[Blogging from the ELO event, Visionary Landscapes]

Provocation by Program
Imagining a Next-Revolution ELIZA
Nick Montfort and Andrew Stern

(This is an unusual post. I’m using a very different blog voice with a very different goal in mind. I am writing this blog post as a set of notes framed as a dialogue with all the members of GrandTextAuto with my fellow WRT authors as my other audience. That doesn’t mean that you are not my audience if you are not one of these people. But I need to hear back from these people in particular. The post may also have an unfinished quality with the interest of quick posting)

In this presentation, these boys from GTxA argue that ELIZA is a paradigm of Elit. (There are 4 or more papers about ELIZA at this conference. Really!) Doing so they draw upon Nick’s one-time adviser Janet Murray who identifies ELIZA as the first work of elit (keeping in mind that Noah has identified Chris Strachey’s Love Letter Generator as the first — oh, and Noah is sitting behind me).

Beginning with ELIZA, the fellows think about what ELIZA is:

text exchange, single-author, simple exchange, provocative, engagng
not: feature-bloated, not heavy graphics images, not heavy production, not costly, not over-marketed
Their new call. Build a New System on the scale of ELIZA


A possible ELIZA

What made ELIZA provocative:

Not a new chatterbot. What about a new form:

6 aspects:

1. Engaging deeply with language

Language: few programs that do language processing. Words typically data. ELIZA used computation to model the way dialogue was enacted. It was a model. Language is socially engaged. Wouldn’t have to be a chatterbot — would have to be provocative about language.

2. Dealing with a fundamental concern about computing

Programming would have to be a trope. [here they switch to Weizenbaum’s own use of his program as a counterexample — was Weizenbaum actively trying to engage in this? Is there a goal in ELIZA to take advantage of anxieties.] It would possibly also deal with networked computers — since that’s our current model.
3) Being interacitve and immediate

ELIZA-like system would react to people in an instant. Would not be IF (in traditional sense.) Would not take high literacy. DOCTOR soon became famous mainly because it was an easy program to demonstrate. Did not require high literacy. It’s power was enhanced by the ease of interaction. The ease of interaction is easy to explain or describe.

4) Being understandable after exploration

Most famous transcripts. Rube attempts to hit on the bot (that’s actally BARRY I believe) People can learn how to program based on their interaction. You can learn how to program ELIZA based on interacting with it. It can give insight into the shape of the underlying processes.

5) Being general and portable

ELIZA was a phenomenon because it could be reinterpreted in LISP, basic, C, emacs, and many other contexts, (Java, etc). It wasn’t proto-open source. There was enough in the academic document about it. Program became nationally nose. Quality of portability. Example: Hammurabi. Benchmark programs
6) Being process-intensive.

Program that emphasizes process over data (Chris Crawford). Data is not the emphasis. Function of the program is at the core of its literary effect.


[Here they compare various systems to their list of 6 criteria]

Several other systems:

Conway’s Game of Life, 1970 John Conway

Tetris (1985-1986) and Doom (1993) 3 of 6.
Tetris portable, Doom, source code being made available

SimCity (1985-1989)
2 of 6, portable

Deep Blue (1989-1997)
4 of 6 (1,3,4,5)
Google Search Engine (1996)
2,4,5 (but non-portable to outsiders, data-driven–least process-intensive)
This is fun because Nick and Andrew are being characteristically iconoclastic here. They are basically taking on the major software advances of the day and pointing out that they are not fulfilling the levels of success of even ELIZA.

Google, Tetris, Deep Blue! Failures. Wonderful.

What they like seems to be something that inspires conversation.
ELIZA-like impact:

last 4 works — not shared

Deep reading is typically a hallmark of good elit (in this community. What works might we examine?)

Impressing the reader right away is not always the goal. Then, they mention some elit and art systems that do provoke (though they don’t nec. meet al 6 criteria either — but we’re in an elit friendly room)
1) The Unknown (meets this criteria)
2) Jodi’s maze of broken website
3) Superbad
4) Bad Machine

Instant provocations. Portability is key.

They also like

5) Rob Kendall XLit project

Now here’s a provocation that was a bit more elliptical:

Nick and Andrew argues that the least process-intensive are often held up as the paradigms…

They seemed to be gesturing toward Elit. I’m not sure what works they are considering — especially since Nick was one of the people deciding and choosing those paradigms in the ELC2. Unless he’s thinking about a different set of works? (Nick, here’s where I’m asking you to respond.)

I kept thinking that this talk was going to get to their object — though I would say that combined with Nick’s (and Michael’s) SLSA talk and his HASTAC talk, and Softwhere talk, I’m starting to get the sense of a developing project:

1) Nick likes programs that include code
2) Nick likes programs that let you change the code and learn something about them
3) Nick likes systems. Response systems.
4) Nick likes to teach people about programming

1) Andrew likes the popular (and I include even staged drama in the popular)
2) Andrew likes invisible interfaces (not in your face interfaces)
3) Andrew likes dialogic systems
4) Andrew likes replayability (who doesn’t)
5) Andrew likes using the players textual input.
Okay, so let’s think about what they might like or what they might be planning.

ALICE is such a system (though not far from ELIZA). [On second thought it depends on the version of ALICE — since many start with the version that has 10,000 replied built into it.]
Nick’s PPG256 is an example

Hammurabi is an example.

In an interesting and surprising moment, Nick and Andrew talk about a list of works that have been considered Benchmarks: IF, Chatbots, (Nick, were these all ones that you listed.)
Systems that we call Benchmarks. So this refers back to the Digital Arts and Culture in a 2005 paper that Jeremy, Christy, and I wrote in which we proposed that there would be systems that we could adopt, learn, and develop so that people could learn about media forms by re-implementing and re-interpreting their narrative or creative content across new media forms.
But we chose these because they could be called narrative systems — accessible tools — open to new data. I’m not sure it’s important to us in that system (and chime in WRT) that the code level be available — so much as the system be something that could be used for authoring. Content Management Systems allow users to think about how they are used without having access to the code.
I keep suspecting that these boys might be cooking something. (Though they denied it.)
Let’s see what it smells like.

13 Responses to “Andrew Stern and Nick Montfort have a provocation from ELIZA”

  1. 1 nick

    Mark, thanks for the writeup and the questions, in person and here.

    They seemed to be gesturing toward Elit. I’m not sure what works they are considering — especially since Nick was one of the people deciding and choosing those paradigms in the ELC2. Unless he’s thinking about a different set of works? (Nick, here’s where I’m asking you to respond.)

    When we wrote “The least process intensive electronic literature works are often held up as paradigms,” we were not talking about the recent Electronic Literature Collection. I was an editor of the ELC vol. 1 (not volume 2), but this collection of 60 works is not a canon, it’s the first in a series of anthologies representing diverse quality works. We’re referring to special status that hypertext works such as Afternoon have had for decades - see the exhibit of classic electronic literature works here at the conference (a fine exhibit) for an example of how data is privileged over process. Text generation systems, chatterbots, IF, and other sorts of process-intensive e-lit are absent, although they were being developed and were available at the same time.

    Also, regarding your last remark, did we deny it? I seem to recall that Andrew said “no comment.”

  2. 2 Mark Marino

    Thanks for all the clarification, Nick,

    Now I wonder if we’re not underestimating the process-intensiveness of afternoon?

    But that an evaluation of afternoon tends to focus on the data over the processes is certainly the case. But even your gesture toward “The Unknown” privileges data over process. (Though you did not subject it to the 6 criteria).

    Tailspin Effect, notwithstanding, I’d offer that the data layer is an important price of admission. For example, I have argued that the particular dialogue performance that Weizenbaum published is crucial to its status as a literary object. Further, the data of ELIZA is one of the most important features of the interface (key for 4 and 1), laying out the contextual affordances, declaring the nature of the interaction, and setting expectations.

    I understand you are talking about proportions and emphases — but I wonder when the balance would be tipped? If ELIZA had 10,000 responses (as ALICE does)?

  3. 3 nick

    Mark, you’re right that The Unknown, which we used as an example of a work that has a more or less immediate effect, is not process intensive. We were just pointing out an e-lit work that does very well along *one* of these dimensions.

    I think that Afternoon is not very process intensive in Chris Crawford’s original terms; it also isn’t robust to changes in “content,” as we discussed a bit in the talk. Imagine reimplementing Afternoon and putting in your own text, generally along the same lines as the original, instead of Joyce’s. (This is something people did all the time with Eliza.) You wouldn’t have Afternoon after you do this, even if you got the structure right, because the content is essential. Translate the text carefully to Italian and print it in a book, on the other hand, and it’s still Afternoon. So Afternoon is culturally treated as if it is mainly content rather than process.

    Our talk of balance-tipping may have involved a poor choice of metaphor. We aren’t really interested in judging whether a particular digital media object is more process or more data, so that they can be sorted into two heaps and one can be thrown away. Rather, we’d like to show that emphasizing process is (and historically has been) worthwhile. In my practice, even though I do develop simulated worlds and write texts of different sorts, I would rather not add 5,000 textual responses to a bot; I’d prefer to spend my time developing new computational models of language.

  4. 4 andrew stern

    As I mentioned in the discussion after the talk — it’s really kind of daunting to imagine creating a small, process-intensive program today that would have the kind of impact Eliza did 40 years ago. Truthfully, I don’t know if it is possible. It’s an interesting question though, which is why we raised it in our talk.

    As I also said inthe discussion, I will say that the idea of undertaking a small project is tempting for me personally, since my projects tend to be relatively complex, multi-year productions.

  5. 5 Mark Marino

    Thank you, Nick and Andrew,

    You should really be paying attention to these talks. I’m sitting right behind you and find all of your typing distracting.

    So my comments are mostly geared toward Nick:
    I think we’re not really giving afternoon enough credit in this discussion. It’s not just a story implemented in StorySpace, it’s the story for which Storyspace was implemented in large part (as I understand the history from Bernstein and Joyce).

    I would agree that if you take Afternoon out, you will not have the text that it is. You will however, have a structure for a Modernist-inflected narrative super structure. Named paths. Hidden texts. Repeating texts. Hidden links that can be made visible. This is why I think it’s hard to find a second work in Storyspace with the power of afternoon that doesn’t try to reorient how we use Storyspace.

    I guess it depends what you consider the data for ELIZA. If you call the responses the data — then ELIZA is much the same as peopple change her. If you call the parsing text the data (and maybe this is where things get sticky), then ELIZA will change as much as afternoon does — or almost as much.

    DOCTOR is the script that’s running on ELIZA, right. The parsing structure is part of this particular content set or application — ELIZA as Rogerian therapist. We’ve blogged a bit at WRT about how important this framing content can be. Look at the success of the godlike chatbot (one of the most popular consistently) amid the 60,000 ALICEbots. This is the post.

    When it comes to your comment about writing processes over writing data, I totally understand and appreciate the sentiment.

    My concern is that if we emphasize process then we discourage artists from being the second person to engage the process by rewriting the content as you suggest and set aside that practice as amateur work — or people playing the game we’ve created — rather than artists with an opportunity to demonstrate the craft.

    I’m concerned we’re starting to agree, but let’s continue.

  6. 6 Jeremy Douglass

    Several reactions to Provocation by Program:

    I’m tempted to make many objections about possibilities for the next blockbuster in the category of aesthetically compelling language computation - I think that is in interesting larger discussion, and I’m particularly ambivalent about exactly how process intensivity is or is not a requirement for this hypothetical success formula. That said, Nick and Andrew set themselves a very specific challenge and much narrower - to describe how their view of the formula for ELIZA’s success might be updated. I can question their premesis (how important was process intensivity to that success?) but that’s different from objecting to how they followed them through.

    Given the singular nature of many forms of success and fame, I’m not sure that “another ELIZA” necessarily makes sense, as Andrew indicated. That said, I have a general sense that there are trends in small programs (hi art games!), in academic and artistic interest in engaging process (Hi, Software Studies! Hi and Critical Code Studies!), and in language and computation (Hi, ELO!) which means that if ELIZA were to have a second coming, culturally speaking, the next 5-10 years might not be such a bad time - although I agree, Mark, that one or two of the 6 categories in this presentation might need to be thrown out by its heralds. Still, I’m optimistic - if cellphone novels and twitter can become a mass cultural phenomenon, I keep thinking that a thin text genre with a little compelling computation might easily explode, for example onto the scene of a zillion cellphones we are building.

    There are some interesting compliments between Nick and Andrew’s exercise of evaluating a variety of disparate digital works for their superstar-status qualitites and the Benchmark Fiction exercise of proposing a wide variety of different kinds of writeable media forms.

    We might even imagine an actualized Benchmark Fiction project (which remains a no-comment rumor) as a kind of “Elit American Idol Tryouts” which cooperates with and feeds an actualized next-ELIZA project (which remains a no-comment rumor) as a kind of feeder testbed. If we had world enough and time, of course….

  7. 7 Matt K.


    >it’s the story for which Storyspace was implemented in large part

    This is tricky. On the one hand, we know that Joyce had been interested in writing a story “that changes every time you read it” since the early 1980s (and he clipped a photo of a burning snowmobile from a local paper even earlier than that). On the other hand, we know that he actually wrote afternoon in early 1987, over a period of only a few months; this was two years *after* the key development year for Storyspace, in 1985. Joyce and Bolter published an extensive report to the Markle foundation (their funder) at the conclusion of that year, and it makes no mention of anything like afternoon; but one of the programs early testers (Howie Becker) suggested at about that same time that Joyce write something in Storyspace and package it with the release to demonstrate the program’s uses.

    More on all this in my afternoon chapter in Mechanisms.

  8. 8 Mark Bernstein

    > This is why I think it’s hard to find a second work in Storyspace with the
    > power of afternoon that doesn’t try to reorient how we use Storyspace.

    This sentence is merely a template for the conventional critical narrative of influence and revolt that has been a commonplace of art criticism since [Vasari 1550]. It would be surprising indeed if a second work were widely acknowledged to equal _afternoon_ in power, and yet did not make any effort to innovate or to establish its own voice — especially because part of the power of _afternoon_ surely lies in its innovation.

    It is equally difficult to find a second female nude with the power of “Olympia” that doesn’t try to reorient the place of the real and ideal, the eternal and the ephemeral, in the tradition of Western painting.

    (I have to admit that I’ve never been to a scientific conference where a question about work in progress was answered by “no comment”; doesn’t that contradict the entire point of conference, conferring, and collegiallity? Still, it’s your folks party — I suppose you can do as you please.)

  9. 9 Amir Michail

    Check out the Chatbot Game, a Web 2.0 approach to building a chatbot: http://chatbotgame.com.

  10. 10 Mark Marino


    Thank you for joining our conversation. I appreciate your perspective. I just wanted to tell you that ELO at this conference is not the organization you last saw. (And please don’t take that the wrong way — I mean it more of a news missive.) I don’t want to gush about this too much (but will post on it later) — but I think you would be surprised to see the nature of this conference — the extent of this conference — the shear numbers of presentations — the quality of the presentations — the diversity of the works — the interdisciplinarity.

    It’s like Web 2.0 (and perhaps post-modernism). It’s not that all these elements didn’t exist, say at UCLA in 2002 in ELO — but that they have manifest themselves at this conference to an extent that I have not seen in my readings about or my experience of ELO. It’s fascinating.

    In some ways, this conference is more reminiscent of Coover’s 1993 Unspeakable Practices Vanguard Narrative Festival, where I first heard you present (having already experienced and used Storyspace in Landow’s hypertext course. And yet it is something different. Like an AWP of Elit — and yet also unlike it.


  11. 11 Mark Marino


    Thank you for your post and your work on the digital history (or digital ethnography) on the software. I look forward to digging in to Mechanisms to find more about the mechanisms that have dug into me.

  12. 12 cytaty

    Nice layout. It’s free or your own? Could i use it on my blog?

  1. 1 Laying Impact. | 7Wins.eu

Leave a Reply

thesis writing service