‘Not to be killed without warning’, 2012

Last year the Intelligence Advanced Research Projects Activity (IARPA), an American intelligence agency, launched ‘The Metaphor Programme’: an open call to tender for methods of computer-analysis of English, Farsi, Russian and Spanish, to map their respective metaphorical economies. The idea, presumably, is that by developing a systematic model for each language’s network of concepts – in English, for example, heat’s association with anger (hothead, fiery temper, boiled over etc.) – vast quantities of writing and speech could be quickly parsed by software seeking to register subtle shifts of cultural attitude or public mood. The call was aimed at commercially-minded academics and is presumably catnip for bleeding-edge ‘natural language processing’ types, at least those who don’t have any qualms about providing “our nation with an overwhelming intelligence advantage over future adversaries.”
I’ve been struggling with another form of natural language programming: I am trying to learn how to write a text adventure or, as I’ve been learning to call it, an interactive fiction (I.F.), using a programme called Inform 7. Text adventures are graphicless adventure games, played in a call-and-response style by typing simple commands into a parser, and were one of the earliest forms of computer game. I had at least one text adventure on my 32k BBC Micro as a child, Dr Who and the Warlord (“You emerge from the Tardis to find deep mist all around…”). If you’ve never played one, here is a sample game session (player commands in CAPS):

In fact, this is a rather idealised version of how most interactions would play out. It is never very long before you generate the response, “That’s not a verb I recognise,” or perhaps some jokier equivalent. Every text adventure follows the trajectory from a horizon of apparently infinite possibilities – the mirage of an environment bounded only by language itself – to the shrunk-down confines of a lexicon in the low hundreds, and in fact that is perhaps even one of its pleasures. Nick Montfort, in his groundbreaking book Twisty Little Passages, quotes Jim Menick, author of Basic Adventure and Strategy Game Design for the Apple (1984): “The first step for the player is figuring out what language the game speaks… One of the joys of adventuring is that discovery of the extents and limitations of the game’s vocabulary.”
Interactive fiction peaked early: from its origins with the game Adventure in 1975 (a cave-based fantasy based on the real spelunking of its original author Will Crowther), it entered a mid-‘80s ‘golden age’, particularly embodied in the company Infocom which published – amongst many others – the Zork trilogy, and Douglas Adams’ adaptation of his own best-selling Hitchhiker’s Guide to the Galaxy. (Adams later authored another interactive fiction for Infocom, Bureaucracy, which, with its deliberate tenor of data-entry bathos, appears to be both a satire of 1980s British public services and of interactive fiction itself.) This mid-decade moment was the utopian high-watermark of I.F., which drove Mary Ann Buckles, for example, to divert her doctorate on German literature at the University of San Diego into a study of Adventure (now belatedly acknowledged as the first academic study of computer games). Rapid advances in home computer processing power and the advent of graphics, however, meant a precipitous decline in the text-only genre, to the point of its commercial extinction by the end of the decade. (Buckles gave up academia and became a massage therapist.)
What is remarkable is that, from the early ‘90s onwards, a self-sustaining community of amateur I.F. enthusiasts has kept the flame and the form has continued to develop, despite the commercial vacuum. A number of cross-platform interpreters for Infocom’s Z-machine programming standard were made available such as Zoom and the wonderful Frotz app, which makes a huge library of classic games freely available on iPhone (touch screen copy-and-paste functions are actually ideal for I.F.). In 2003 MIT published Montfort’s Twisty Little Passages, the first book-length academic account of I.F., and in 2010 Get Lamp, Jason Scott’s highly enjoyable talking-heads history of I.F. was released on DVD. [On YouTube it’s a Google Tech Talk – don’t be put off by the intro, the film proper starts about 7:30 – and it’s actually quite nice to have the live audience laughter in the background at some of the in-jokes.]

One of the most significant reasons the I.F. community has continued to thrive is that new platforms for developing games have been made and distributed for free, in particular the Text Adventure Development System (TADS) and Inform. Inform was originally developed by Graham Nelson, an English poet and differential geometer, who famously declared that an adventure game was “a crossword at war with a narrative.” Nelson is the author of The Craft of Adventure, a short book covering the history and principles of I.F., including his ‘Bill of Player’s Rights’ (“Not to be killed without warning… not to be given horribly unclear hints… not to have to type exactly the right verb…”).
With Inform 7, developed in collaboration with Emily Short and released in 2006, Nelson completely rethought the programme and created a kind of symmetry between the programmer and the player, by giving the design itself a natural language interface. To develop the game, you could describe your world, the rules governing it and the parameters for the player to interact in grammatical sentences rather than in complex code:

The programme parses your sentences in order to construct the game, much as the game will ultimately parse the player’s input. What is extraordinary is that it does so by a degree of inference, so that a simple subject-verb-object sentence will be parsed to generate characters, things and relations. Of course, there are still relatively complex protocols to learn in order for your instructions to be understood, clustered ‘if’ clauses and so on – and I’m still at the most remedial end of the spectrum, sadly. I’m certainly not yet ready to tackle, say, liquids or rope (which are apparently among the hardest objects to account for in I.F.).
IARPA are also unlikely to be calling on Nelson. As well as the limits of Inform 7’s natural language parsing, both for the programmer and player, others have raised questions about the whole model of language and literature implied in I.F. Sarah Sloane, in her book Digital Fictions: Storytelling in a Material World (2000), complained about the “Objectivist epistemology” she believed structured most I.F., arguing, for example, that the Infocom games labour under the “limitation of seeing language as transparent, neutral, and, in essence, non-rhetorical; language is just another set of symbols that can be matched easily to a pre-existing, singular world.” The player, she suggests, is always in the straitjacket of an ‘ideal reader’, without the capacity for genuinely creative perception or, in fact, access to the metaphorical resources of language. IARPA’s instrumentalised vision of a parser which could process the whole of a language, and map rhetorical meaning beyond even speaker’s intentions, might in some ways represent the fulfilment of the I.F. community’s dreams. But at the same time we might wonder if even a more metaphorically-attuned form of I.F. would be able to create a real reciprocity between language and world, a full-bodied interactivity in which the player’s language helped to shape their environment rather than simply providing access to it.
Also still unresolved is the thorny question of I.F.’s relation to less adventurous texts, sometimes known as literature. Literary writers have certainly dabbled in I.F., notably U.S. poet laureate Robert Pinsky with Mindwheel and Thomas M. Disch with Amnesia (according to Nelson his initial enthusiasm was followed “by total disillusion when it was not marketed and received as a novel might be”). Such authors have had to adapt to the constraints of I.F., whether of format (the early games could only contain about as much text as a 30 page novella) or expectation (Pinsky balked at the number of puzzles he was expected to inject). One interesting related question has been the value or otherwise of game transcripts – Pinsky, indeed, published a session from Mindwheel, but in Twisty Little Passages Montfort argues that, “it is the effect of the narrative in the process of being generated that is important, after all, not the quality of the text that is output when the session is over, and not the effect of any post hoc reading of that output text.” Seth Price perhaps found the ideal medium for a transcript that is true to the process with his video Romance (2003), which documents a session of Adventure.
At this point, I’m keen to examine these questions from the inside out. Once, hopefully, I have got to grips with Inform, I'm going to work with the artist and writer Ed Atkins on a game for the Annual I.F. competition in the autumn. We're still deciding on our approach but we may ultimately attempt a literary adaptation (Trakl?), despite Graham Nelson's warnings about the pitfalls ("If the player decides to have tea outside and not to go into those ancient caves after all, the result is not A Passage to India"). In any case: we'll post any resulting game here too.