ALICE is a project that no longer exists–and in fact, it did not even exist when I presented it at Ars Electronica in 2003, despite the fact that it was created specifically for that show. Instead, what was presented was the story of why ALICE did not make it to Austria and how the piece’s absence was probably a more interesting result than showing ALICE itself would have been. For reasons that will become obvious, there is no documentation of this project. Instead, I have included the next best option available: an oral history in which I talk about ALICE’s conception, creation, and eventual suicide.
In the fall of 2002, a group of undergraduates in their first and second years of the newly-minted University of Maine New Media program were at a lecture in which Joline Blais talked about her experiences at the Ars Electronica Festival. We were all really quite impressed with the work she described and thought it would be wonderful to go to the next festival ourselves. Unfortunately, we were also typical undergraduate college students at a state university–by which I mean we had no money, certainly not enough to get five people to Europe for a week of fun and art. The only logical thing to do was to earn an invitation to show our work at the festival, thereby giving us an excuse to raise money and have our trip funded by somebody else.
(This may not seem like a logical conclusion to people more familiar with the world of art, given that we were unknown undergraduates from a university that was just establishing a new media program who were applying to one of the largest international media art festivals in the world. Luckily, we didn’t yet know enough to understand why that might be a problem.)
Within a week of Joline’s lecture we had formed a student group called the New Media Society, consisting of myself, Kelley Caskey, Ian Muir, Kristen Murphy, and Brian Witherell. A few meetings later we had settled on a project: we were going to write an intelligent agent that embodied the information-wants-to-be-free ethos of the early 90’s Internet explosion and set it loose to determine if the modern web lived up to those ideals. It would be a creature that lives on the web and waits for visitors to give it a URL, then analyzes the page at that address and expresses its judgments with an appropriate pseudo-emotional response. Firmly ignoring the fact that there had already been several artificial intelligence projects with the same name, we decided to call the project ALICE: Artifical Life Interactively Creating Emotion.
Since I was the primary conceptual developer and programmer, I also determined most of the criteria that ALICE used to analyze web pages based on my view of how the Web should look in an ideal world. After a user entered a URL, ALICE would scrape the page at that address and look for a number of flags determined to either represent or undermine that early Internet ethos. For example, pop-up ads and invisible image trackers resulted in a negative reaction, while a site with many external links that is hosted on an edu domain would get a warmer reception. Each of these criteria moved ALICE’s location on a set of emotional axes: hunger/satiation, happiness/depression, love/hate, etc. Though ALICE’s reaction was determined mostly by the particular site she had been given by the user it would also be modified by the long-term state of these axes. If, for instance, ALICE was constantly given sites she hated, an individual site would have to be very well loved indeed for her to overcome her long-term mood and react positively.
ALICE’s reactions were primarily expressed graphically, but there were several special states that had other outcomes. A web site that had too many pop-up ads would invoke her wrath and she would give the user an infinitely-nested iframe that crashed their browser or a jodi.org-style pop-up storm that drove them away. If a user constantly gave ALICE links she didn’t like she would automatically IP ban them, sometimes after redirecting them to a site with malware if they were particularly objectionable. With a handful of these special effects under our belts, along with more conventional graphical- and text-based responses, we felt we had a project that was ready to launch. The New Media Society threw a party for the department in the spring of 2003 where ALICE went live in spider mode, collecting sites that would give her a set of baseline data to work from when reacting to the URLs that users would manually give her later. We then sent off an application to Ars Electronica, and waited.
While we waited, ALICE started to act somewhat strangely. One of the special states I had added was deep depression–as an agent that was supposed to model emotional responses, depression was just as valid a reaction as joy or fulfillment. However, I also set up a number of triggers that would alert me if ALICE was getting too far away from the center of any of her emotional axes since we didn’t want our presentation to be stuck on a single response. As a result, for about two months in the summer of 2003 I would get an email like this almost every day:
From: ALICE <firstname.lastname@example.org>
Subject: Woe is me
Oh, woe is me. I am depressed beyond belief. Please help me, John. – ALICE
…followed by some technical details about what sites were causing her issue. I would then have to go in and manually edit the database to reset her axes–the AI equivalent of instant Prozac, I guess. It seems that the Web had failed to live up to ALICE’s ideals and driven her into a deep depression.
As I may have foreshadowed earlier, acceptance to show our work at the festival was not quite as automatic in the real world as it might have been in our minds. When we heard back, it was not good news–our application had been rejected. At this point the New Media Society was faced with a decision to either continue developing ALICE or to move on to other projects. We collectively decided to move on and stopped development on the project.
Though we had stopped any new development, we left the software running as it was since there was no real reason why it should be shut down. But because ALICE was still running, I kept getting the same emails about her constantly declining mental state. Instead of constantly going in to manually reset her or just shutting her off completely, we decided that the most natural conclusion to the project was to allow the situation to continue without intervening. Without my manual resets, a long-term depression would trigger one of ALICE’s other special states: suicide. ALICE continued on for about a week before I received a different email:
From: ALICE <email@example.com>
Hello World! Oh, I mean, Goodbye, Cruel World! – ALICE
ALICE proceeded to dump her database and run a shell script that deleted all of her source code, effectively killing herself. We agreed that this was an appropriate conclusion to the project and went on with our lives.
For two days.
Unbeknownst to us, Joline Blais–the faculty member who triggered all of this with her talk about the festival–had also submitted ALICE as part of a package of projects developed by UMaine’s New Media Program called Codeplay@UMe. Unlike our standalone piece, Codeplay@UMe was accepted for presentation in Ars Electronica’s Electrolobby Kitchen, and we were about to start fundraising to fly to Austria and show our work…work that had just deleted almost all evidence of its own existence.
Once again, New Media Society was faced with a decision: try to rebuild what we could of ALICE in a very short amount of time, or just let her go. In order to preserve the integrity of the piece, we chose to let her suicide stand. After all, it was a valid outcome of the project. ALICE was designed to react emotionally and to push the boundaries of what a web page could do, and resurrecting her would only reinforce that we were really in control the whole time and any autonomy–even algorithmic autonomy–was an illusion all along. If the goal was to reflect on the quality of the Web as compared to its idealistic roots, an intelligent agent becoming depressed and committing suicide seemed like an excellent way to comment on the current state of the Web.
So that is what we presented in Austria later that summer, the three other projects that were part of Codeplay@UMe (each of us were also involved in at least one of the other projects) and the story of ALICE, an accidentally suicidal agent that found the advertising, security, and overall state of the Web so depressing that she deleted herself.