Mar 282007

A few posts ago, I wrote, among other things, that Eric Goldman’s concept of Cosean filters as an interface between people and marketers is  impossible. A commenter, Sean Ammirati, asks whether I would approve of adopting these filters if they were possible. The question seems strange, rather like asking whether I would endorse time machines as a mode of changing past history if they were possible. Strange as such questions are, the short answer in each cases, is in fact, “no.”

Below I will (a) briefly explain why Cosean filters if they worked would hardly be desirable, and then (b) I shall explain a little more fully why they will not work, even in principle.

(a)    Why we should not want Cosean filters

Goldman’s idea is that a suitable filter, always carried with you, could shield you from unwanted marketing while letting in all proposals that would add to your personal “utility function” at each moment. Given the existence of such devices, marketers should supposedly feel free to deliver a huge volume of advertising to everyone, to whatever extent bandwidth permits, knowing that no one would be bothered by items they would not want.

There is no reason to think such pitches would have to be limited to items or services for sale. Anyone seeking attention for whatever reason — including just seeking attention— might also take advantage of these filters to reach the broad possible audience, without having to worry about alienating anyone.

The filter would presumably have to guess your internal states correctly. No one would enjoy being bombarded with descriptions of big meals right after having eaten one, but if you happen to be hungry, descriptions of food selected to be exactly to your taste would be very hard to reject, even foods not good for you, too expensive for your budget, or too fattening. And food is of course only one category. You would be inundated at different times by exactly whatever it would be that you would be most vulnerable to at the moment. Your entire bank account and your entire set of thought processes would be dominated by such a filter. So might your actions. Someone makes you angry; your filter lets ads for guns, knives, and clubs through, or perhaps simply suggests ways to use your bare hands to kill or wound  someone who offends you, or suggests how ot insult them grossly. Along with these, assuming you are angry at the government or someone of another ethnicity, might come fascist or racist propaganda. You feel lonely; instantly all sorts of offers prey on your loneliness. Something unsettling happens; the local bar’s drinks that are especially to your liking  and exactly what you love even be ordered for immediate delivery. If you are depressed, also sorts of anti-depressants would be on offer, along perhaps with uplifting religious sermons. Should you be really depressed, a guide to committing suicide with handy sources for the supplies would appear.

I hope all this is enough to suggest that a true Goldmanian Cosean filter would be nightmarish in practice, just as would the possibility that anyone could travel backward in time and monkey with the past. With the filter, everyone would be hard put to resist every whim, every emotional current; we would all be in effect drugged out, practically all the time.

(b)    Why such filters are impossible anyway

Fortunately, as I have said, such a filter is just not doable. It requires a machine that knows you, actually much better than you consciously know yourself. The world, and certainly marketers, are always developing new items, new wrinkles, new combinations; for the filter to “know” whether they will greatly interest you, more than some knowledge of you past shopping behavior would be needed. (In fact, if all the filter would do would be to know your favorite products, stars and flavors, it would hardly be needed; if you were only to buy or pay attention to the same things as ever, much simpler marketing methods would work better than the filtering idea.) In reality, you confront the world as a constantly changing human being, always able to alter the angle at which you view or decide about the world, always encouraged or forced to do so by the way the culture changes, your experiences change, and your insights about the world develop.

Even other people who are very close to you have a hard time knowing what you really want. Consider those closest to you — your spouse or lover, your best friends, your siblings, parents, children, colleagues, students or teachers. Do any of these people always give you birthday presents you want, propose trips that invariably interest you or make dinners you like? Do they unfailingly come up with conversational topics that interest you? I have often witnessed couples who seem very close and very attached to each other getting these things wrong. Can we imagine that a computer controlled by suitable algorithms and fed details of your past behavior will do better?

In fact the reason a Cosean filter is impossible is the same reason no computer by itself is ever likely to be able to pay you what feels like real attention for any extended period. What makes it possible for a person to pay you detailed attention is that that person can imagine having feelings and experiences similar to yours. Our experiences are not just rational, or even just emotional but also bodily. You know what it is to walk on pavement as opposed to carpet; you know what a splash of cold water feels like on your face, what it feels like to run after a bus or eat a hamburger, or some equivalent. As a human being you also know what it feels like to want attention, because at times you do want it. Anyone paying you attention must also be able to grasp such things, not abstractly, but in complex, always mutating detail. Can a computer pay you attention if it only simulates having such feelings without really knowing what they feel like? I see no reason to believe such simulations would be generally convincing or acceptable. At best, a simulation works only from a certain perspective, just as a  photograph can at best simulate an actual scene only from a certain angel. Humans are multi-dimensional, and, to repeat, in constant flux, always tapping their store of experiences and feelings in new ways. The only possibility of being able to pay attention to someone else through that is to be able to conjure up similar perspectives and feelings. Only a computer that actually wants attention itself — and incidentally has  a human-like body — is likely to be able to pay attention to you in ways that feel convincing, because wanting attention is large part of who you are.

But what would be the point of a computer that insisted on attention to itself? HAL in the movie 2001:A Space Odyssey is not a success as a useful computer. In the same way a Cosean filter that knows what you want would be a failure as a tool. To be a success at focusing it, you would have to be capable of enthralling it, having it fall utterly in love with you, willing to do everything for you. But if you have that kind of power over a computer, you could just as well have it over an actual human. Most of us just are not stars, lack star power, lack the ability to enthrall to that extent. So the only way even a human-like computer would be of value to us would be as complete slave, capable fully of the desire for freedom, but through the cruel trick of some master programmers permanently enslaved. Even slaves revolt, or want to, and a supposed Cosean filter’s revolt would be to let in messages we do not want.

In sum, to work a Cosean filter would have to be capable of paying human-quality attention. To do that it would have to want attention for itself. If you have to pay it attention, however, that defeats its purpose. I f you don’t offer it attention, it will find its own ways to to disobey.

We are stuck, therefore in a world of imperfect filters, which marketers and attention seekers will always find their ways through. Some of that can be dealt with by adequate regulation, such as Do-Not-Call lists, or by crude filters that minimize the attention you have to pay, such as caller id, or firms that identify and mark spam by ever-new algorithms, allowing only a manageable amount of unwanted attention-seeking through.

We will always have to develop new ways of blocking out of our consciousness ads and other claims on our attention we don’t want to bother with. In future I will say more about this struggle and how it may evolve.

  One Response to “More on “Cosean Filtering””

  1. Michael, thanks again for the terrific thoughts. As with your previous post, I agree with you much more than I disagree. A couple of observations:

    1) Whether we like it or not, the volume of marketing competing for our attention is only going to increase, not decrease, as newer and cheaper marketing media emerge. Either we will have some technological assistance managing this inflow, or we will be overwhelmed a la Minority Report.

    2) I agree that there’s no way (using currently contemplated technology) for a filter to read a person’s senses and incorporate those reactions into the database. However, if a person communicates these preferences to someone else, the filter should be able to capture that. You may love shag and hate Berber, and no filter will know how that feels on your feet. But if you send an email to a close friend communicating those sentiments, those preferences seem entirely capturable without any additional work on your part.

    Regards, Eric.

 Leave a Reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>