From the Mind of Trevor Stone

This song is called “Alice’s Algorithm”
It’s about Alice, and the algorithm
But “Alice’s Algorithm” is not the name of the algorithm
That’s just the name of the song
That’s why I call the song “Alice’s Algorithm”

You can send anything you want with Alice’s algorithm
You can send anything you want with Alice’s algorithm
Send a syn, you’ll get an ack
Two prime numbers are hard to hack
You can send anything you want with Alice’s algorithm

Now it all started two Thanksgivings ago, two years ago, on Thanksgiving
When my friend and I wanted to send a message to Alice with an algorithm
But Alice doesn’t get mail through the algorithm,
She gets mail on a server running the algorithm,
In the home directory, with her copy of RSA and Carol the sysadmin
And livin’ in the home directory like that, they’ve got a lot of room in /var where the spool used to been
Havin’ all that room (seein’ as how they deleted all the spools)
They decided that they didn’t have to take out their garbage for a long time
( Read more... )

Context for non-computing folks: computer messaging protocols are often described with Alice, Bob, and Carol sending messages and Eve trying to eavesdrop. There's also a lot of Unix references in there.
Context for non-American-folk-music folks: Alice's Restaurant by Arlo Guthrie is one of the greatest satirical songs ever performed. Listen to it first to get the guitar portion rolling around your head while you read these filk lyrics.

At the Aaron Swartz memorial in San Francisco (video), some interesting themes emerged.

The first is Aaron's passion for machine-readable public information. This principle is at the core of much that Aaron did, from enabling search engines to find public domain and CC-licensed content to downloading swaths of paywall-guarded documents so that the public can have access to its own information.

The second is the unbalanced power wielded by prosecutors. Aaron killed himself in part because he felt helpless when faced with a multimillion dollar federal trial featuring 13 felony counts. If Aaron couldn't face this, what hope have ordinary folks who aren't close friends of Harvard law professors, rights advocacy organizations, and expert witnesses, not to mention a chunk of cash from selling an Internet startup. Faced with expensive defense lawyers and the fearsome specter of the government's prosecutors, only 3% of cases make it to the trial which we're constitutionally promised.

The third, expressed by his girlfriend Taren Stinebrickner-Kauffman and fellow document liberator Carl Malamud was a call to the technologists and scholars and activists to become radicalized. Aaron did big things because he thought they mattered. Like Peter Singer, he stressed about the opportunity cost of not doing the most important thing in the world. His death has become, in part, a call for people in the free culture movement to step up and do more.

So here's an interesting challenge that combines all three: write a program that interprets and presents law. Though they predate computers by a few thousand years, laws are meant to be something like human-runnable source code. They're detailed, they're written explicitly, and they apply to everyone. And yet in many cases it takes someone with a graduate degree to understand what they say. People with graduate degrees are expensive and unevenly distributed.

Imagine we had a program which could turn laws and judicial opinions into machine-readable format. We could then write programs that took those laws and presented them in various ways, helping lay people understand both core details and subtle interactions. We could write other programs to organize this legal information into arguments given the evidence about a case.

Compared to people with graduate degrees doing stuff, running computer programs is free. Someone without a lot of resources could understand what they're charged with, explore similar cases, and collaborate with friends on a defense. There'd still be a role for lawyers to conduct the defense at trial and advise on the best way to convince a jury, but the time spent at trial is today dwarfed by the time and expense preparing for it. Let the humans do what they're good at&endash;convince humans of things&endash;and let the computers do what they do best&endash;tirelessly and cheaply examine lots of data and find useful patterns.

Like a patient who comes to a doctor after reading the medical literature and closely observing his body, a defendant who comes to a lawyer with a solid understanding of the relevant laws is in a much better position to face the plaintiffs and prosecutors who have the deck stacked in their favor. If we can make computers understand law, we can empower all citizens, regardless of income, to make fair use of the due process granted them by the constitution.

Building such a system wouldn't be easy. Human language is still hard for computers to understand. And legalese is even hard for humans to understand. There are all sorts of powerful people and organizations, private and governmental, with interests vested in law and courts being expensive and difficult to access. It's not easy, and that's why it should be done. A hard, ambitious, and meaningful project like this would capture the spirit that's been raised in Aaron Swartz's wake.

I had a thought at work the other day. Java has a strict one-parent-class rule. Philosophically, this is a good match in domains like taxonomic biology (a phylum is in one kingdom, a genus is in one family). But it's a terrible match in domains like cooking. A soup inherits flavors from many sources. You can add meat to what was originally a vegetable stew. Ruby is a much better language for programming a curry.