http://www.niemanlab.org/2012/06/theres-no-such-thing-as-an-objective-filter/
The Nieman Journalism lab highlights in this article the challenges in trying to create objective filtering algorithms for the news. The struggle between "we have to make it" technologists and "human needs are not taxonomy" humanists continue.
But like many wars, once you start to speak the same language, once you have that shared platform that enables you to see yourself in "the other", the bloodshed begins to die down.
In regards to the news, the common ground for technologists and humanists is expanding, and is called "design." After Steve Jobs, the term design means something very different than it used to. There's a strong element of empathy in what we currently call design, an empathy for the user, the individual, and his - well, to be completely redundant - individualism.
To the cult of Steve, this is pretty ho-hum stuff. We forget though that for a very long time, and as a result of an industrial age that thought of organizations as machines, we loved to group people. It was our collective bad habit that grew into addiction because of its usefulness. It was like Windows 98 in 2004, a useful but ancient artifact that became inherently limiting. We've learned a bit more since then but we still love to group people. It's easy, locally efficient, and scalable. Religion, nation, education, politics, state, age groups, generations, digital immigrant status - it goes on and on.
Our desire to algorithmically produce shared "content results" derived through a common filter stems from that old mindset, one where standardization was cheaper. But the marginal cost of data is zero and personalization is better. The Nieman Lab gets to the heart of the matter - what is relevant for you is fundamentally different than what is relevant for me. It's also true that what is relevant for me today could be very different from what is relevant for me in a month - though we should recognize that I generally change much less in a month than the average difference between me and the next person.
I think about this all the time for my day job - this idea of the long tail of information needs. Every set of metrics I see about information consumption in a global firm of 180,000 points to that long tail of needs. And frankly, while my firm is large and diverse, there is an inevitable and specialized kind of homogeneity based on the selection process to even be hired. Yet, despite that similarity, still there's an extraordinary divergence in information needs. The top keyword search on our intranet in a given month - "learning program" for the curious - represents only 3% of the total, and even that percentage drops off steeply as you go down the list of keywords.
How can we serve that diversity of information needs, both densely and efficiently? We talk until I want to puke about personalization, but what does that really mean and where have we seen systems and institutions treat people like the whole persons that "personalization" implies. Historically, the rigor of programming logic, taxonomy, and ultimately actually having to put fingers to keyboard and make the damn thing has forced us into a pseudo-personalization as the best alternative available. Whether it’s favorites, bookmarking, opt-in newsletters, or social, we call the result personalization because each person at least sees something different, and because it is admittedly a notch further along in the general right direction. There’s usually a better before you get to the best.
But how about the whole person me -
So – what information do you have for me? What filtering algorithms will you create? Will they pick out my firm's acquisition of a digital agency, the opening of a new Charles Phan restaurant around the corner from my house, a new talk by David Eagleman, a beautiful letter from Cheryl Strayed as Dear Sugar, the engagement of my friends Geraldine & Felix in London, a new Skillshare class on mapping spatial data in SF, my friend Wes' desert race? How do you take the gold pebbles I now have to tediously mine throughout the day, and hand me a pure bullion bar every morning? My sister, who's probably the person most like me in the world, couldn't do it. It's like your best friends setting you up on a blind date - how often do they go well. We individuals are so specific, there’s no way someone else could tell gold from dross, consistently.
Those kinds of filtering algorithms can’t be created; they have to grow. They have to grow in the same way we grow, from a seed pattern, with variable expression, adaptable, evolving – human beings are the ultimate example of mass complex individualization, and we are grown. There’s no other way to do it. In 246 BC, the first Emperor of China, Qin Shi Huang, created 8000 unique terra cotta soldiers, each modeled after a real soldier, to protect him in the afterlife – but it required 700,000 workers and decades to complete. And our outsides are far less complex and easier to recreate than our insides.
To take the metaphor to its stretching point, growing a young filtering algorithm requires the right fertilizer – the kind of rich, nuanced, sometimes paradoxical information about you that no single person has all of, not even yourself. You need a heck of a lot of information to mine to even begin to codify the depths of someone’s values, dreams, loves, life.
The technologist will say, “This is still crazy talk.. we eventually have to put fingers to keyboard. The perfect filtering algorithm is a false Shangri-La.” But let’s for a moment create a reality distortion field, of the variety that Steve Jobs liked to construct. How would we, not necessarily aggregate, but rather connect, all that information about an individual? Where would we start? Who would we trust with it? Maybe no one. Perhaps only ourselves. Or – maybe everybody, under the right circumstances and enabled by some system of clever mechanisms and protocols. No one, ourselves, everybody, how about all of the above.
Or maybe it’s impossible. Maybe. But how many times has someone said that, only to be proven wrong? On first glance, we seem impossible, the complexity of our brain structure, the intricacies of our social interactions, the out-of-nowhereness of our art, the permutations of us. But we exist, and we grew and evolved. We are not perfect but we tend to get better. Call the comparison (of growing filtering algorithms to evolution) macro-biomimicry or conceptual skeuomorph or just a mere instance of AI machine learning – whatever the case, it seems hard to deny that evolution has time on its side.
Is it just a matter of enough cycles? I don’t know. I do know that the scale of the conversation today between the Two Cultures – sciences and humanities, technologists and humanists – is unprecedented. Edge.org, TED, Apple products, neuroscience, game mechanics, and on and on. I can’t think of a better place to focus their efforts than the design of filtering systems that grow a thumbprint algorithm for each of us.
Why does personalization matter? It sometimes sounds like glazed-over jargon. Here is why it matters: Because time continues to pass whatever we will, and so life is short and information infinite. If I value that time, what can be more important than filtering systems that delight me with their relevance, let me quickly consume the most important information for me - drawing the cutoff where I decide - and free me up to spend the rest of my sunny day on people and other things I love.
From the children's book Cheaper by the Dozen, after father of twelve Frank Gilbert had died of a heart attack:
"Someone once asked Dad: “But what do you want to save time for? What are you going to do with it?”
“For work, if you love that best,” said Dad. “For education, for beauty, for art, for pleasure.” He looked over the top of his pince-nez, “For mumblety-peg if that's where your heart lies."
The Nieman Journalism lab highlights in this article the challenges in trying to create objective filtering algorithms for the news. The struggle between "we have to make it" technologists and "human needs are not taxonomy" humanists continue.
But like many wars, once you start to speak the same language, once you have that shared platform that enables you to see yourself in "the other", the bloodshed begins to die down.
In regards to the news, the common ground for technologists and humanists is expanding, and is called "design." After Steve Jobs, the term design means something very different than it used to. There's a strong element of empathy in what we currently call design, an empathy for the user, the individual, and his - well, to be completely redundant - individualism.
To the cult of Steve, this is pretty ho-hum stuff. We forget though that for a very long time, and as a result of an industrial age that thought of organizations as machines, we loved to group people. It was our collective bad habit that grew into addiction because of its usefulness. It was like Windows 98 in 2004, a useful but ancient artifact that became inherently limiting. We've learned a bit more since then but we still love to group people. It's easy, locally efficient, and scalable. Religion, nation, education, politics, state, age groups, generations, digital immigrant status - it goes on and on.
Our desire to algorithmically produce shared "content results" derived through a common filter stems from that old mindset, one where standardization was cheaper. But the marginal cost of data is zero and personalization is better. The Nieman Lab gets to the heart of the matter - what is relevant for you is fundamentally different than what is relevant for me. It's also true that what is relevant for me today could be very different from what is relevant for me in a month - though we should recognize that I generally change much less in a month than the average difference between me and the next person.
I think about this all the time for my day job - this idea of the long tail of information needs. Every set of metrics I see about information consumption in a global firm of 180,000 points to that long tail of needs. And frankly, while my firm is large and diverse, there is an inevitable and specialized kind of homogeneity based on the selection process to even be hired. Yet, despite that similarity, still there's an extraordinary divergence in information needs. The top keyword search on our intranet in a given month - "learning program" for the curious - represents only 3% of the total, and even that percentage drops off steeply as you go down the list of keywords.
How can we serve that diversity of information needs, both densely and efficiently? We talk until I want to puke about personalization, but what does that really mean and where have we seen systems and institutions treat people like the whole persons that "personalization" implies. Historically, the rigor of programming logic, taxonomy, and ultimately actually having to put fingers to keyboard and make the damn thing has forced us into a pseudo-personalization as the best alternative available. Whether it’s favorites, bookmarking, opt-in newsletters, or social, we call the result personalization because each person at least sees something different, and because it is admittedly a notch further along in the general right direction. There’s usually a better before you get to the best.
But how about the whole person me -
- the one that seeks to live a spartan life but is constantly surrounded by luggage in disarray
- is terrible at spatial orientation but good at maps (except 3-D maps, where I get just as lost as in real life)
- mostly agrees with libertarians but believes in universal healthcare
- is totally okay being mainstream
- thinks Obama has done a pretty decent job over the past 4 years, but also that Bush wasn't as bad as everyone on the coasts thinks
- drives a 12-year-old Honda Civic but spends its Kelly Blue Book value thrice over in Virgin America flights every year
- likes soup, coffee, vodka tonics and whiskey, in that order
- never gets angry except when she blows up
- lacks every domestic capability (except cooking) but just hired a yuppie home color consultant
- loves her parents more than anything but never sees them enough
- would probably call herself a possibilian if you have to subscribe to one religion, which I'm fundamentally against because I'm not really a joiner
- seems to be the only person in SF with a deep abiding passion for LA, combined with a darkly skeptical perspective on hipsters
- has never done a cartwheel but secretly aspires to complete one once in her life
- and on and on......
So – what information do you have for me? What filtering algorithms will you create? Will they pick out my firm's acquisition of a digital agency, the opening of a new Charles Phan restaurant around the corner from my house, a new talk by David Eagleman, a beautiful letter from Cheryl Strayed as Dear Sugar, the engagement of my friends Geraldine & Felix in London, a new Skillshare class on mapping spatial data in SF, my friend Wes' desert race? How do you take the gold pebbles I now have to tediously mine throughout the day, and hand me a pure bullion bar every morning? My sister, who's probably the person most like me in the world, couldn't do it. It's like your best friends setting you up on a blind date - how often do they go well. We individuals are so specific, there’s no way someone else could tell gold from dross, consistently.
Those kinds of filtering algorithms can’t be created; they have to grow. They have to grow in the same way we grow, from a seed pattern, with variable expression, adaptable, evolving – human beings are the ultimate example of mass complex individualization, and we are grown. There’s no other way to do it. In 246 BC, the first Emperor of China, Qin Shi Huang, created 8000 unique terra cotta soldiers, each modeled after a real soldier, to protect him in the afterlife – but it required 700,000 workers and decades to complete. And our outsides are far less complex and easier to recreate than our insides.
To take the metaphor to its stretching point, growing a young filtering algorithm requires the right fertilizer – the kind of rich, nuanced, sometimes paradoxical information about you that no single person has all of, not even yourself. You need a heck of a lot of information to mine to even begin to codify the depths of someone’s values, dreams, loves, life.
The technologist will say, “This is still crazy talk.. we eventually have to put fingers to keyboard. The perfect filtering algorithm is a false Shangri-La.” But let’s for a moment create a reality distortion field, of the variety that Steve Jobs liked to construct. How would we, not necessarily aggregate, but rather connect, all that information about an individual? Where would we start? Who would we trust with it? Maybe no one. Perhaps only ourselves. Or – maybe everybody, under the right circumstances and enabled by some system of clever mechanisms and protocols. No one, ourselves, everybody, how about all of the above.
Or maybe it’s impossible. Maybe. But how many times has someone said that, only to be proven wrong? On first glance, we seem impossible, the complexity of our brain structure, the intricacies of our social interactions, the out-of-nowhereness of our art, the permutations of us. But we exist, and we grew and evolved. We are not perfect but we tend to get better. Call the comparison (of growing filtering algorithms to evolution) macro-biomimicry or conceptual skeuomorph or just a mere instance of AI machine learning – whatever the case, it seems hard to deny that evolution has time on its side.
Is it just a matter of enough cycles? I don’t know. I do know that the scale of the conversation today between the Two Cultures – sciences and humanities, technologists and humanists – is unprecedented. Edge.org, TED, Apple products, neuroscience, game mechanics, and on and on. I can’t think of a better place to focus their efforts than the design of filtering systems that grow a thumbprint algorithm for each of us.
Why does personalization matter? It sometimes sounds like glazed-over jargon. Here is why it matters: Because time continues to pass whatever we will, and so life is short and information infinite. If I value that time, what can be more important than filtering systems that delight me with their relevance, let me quickly consume the most important information for me - drawing the cutoff where I decide - and free me up to spend the rest of my sunny day on people and other things I love.
From the children's book Cheaper by the Dozen, after father of twelve Frank Gilbert had died of a heart attack:
"Someone once asked Dad: “But what do you want to save time for? What are you going to do with it?”
“For work, if you love that best,” said Dad. “For education, for beauty, for art, for pleasure.” He looked over the top of his pince-nez, “For mumblety-peg if that's where your heart lies."
RSS Feed