Keyword Poetics

This short essay first published on 22nd July 2006 at WRT in response to Mark Marino’s post ‘Keyword: A Link By Any Other Name’ in which he asks: “What is the relationship between a link in a hypertext, a keyword in a chatbot, and an a noun/object in IF?”

Excellent topic Mark! It is an interesting question that does indeed slide along my cross-media poetics concerns, and so I’m keen to share some of my thoughts on the subject. As Adam recalled of Nick’s observation, IF (and botfiction) is two-way, it is synchronous communication. But there is a common thread through them all, as you’ve highlighted Mark: the hyperlink, noun and keyword. In the context of ergodic literature, these words share a function: they are all are an illocutionary act. ‘Illocutionary act’ is a linguistic term to describe an utterance that is trying to affect another; they are commands, suggestions, inquiries, vows and so on. If successful it is a ‘perlocutionary act’: ‘a speech act that produces an effect, intended or not, achieved in an addressee by a speaker’s utterance’. Examples of perlocutionary acts are persuading, convincing, scaring, insulting and ‘getting the addressee to do something’. In ergodic literature, the hyperlink, noun and keyword is the system (as representative of the human creator) trying to persuade the user to use a particular mode of interaction. They are basically saying to the user ‘this is how you can change me’.

I am far from a linguistics expert so everyone feel free to correct me, but here is an overview of how I think speech acts correlate in hypertext fiction, IF and Botfiction:

Hypertext Fiction:
Utterance: Rabbit
Locutionary Act = This link will take you to (further) information about rabbits
Illocutionary Act = Trying to persuade or command user to clink
Perlocutionary Act = Causing the user to clink

Utterance: The room is bare. You see a bird, a plane, and a rabbit.
Locutionary Act = Rabbit is a noun you can enter to move forward in the game
Illocutionary Act = Trying to persuade or command user to enter a noun, possibly ‘rabbit’.
Perlocutionary Act = Causing the user to input a verb with the noun ‘rabbit’

Utterance: That’s great. Have you thought about rabbits?
Locutionary Act = Rabbit is a keyword you can enter to continue the conversation
Illocutionary Act = Trying to persuade or command user to enter the keyword ‘rabbit’
Perlocutionary Act = Causing the user to input a statement with the keyword ‘rabbit’

Notice the relationship between the illocutionary and perlocutionary act? Before I go into how I think this applies, I need to draw the distinction between commands issued by the system and those by the user. Chris Chesher views hyperlinks as ‘avocations’ (they call the user to invoke) and the user’s response of clinking them, ‘invocations’ (they invoke). But, unlike human-human interaction, the illocutionary act in ergodic literature works best when both sides share the same representation of an illocutionary act. For Chesher, the URL ‘names the invocationary act, and performs it’, but so too does the noun in IF and the keyword and topic in botfiction. It appears that illocutionary acts (getting the user to do a particular thing) occur in ergodic literature when the illocutionary force of the system is delivered in the same manner the user can exert illocutionary force. That is, the system offers a hyperlink, verb, noun, keyword and topic as an illocutionary act to instruct the user as to how they can act to change the system. When the user then clinks, or enters a verb and noun or keyword it is the evidence that the system performed a perlocutionary act but also that the human is exerting illocutionary force onto the system. Successful ergodic works have utterances that are illocutionary and perlocutionary, the
cycle goes round and round.

Lots of big words repeated here. Perhaps I’ll just say: the system provides a key to the user that they use. Chesher observes that the ‘URL is the key to the invocation, as it is both comprehensible to humans and readable to the machine’. I like the term ‘command’ because it is at once a term used to describe an utterance with illocutionary force issued by a human to another human and to a program and it is the perlocutionary act. Indeed, the use of illocutionary force and the subsequent perlocutionary act in ergodic literature can be seen as a trope, even a command aesthetic.

Failed interactions:
Chesher says that a failed invocaton with hyperlinks can happen because of ‘a misspelt address; network overloading; technical misconfigurations’. But with hypertext fiction, IF and Bot interactions, the lack of illocutionary force is a reason for failure. They differ in their capacity to enact illocutionary force (command) and therefore succeed in a perlocutionary act (cause user action).

The IF command rests between a hyperlink and natural language input. The difference between IF and bots is not in the mode of input: both allow the input text, any text. The IF system responds to a narrow set of inputs and with a narrow set of responses. The bot response is designed to respond to any input because it is meant to be conversational (therefore implying free-will). In botfiction, no input is less acceptable. All inputs are meant to be addressed equally. Every input is meant to have a response that is preferably unique.
The supplying of information to the user as to how to perform an illocutionary act to cause a change in the system is seen by Chesher as a specific ritual: ‘New media artists prepare special vocabularies of invocations that visitors perform. Unlike useability engineers, they are designing specific, not generic interfaces’. (Invocation, Evocation and Avocation in New Media Art).

Unlike a failed hyperlink, a failure message in IF can assist in informing the user on how to perform an illocutionary act/invoke. For instance, take my interaction with the ZPlet by Matthew Russoto of Crowther and Woods’ Colossal Cave Adventure at Dennis G. Jertz’s site:

You are standing at the end of a road before a small brick building. Around you is a forest. A small stream flows out of the building and down a gully.
> do forest
That is not a verb I recognise.
> get forest
That is hardly portable.

This feedback helps the user to narrow the field to possible commands/invocations by letting the user know the interaction didn’t work and also gives clues as to what would be a successful input: obeying the laws of physics and using a particular set of verbs.

The bot can give feedback when there is a failed interaction in the form of offering a keyword or default to a statement that explains how it does not understand. But the bot, because it is meant to deal with EVERY input equally, does not provide clues as to particular commands or invocations. Instead, the system response is to SOMETIMES offer keywords or a particular topic for the user to then use, according to the design employed by the botmaster. Providing information about how the user can affect the system has not really been a convention in botfiction. But, one can see how the use of questions (an illocutionary act) to get the user to answer (a perlocutionary act) was the main reason for ELIZA’s success, in terms of interaction.

But anyway, I could keep burrowing this topic for days, I mean pages. Any responses?


Austin, J.L. (1962) How to Do Things with Words, Oxford University Press, Oxford.

Chesher, C. (2001) ‘Computers as Invocational Media‘ [PhD] School of Media and Communications, University of New South Wales, Sydney (chapter only)

Chesher, C. (2004) ‘Invocation, evocation and avocation in new media art’ presented at European Association for Studies of Science and Technology (EASST) / 4S conference, Paris, August, published by University of New South Wales

Chesher, C. (2004) ‘Hyperlink as Invocationary Act‘ presented at ANZCA (Australian & New Zealand Communication Association) Conference, Sydney, 7-9 July, published by University of Sydney

Searle, John. 1969.Speech acts: An essay in the philosophy of language. Cambridge, England: Cambridge University.

Note: Clicking on a link or ‘clinking’ is m.c. schraefel et al.’s wonderful portmanteau.