31 August 2005

Identity Is Subjective

Pam commented:
"If you were to open an old photo album, and see a picture, let's say this picture was taken by an aunt or uncle. And this picture showed one of your children at christmas, looking up with delight just after they found out what their present was. Would you look at that picture and see that the lighting was all wrong, and that cousin Mervin was picking his nose in the background - or would you register that this loved one of yours was experiencing a moment of joy? Isn't it possible that you would register both? And that the emotion that is valid for you and a small handful of people within this very specific context makes up for the artistic absence?"
This is precisely right! I would register both a (negative) feeling for the photographic aesthetics and a (positive) feeling about my child. And a small group of people who know my children would register the second feeling, too (they might not register the first feeling, unless they too have The Photographer's Eye) - but most viewers would have either just the first feeling ("that photo sucks") or they would have the first feeling together with a generic feeling of affection toward a child at Christmas.

Why does my feeling about this (hypothetical) photograph differ from the feelings of the multitudes who might view the photo on flickr? Because of the first axiom of identity:

Umberto Eco has said that a novel is a machine for generating interpretations; the same thing is true of a picture. But which interpretation a picture generates depends on one's experience.

When I see a picture of my own child, I recognize the child. Because of my experience, I know a rich, detailed story about the child, and I associate the picture with that story (the story is, from my point of view, my child's identity - since An Identity Is A Story).

A stranger - someone who doesn't know me or my children - has nothing to associate with the picture when she sees it, but she has to react anyway.

Because the stranger's experience does not provide her in advance with a story to go with the picture, she has two choices:

The picture doesn't contain either my version of my child's identity story or the generic story which the stranger makes up when she sees the picture; it's just kind of reference to those stories. Over time, more and more people either forget the stories, or forget what the subject of the stories looked like; this tends to disassociate the picture from the stories and make the picture less useful as a reference. (I remember a photo.net thread which asked "what do you most wish were in old pictures?"; the best answer was "name tags".)

There's an important lesson here for people who want to use biometrics as identifiers; biometrics are essentially pictures of people, and people change over time. The practical effect of this is that the biometric database, over time, will tend to "forget" what the subjects of its stories look like (because it will be relying on old pictures) - and indeed one of the design parameters for biometric systems is the rate at which peoples' physical features change.

In fact, of course, everything about a person changes over time - his physical appearance, his attitudes and beliefs, his creditworthiness, his address, his name (OK, more often her name), his bank account number, his employer, and so on. This is in fact our second axiom of identity:

This is blindingly obvious if you think about it; if An Identity Is A Story, then of course an identity will change over time - because the story keeps developing (unless you're reading some awful psychological novel or play where nothing ever happens).

But let's leave discussion of the second axiom for a future post. We haven't yet exhausted the riches of Eco's observation that a story is a machine for generating interpretations.

Anytime there's a story, there's also a storyteller and an audience. The storyteller has an intention in telling the story - just as I have an intention in taking a picture. But the members of the audience don't necessarily know what that intention is, and they don't share all of the storyteller's experiences; they bring their own attitudes and experiences to the the campfire around which the story is told.

Each listener's attitudes and experiences generate a unique interpretation of the story, just as Eco observed. And this means, of course, that if I tell an identity story, each member of my audience hears a different identity story. So when our first axiom says that IDENTITY IS SUBJECTIVE, it's not just saying that different observers know different parts of the same story. Even if two listeners hear exactly the same story, each of them feels and remembers a different story.

If you think about it, this is why more than one credit agency can exist; if all credit agencies had the same algorithms for taking information about me and turning it into a credit report, or a credit score, then they would all be delivering exactly the same product, and there would be no basis (except price) for competition and no reason to consult more than one agency. It's precisely the subjectivity of identity that creates the possibility of, and the need for, competing services.

Eco is careful to note that no interpretation should be considered privileged or canonical (as indeed the credit agency example makes clear; if one agency's interpretation were correct, that agency would be able to put the others out of business quickly).

The storyteller's own interpretation is particularly suspect (Eco writes "The author should die once he has finished writing. So as not to trouble the path of the text.") What he's saying here is that interpretations are essentially subjective - that there can be no such thing as a true interpretation. And this too is true of identity stories; certainly the person the identity story is "about" is an unreliable narrator - he's got too much invested in the happy ending to be trusted to give us the unvarnished truth - but he's also the only one who knows all the facts!

26 August 2005

The Photographer's Eye

Go visit flickr. Pick out a photo with some people in it (pick one where they're looking at the camera). Now let me describe what you're looking at.

There are two to five people in the photo. Some probably have their arms around each others' shoulders or waists. They're looking straight into the camera. Probably there's a beach, or a picnic table, in the picture. Or maybe they're inside, possibly at a restaurant table, with their faces brightly lit because of the flash and the background hidden in shadow. Wherever they are, they're smiling broadly.

There's other stuff in the picture too. maybe half-finished plates of food, or a 4-wheel drive vehicle, or toys and a plastic wading pool. Maybe there's even a power line entering the right side of the back of someone's head and exiting the neck on the left side, like the Warren Commission's magic bullet.

What exactly is going on here? I asked my friend Pam about this yesterday, and she gave me a great answer, which started me thinking...

My questions to Pam were:

  1. Who are these pictures for?
  2. What are these pictures for?
She got it right in one try: who they are for is the photographer and his or her friends (they're not for you and me), and what they are for is to recall the scene to the mind of the audience.

In other words, there are really two pictures: the picture you and almost everyone else on the Internet sees on the screen, and the picture the very small group of people who were actually there at the time the picture was taken see in their minds.

Now the interesting thing is that while you and I see the assassin's power line, the Nissan Pathfinder, and the half-eaten Pad Thai, the people in the intended audience do not see these things. The first picture - the thing on your screen - is just a mnemonic to them; it's the Platonic shadow of the second picture (the "real" picture in their minds). When the people in the intended audience look at the first picture, they see the second picture.

I couldn't understand why people were putting these pictures on flickr, because I have a particular type of brain damage which caused me to forget that the second picture exists. The type of brain damage I'm referring to is "the photographer's eye".

I've been photographing things for a long time. When I started photographing things, my photographs were mnemonics, like the ones on flickr.

If you photograph long enough, however, you'll eventually have the experience I had: you'll take a "real" photograph. What I mean by a "real" photograph is one which does not require the viewer already to have the real photograph in her mind in order to see the real photograph.

This experience of duplicating the real photograph you have in your mind with the one you put on film changes the way you see forever. Like certain psychotropic drugs, it actually re-wires your brain.

This brain damage makes you look at pictures differently. Because you know it's possible to make the mnemonic (the picture on screen, or on paper) look like the "real" photograph in your mind, you start to think about the picture in your mind and make sure there's nothing that doesn't look like it in the viewfinder before you press the shutter button. So when you see a Nissan Pathfinder, you walk around until it isn't in the picture.

As the condition progresses, you spend more and more time moving around and fiddling with the camera to make sure that what the film sees is what's in your mind. I take pictures of my friends like everyone else does, but they don't look like the ones on flickr. When I take a picture of Pam at a conference we're both attending, for example, it looks like this:

I didn't use a flash, or studio lights, or a fancy background, for this photo. I just asked Pam to sit next to a window, fiddled with my (manual) camera settings a bit, and took the picture. What you're seeing here is what I saw. But unlike most of the pictures you'll see on flickr, you're not also looking at whatever else happened to be in the area when I saw what I saw.

(Full disclosure: I did a little dust removal in Photoshop, took out a stray hair or two, maybe touched up the odd mole, and cropped the picture from 24x36 proportions to 8x10 so that it would print on standard photographic paper - but otherwise you're looking at what came straight out of the camera.)

My picture is better for me than the ones on flickr, because the brain damage has rewired the part of my brain where the mnemonics live. I still take mnemonic pictures, by accident, because it's hard to take a "real" picture. But I don't post them on flickr - I just file them away in their little boxes in the closet. They annoy me, because instead of making me recall the scene, they make me think "the scene didn't look like that!", which makes me feel like I've failed as a photographer.

My picture is better for you too, especially if you don't know Pam. It's better for you because you don't have to read my mind ("Was the photographer looking at the girl, or at the Nissan Pathfinder?"), and you don't already have to have a picture of Pam in your mind, to figure out what I saw. I removed everything that wasn't what I saw from the picture before I took it. All that's left is what I saw, so that's what you see - even if you weren't there at the time and even if you've never met Pam.

22 August 2005

An Identity Is A Story

Ceci n'est pas un Bob.

After thinking for a long time about Kim Cameron's Laws of Identity, I gave a talk on Identity at the Burton Group's Catalyst conference last month. Jamie Lewis and Kaliya Hamlin (who blogged my talk) approached me after the session and urged me to start my own blog in order to join the online identity discussion. If you like what you read here, you should thank them for pushing me past my deep ambivalence about blogging.

Here's a little taste of what you might see here at ceci n'est pas un Bob over the next couple of weeks.

I think identity behaves in consistent and predictable ways in the real world, BUT most contemporary discussions of identity are completely out of touch with what identity really is and how it really works. To understand how identity behaves, it's necessary to distinguish the different uses people make of identity, and consider each of those uses individually.

I think a set of axioms of identity can be defined which describe what identity can and cannot do, and what it will and will not do in particular circumstances. We can enumerate these axioms by looking at centuries of thought about identity and examining that thought in the light of situations which occur in the real world today.

I think that systems designed with the axioms of identity in mind will be more effective than systems designed without regard for the axioms.

I think that the axioms define how identity and privacy are related, and can help illuminate when we can determine identity, when we can protect privacy, when we must choose, and when we are out of luck on both counts.

As usual, it's best to start with definitions.


The problem of identity has a long and difficult history. Nietzsche, surveying that history, threw up his hands:
"Belief in the identity of different things, or in the identity of the same thing at different times, is a fundamental philosophical error"
This was already an old idea in the East, of course; Buddha denied the existence not only of identities but even of things before 500 BC.

The West conducted a nuanced discussion of identity for centures, until the industrial state decided that identity was a number you were assigned by a government computer. Aristotle's tortured treatment in the Topics set the stage for centuries of debate about what exactly it is that is "identical" in the identity of a person.

Serious modern discussions descend from book 2, chapter 27 of Locke's "An Essay Concerning Human Understanding", which distinguishes between the identity of a body, the identity of a man, and the identity of a person, and which makes the argument that the last of these is what is of interest in cases of law, because without continuity of memory and intent, there is no rational basis for the assignment of praise and blame.

Literature has mined the philosophical discussion deeply; mistaken identity is one of the great themes. Think of Twelfth Night, Dr. Jekyll and Mr. Hyde, The Man in the Iron Mask, Zorro, The Phantom of the Opera, and The Picture of Dorian Grey, for starters. If you haven't read it, Christopher Priest's "The Prestige" is a great recent addition to the genre.

If you want to follow the philosophical discussion into the modern day, start with "The Identities of Persons", by Amelie Oksenberg Rorty.

You should want to follow the discussion, because the twentieth-century statist notion that each person has an unchanging identity which can be observed by a third party (for example, a clerk at the Department of Motor Vehicles) and with which a number can be uniquely and persistently associated has broken down, and the consequences of the breakdown are intruding on your personal life, liberty, and privacy right now.

Don't believe me? Let's look at an example. ICAO is in the process of finalizing standards for incorporating biometrics into passports. The motivation for putting biometrics into passports is that better identification will help in "the war on terror". If the problem in the war on terror were preventing known terrorists from crossing national borders at official checkpoints using genuine passports issued in their own names, biometric passports might help.

But let's look at something Rorty says in her introduction:

"Why are we interested in someone being the same person, and not merely the same human being or physical object? One reason is primarily retrospective: we need to know whom to reward and whom to punish for actions performed when "they" were acknowledgedly different in some respects from the present population. But we have more forward-looking reasons as well: we want to know what traits remain constant so that we can know what we can expect from the persons around us. We assign crucial responsibilities to individuals, assume important continuing relationships to them in the belief that certain of their traits are relatively constant or predictible."
If you buy this argument, you're forced to conclude that "whether a person is the same physical object as at some past time" (which is what a biometric sensor measures) is not what you need to know, either to punish the person for a past misdeed or to predict whether they're going to do something bad in the future.

How much taxpayer money are you willing to spend on biometric passports before you know whether you believe this or not?

In the National Academy of Sciences Committee on "Authentication Technologies and their Privacy Implications", we tried to define terms like "Identity", "Identifier", "Attribute", "Identify", "Authenticate", and so on - as a contribution to settling these arguments before systems began to be designed and deployed. If you're interested in the topic you can read our entire report; if you want to keep up with the identity discussion you can just take a look at the definitions.

I still like our definitions very much, but I'm going to venture my own personal (new) definition of "identity" here. I think it's consistent with the committee's definition, but it's a lot catchier:

It's that simple. One of the reasons it's simple is that I've left out lots of things which are commonly but falsely assumed about identities.

For example, I might have said "An identity is a story about a person". But this overconstrains the definition: lots of identities exist which are not stories about people. You know many of them:

  • Captain James Tiberius Kirk
  • Lara Croft
  • The current King of France
I might also have said "An identity is a true story". But of course truth isn't a requirement for identity attributes. The only time Richard Nixon ever told you "I am not a crook" was after he had committed a crime but before he was pardoned for it. In almost every identity story, some elements are true, others are false, and others have an indeterminate status (that is, people disagree about whether they are true or not, but there is no way to settle the argument).

John Demjanjuk has been associated with at least three identity stories. These stories are inconsistent with one another, so they cannot all be true in all their particulars, but courts in two countries on different continents have been unable after extensive investigations to pick the stories apart and determine what's true and what's false.