can see about the world. and for most, for, essentially, a lot of people who think about the internet, either think about it for a living or just are interested in the what's happening, that thesis -- anything having to do with the thesis of the internet's effect on political conversation was first introduced with cass sunstein's book, republican.com.

republic.com. introduced the idea of echo chambers and self-selecting enclaves on the interto, you know, just as a general concern. what's -- how would you position your thesis relative to sunstein's thesis from republic.com? what's the same, what's different? >> guest: well, sunstein, you know, especially in the first round that people are most familiar with, with talking about the sort of explicit self-selection i'm a neo-nazi who's going to the neo-nazi news online, and, you know, i'm getting indoctrinated, or i'm amping myself up, and i can exclude the whole rest of the world. i think the number, you know, the first thing that's different about this is that it's, it is this passive experience. it's not that people have to seek it out or know they're making decisions that shape what

they see, it's just encreesingly -- increasingly visiting web sites that -- you don't see the internet at work, and you're certainly not making that decision. i think the decision is important because i think, you know, understanding what the editorial sensibility that you're dealing with is important. i mean, i think if you're trying to see what the world is knowing ah, yes, i'm looking at this through fox news, probably not going to have a lot of good stories about obama. i should, i should watch that. you know, i should make sure that i'm not, you know, or i'm looking at this on msnbc and, you know, the opposite. so that, that part about selecting, choosing the filters through which you see the world, i think, you know, he had that as sort of, okay, people are going to go out and do this. i'm saying you don't even get to, you don't even get there. you don't even get to --

>> host: you didn't get to sunstein's moment of explicit choice because it's just -- >> guest: it just comes to you. certainly when you hear people talk about the future of tv, you know, it is very much this, you know, the future of tv is you've got show after show that is a show that someone like you will like. i also think, though, that because of the nature, the way that these algorithms are constructed, this is where it departs more from -- [inaudible] world in a way. there are things that fall out sort of as unintended consequences of doing that kind of sorting. so, again, to go with the future of tv model -- >> host: uh-huh. >> guest: -- you know, presumably the news come on very rarely for a lot of people in that future tv model. >> host: right. >> guest: or certainly a half hour block of news, that's just not built in -- >> host: cnn has already gone to

much more of celebrity news and much less on its focus on international. >> guest: right. and there's a political scientist, marcus pryor, who i think does some really interesting work on this who tracked, you know, sort of old three-channel tv world versus cable world versus internet world. what does this do to the shape of how people -- of political knowledge? and he, basically, says that there's good reason to believe that in that old world where there were three channels and the news was on at 6, and if you wanted to watch tv at 6, you had to watch the news. >> host: cronkite or brinkley. >> guest: you know, you had a belt curve. most people grouped in the middle in terms of what think knew about politics -- what they knew about politics. and as you increase the amount of choice, you have news junkies that know amazingly much. >> host: we're in the heaven right now. yeah, straight into the vein.

>> guest: but the average, you know, moves downward actually on the curve because on balance people don't necessarily prefer, you know, should i watch the news now, or should i watch another episode of, you know -- >> host: desperate housewives or whatever. >> guest: yeah, yeah, yeah. so this is, you know, again, it's this challenge of, um, if you're going to look at the world through these algorithms that are trying to figure out an idea of relevance or an idea of what you want, how do you balance those conflicting impulses? the how do you get it to do what i would argue the media of the 20th century have done which is, you know, yes, the new yorker has these long articles, but it also has the cartoons and the little bits and pieces, or -- and most magazines are like that. that's part of the form, it's entertaining enough that you'll read it, and you'll get to the -- >> host: you've never seen tiger beat. not all magazines are like that.

[laughter] >> guest: that's true. but, i mean, "rolling stone" is an example. a lot of pop nonsense and good music reporting and also matt taibi who's doing this expose on the political world. that's actually a nice format for having people know about the world. as you get the bundling and all of the different pieces available without having to see what other pieces they're grouped with, you lose some of that, and it requires these algorithms to be much better about building some of that back in the. >> host: so you are simultaneously saying that the large-sized chunks of media whether it was we only had three networks, and we had the news for an hour every night, and we had magazines that had to have a little bit for everybody, for whatever version of everybody they wanted to reach in order to get everybody to be willing to

support the subscription, everybody had to subsidize everybody's stuff a little bit. >> guest: yeah. >> host: i'm struck by how much all of that stuff is in the engineering terms absolutely inefficient. >> guest: right. >> host: and we now have a world that works unbelievably much better, and yet we lose the kind of subsidy of the -- >> guest: but the question is, what -- how do you define efficiency, right? >> host: well, i'm dining it specifically in -- defining it specifically in engineering terms. >> guest: but if you say a five-star movie on netflix is worth way more than a four because, actually, what you want is not only for people to have a good time on netflix, but you want people to have movies that stay with them for 20 years or whatever. >> host: well, but so here's the question. engineering almost never reverses. we never get to the point where we say, oh, we missed the horses. we're going to roll the internal

combustion engine back. so we've got this explosively efficient network now that is casually unbundling and casually destroying the engineering efficiencies that made the media work in the 20th century the way it did. and you are not a notal gist. so we can't use the old methods to get the old values. but what it sounds like you're saying is the old values of exposure to lots of things, exposure to the ideas and concerns of lots of people, cross-subsidies of one another's interests are still important. so starting from be here and moving forward rather than backwards -- >> guest: yeah. >> host: -- what do we do? >> guest: well, so i think there are two -- in ways i'm pointing to two related but pretty different problems. one is personalization isn't very good right now. it's not very -- >> host: but if it gets better,

the problem gets worse if i understand your -- >> guest: well, no, because it doesn't understand -- i mean, again, it doesn't have a very good view of the nuances of how we, what we want and how we -- >> host: but we're getting five second personalization but not five year personalization. >> guest: yeah. and we're not getting -- it's still very much, oh, your interested in if -- you're interested in, you know, computers, you're probably also interested in the software. t not drawing some of the larger inferences you can make because you can actually do very well just sticking in these little narrow territories. so you could use it to introduce people to more than they're introduced to -- >> host: you just get the tight little -- >> guest: yeah, yeah. the second problem is, yes, there are some places where you actually want to introduce things that are pulling in some other directions that are not

going to be, at least in the short term, the most efficient solution. so you want to say -- because you want to -- and i suppose the market argument for that would be that that is actually what a satisfying media experience, you know, that discovery and i didn't think i would be interested in this, but it turned out that i was -- >> host: those are the moments you really remember. >> guest: yes. and i think this is why people stick with, you know, i mean, if you look at the media environment, it isn't all -- the old media environment wasn't as optimized as it could have been. [laughter] you know. and magazines, you know, or newspapers didn't just put what sold, there were other things there. and part of it is because it reflects back to you an idea that you are, you know, learning about things, that you're getting either the feeling of getting to know about the world and broadening your horizons, and people like that feeling.

the algorithm ims that we have right now are pretty crude at providing that feeling, and i would argue it's going to a while before we can fully -- knowing that we can get more people to read this piece if it directly follows this piece about something else because they have some kind of intangible, you know, symmetry or they're somehow working together. that is some of the art that needs to be built back into, you know, this stuff. so i'm not saying -- so engineering doesn't go back, but engineering can it learn from, you know, it's like, yes, we have cars now, we're not going back to horses. there are things you can learn from nature about how to learn to build better cars. so how to we learn from some of the finer hoints of ours, how do we take that and figure out can you code this in? then the other piece is how do

you provide people of better ways of -- how do you make this a more multidimensional, you know, kind of sorting system? so to go to facebook like for a minute. like is such a one -- such a -- there are particular things that people are easy to like, there are particularly things that are harder to like. the things that are easy to like, move across facebook more rapidly, things that are harder to like, there's no inherent reason to say there's also an important buttton. and in a way this is, you know, how do we -- we've learned so much on the tagging kind of world about all the different ways you can sort money. but it hasn't applied to -- >> host: to the social media, rye onically. >> guest: i mean, this comes up

on facebook in the terms of facebook doesn't know who actually i'm friends with, who is my family, you know? the it would presumably be a better service if you could figure out a way to do that that was inon truce i. inobtrusive. partly, i think, there are -- you can get a big with gain by just sort of doing this first step and getting to, ah, you're male. we're going to show you more stuff about gadgets. but to get to the place where it's actually good, you know, it may not deliver as much profit and it's harder. >> host: doesn't could -- sound like a winning idea for business. >> guest: they love making people spend more minutes on facebook. and what you're asking us to do with these complicated social questions, you know?

the this is all kind of like, you know, we just want to do the fun and simple stuff where we figure out a clever, you know, they be on the book more. >> guest: you know, and then, too, that these companies start to provide these tools and find that people are taking them up because they have that desire. >> host: so that's, that's the supply side. and i'm struck when you're talking saying we should do this, and we could do that with algorithms and studenting and so forth. -- tuning and so forth. the we is that class of engineers who work on gathering these signals and making predictive models. it's amazon's, people who bought

xl. movie considerations taken to some kind of civic level, but most of us are not sill cob valley ec nears, so -- silicon valley engineers. so let's say i am a citizen who thinks i'm never going to be a news junkie. i'm not going to be someone who nervously consumes anything on any subject they can get their hands on. but i do want to use the network to get more -- a broader sense of the world than i'm currently getting. and i can't go in this and change the google search algorithm. what can an individual do to say i want a new immediate qualm dock. >> chattanooga. >> guest: hasn't done this very much. so, you know, i lost my

conservative friends on facebook. i never see them anymore. i see them on twitter, and that's because twitter has a very simple, transparent, you know, rule which is you follow them. if you're logging on to twitter now and they're tweeting now, you'll see their stuff. >> host: what happened yesterday? >> guest: what happened yesterday is that twitter's actually starting to succumb to the same motivation for the same reasons. so when you search for something on twitter now, you get the most vel rant tweet for you, not -- >> host: not the most recent. >> guest: not the most recent. so this explains a conversation i had with people at twitter where i was with saying it's so great you guys are holding the line on this. and they said, yeah, it's really hard. [laughter] but at the moment anyway, this is a tool that does yeast how you to say, you know, i follow

karl rove and rachel maddow and together i'm going to kind of get a sense of what's going on. >> host: so a twitter stream, if i follow someone, twitter is not affecting -- >> yes. twitter, i mean, the phenomenon of i'm googling climate change, and maybe i'm going to get results that are for people like me -- >> host: uh-huh. >> guest: -- can now move to twitter, i'm googling client change and i'm getting tweets. >> host: so i think there are no climate change deniers or skeptics. >> guest: exactly. you know, again, it's, i think, it's useful to all of us to see that. you know, to be reminded that -- and to actually understand those arguments. >> host: could we have an i really mane it button? it sounds like your suggestion

to the, you know, we're backing out of th algorithm designers. i should actually be allowed to override my own personality if i want to. [inaudible conversations] >> guest: it's even, you know, it is more let me turn on or off your behavioral reading of me. you know, sometimes i may want to just come home and vej out in front of the tv. you know, that's fine, but i want to be able to opt out of that sometimes, and i want to have it be kind of an explicit thing that you're doing. and i think part of the challenge right now is that i think a reason that this is not transparent is that it creeps a lot of people out when it's done explicit ily.

you know f if you track the history of this, it start explicitly, and it started with, you know, microsoft bob and these little personal avatars and -- clippy. and it's annoying and creepy, very bad. i mean, it just with doesn't buck that well when you think of a person that knows you because he, obviously, doesn't know you very well. if you embed it under the surface of things and you're just showing them stuff. they then will click on these lins. i'm not sure if you had a tab at the top of google that said personalize google, regular google, i think a lot of people would choose the -- >> host: be they knew that it

was -- right. if they knew that goggle's bending things in that direction. yeah. i'm struck by, i moon, to the story of bob -- i mean, to the story of bob, even sort of pre-bob there was patty mace's lab up at mit that are worked on participation, they called it collaborative filtering. and it was a service called ringo. and ringo was exactly an explicit you say what you like, and people hated hearing that when they made explicit choices, they actually hearing that their musical tastes were predictable. >> guest: right. yeah. >> host: so so if you like metallica, you may also like slayer. when the people bought xl, it

didn't because no one was with ever asked to explicitly opt in to the system. no one had the cog cognitive dissonance. >> no, that's right. that people are willing on the shown things that they might like, but they don't like being told that, you know, you're going to like -- >> host: right. you're saying at least tell people, right, we're making sausage for you, and this is how the sausage is made. >> guest: right. the other argument i get in again and men with these -- again and again with these engineers, they say, well, people respect really demanding these services that you want. i say, well, how can you have consumer pressure about a product? [laughter] >> host: interesting. >> guest: you can't get e-mail on that. it doesn't exist. >> host: it would be like the cigarette companies saying the additives we're putting into the tobacco, no one's complaining

about them. >> guest: right. yeah. >> host: so let me ask you this again in a vearl explicit way. i am a wine-swilling, coast-loving, liberally-minded individual. i want to see a different slice of my own country, of the world and i'm using facebook and google, so i want to break out of this bubble you've convinced me. i go home, i open up a browser, and then i what? >> guest: well, i would say right now the best, simplest thing is to, you know, go to twitter and start and put together a twitter stream that is, you know, that includes sources that, you know, that get you out of this zone of debt. and the nice thing about twitter is that it does the thing. so one of the things that a

newspaper first does is, and you've pointed this out, you know, you're not going to read all those stories, but these you know they exist. and you can skip over them, and it changes from the known-unknown to the unknown-unknown. when i follow fox news or glenn beck, all of which i do follow, it opportunity mean i read all those stories, but then i start to fete a chens of, okay, this is what fox news was talking about. and i also follow bloggers in the middle of the country. it's actually a really fun cherdz to i saw, okay, what are the different places that i need to go in order to get a pretty good view of what's going on.

>> guest: in philosophical terms you see -- you believe what you see about the world, and we all do that to some extent, you know? as best we try to say i know things are being edited out, you know, i'm going to pay attention to that, it is, you know, we're all susceptible to that. but you can, if you're choosing and you want to get a good view of the world, i think you can do a lot better than facebook world or google world or google news world in doing that. you can get a much better view. >> host: so, and saying, you know, you believe what you see, but you're saying that the awful feedback increasingly through search and personalization, you also see what you believe. >> guest: right. >> host: they can infer your beliefs, the range of what you see contracts, so your beliefs contract and what you see contracts. so i go home, and i find people on twitter who have clashed with mine and are in the areas i care

about. but twitter is, twitter is not yet a news feed although it has some of those characteristics. and i think about, you know, one of the sources of permization that you've raised -- personalization that you've raised the alarm about is google news and yao -- yahoo! news. if i want to change my news docket, what do i do? >> guest: well, i don't know. i think twitter is a proxy for news in a lot of ways. you know, most news agencies are on -- >> host: so i could do 'em with -- >> guest: you know, i think, um, and actually, i think that google news has wrestled with some of these problems more than a lot of other parts of google. i mean, google news has thought this a bit, people think of google news as kind of this