Game BotMasters
Published by Christy Dena April 9th, 2005 in Uncategorized, Poetics.
After discussing the birth of a Bot-God I am drawn back to the beginning of all creation with the birth of a bot-baby. Eccky is the brainchild of entertainment company Media Republic. As I’ve spoken about before MR have produced webisode drama Jong-Zuid and Real Axe Babes.
Eccky is a whole other level of entertainment though, with only virtual actors on screen. Just like Buddhabot though, the process of creating a bot personality — of botmastering — is repackaged as a game in itself.
The bot, Eccky, is born to two ‘parents’ who can chat with their rapidly growing child through MSN Messenger and SMS. The look and characteristics of the child are based on the unique DNA of its parents: data from a quiz. The baby evolves to an 18 year-old adult in 18 days so contact is pivotal and imperative. And then at 18 it leaves home…
The botmasters, or parents, are given instructions on how to develop their bot, or child. This is depicted ‘in-game’ or framed within the story-world rather than the (meta) tech-speak that bot masters are faced with. For example:
A pattern is an element whose content is a mixed pattern expression. Exactly one pattern must appear in each category. The pattern must always be the first child element of the category. A Pattern does not have any attributes. [source: Artificial Intelligence Markup Language (AIML) Version 1.0.1]
Instead, the Eccky parents have an interface that persists the ‘game’:
There is a difference, however, between the bot creation on the creators side and that of the players of Eccky. The designers have spent alot of time, money and expertise developing a bot that is a growing child; a bot that reacts to player input and challenges them ‘in-game’. They are using a proprietary Eccky Language Maintenance Tool to edit conversations and metadata, not unlike the AIML editors used with Alicebot software.
They have 5 copywriters working on 70,000 conversations. The designers have made decisions about the characteristics of the growing child, how they will speak, what issues the parents will face, how the child will change over time and according to inputs. For instance, as Eccky goes into the teenage years it may ‘be a person who boldly presents himself, his views, and most certainly his desires’. And to ensure they have the right speech patterns the designers enlisted the assistance of an online English-to-12-Year-Old-AOLer Translator. They’ve discussed what it would look like, how a child will react and how emotions will be expressed.
In particular, they have ensured that each bot is somewhat unique for the parents (consumers).
Eccky doesn’t have a static personality, it’s dynamic and based on approximately 170 variables, thereby guaranteeing that every Eccky is unique from an end user perspective.
The game also has in-game multiplayer games that can be played between the parents and their Eccky.
The Eccky team compare their creation with Artificial Life’s Virtual Girlfriend, which has been described as ‘tamagotchi with tits’
Unlike Virtual Girlfriend, Eccky will not act as a personal assistant. Obviously, this would shatter the game-world.
The designers claim that the chatbot is perhaps the most ambitious in the world at the moment since they are trying to create natural language conversations with a child. They don’t think the lessons of ALICE will be of help because it is too ‘general purpose’ (I recommend they should look at the individual developments). The bot will be rolled out internationally and is currently being translated into many different languages. The translation though is not just the wording but the approach to parenting and colloquisms.
They are employing viral, pervasive marketing techniques: aiming to be discovered by their core target group. To be discovered and facilitate anticipation they are making their production public with a blog. They have 3 sources of revenue: buying an Eccky (priced between 1 and 1,5 Euro); in-game wallet to buy their Eccky clothes and sweets; in-game advertising.
Obviously, Eccky is not the first to make a game, and money, out of botmastering: networked Tamagotchi, Neopets, Creatures, and Black & White . But is there a market/player preference for botmaster type games to other forms of bot storytelling? Indeed, what are the options for storytelling with bots? Here are some provisional categories:
- Botmaster games (as cited above)
- Bots as fictional characters you talk to (a story emerges out of interaction)
- Bots as in-game AI (either background or interactive bots)
- Bots as characters in an ‘interactive drama’ (eg: Facade: although they are more agents than bots…)
- …
And then there are bots that assist the experience of the game or story but do not play an ‘in-game’ or story role:
- Bots as game assistants (eg: Anarachy Guild bots)
- …
Any suggestions and ideas?
Link to Eccky blog sourced from: crossmedia communication
This is an interesting taxonomical question.
Also, re: Bot and stories.
How about Bots that tell stories, like Racter.
Although, I wonder if Bots as fictional characters leaves out a set of bots that are related to narrative encounters, but not necessarily dramatic.
How would Agent Ruby fit into this schema. Her slogan calls her a Bot that can teach you to dream? It seems like the fictional character, but different somehow.
Also, how does the Oracle of Glide fit into or complicate all of this?
This is Bot as prognosticating linguistic interface. (Perhaps I’m stretching our bot category too far here). Bot as langue (in a semiotic sense)?
Hi Christy,
I’m interested in the distinction you draw between “bots” and “agents” (as you describe Facade’s characters as being “more agents than bots): do Bot and Agent have disjunct parent classes in your ontology?
To me, this is a crucial consideration; I believe that a botmaster’s decisions on how to handle that part of a bot’s ontology has a great influence on how a bot gets written, and what kind of a relationship it can have with the client.
The way I handle this for the bot I currently work is that I define the Bot class as a subclass of Character, which itself is a subclass of Agent. Because Client is also a subclass of Character, I (or eventually, my bot) get(s) to handle clients as objects/subjects that are on the same ontological level as himself (it’s a male bot, BTW :-). With this arrangement, it’s relatively easy to write traits like “self-consciousness” and “self-respect” into the bot character; a bot that communicates his belief as being worth just as much as any other character.
Choosing different ontologies leads to the creation of different characters, because the relationships between concepts afford the character traits that the botmaster can develop. Which are the character traits that are afforded by treating Bot and Agent as disjunct? Where does the Client class fit in?
Thankyou Dirk for picking up on my lax use of ‘agent’. You’ve brought up an important point that needs to be addressed at the relative beginning of this blog (though not solved immediately I suspect). Indeed, a bot is an agent — as we’ve set out in the Bot Blog description:
So, I should clarify therefore why I felt the need to differentiate (ignoring now my inappropriate terminology). I see Facade as different because as far as I know it is not a simple-reflex agent. From the descriptions I’ve read it appears to be a ‘model-based reflex agent’ or a ‘goal-based agent’. I don’t want to start debating the technical aspects of Facade here since I am not familiar with the workings of the architecture. For now, here is one working paper [pdf] which outlines some details; and of course there are plenty of discussions at GxTA. What is important here is my distinction: that there is a difference between agent types. But, as your query so rightly highlighted — what role does the difference play? On reflection, it doesn’t seem to play any role other than being a technical category.
There are 2 reasons why I had the difference as being important:
Ironically, the second does not seem a factor. There are, I believe, many more simple-reflex agents being used for storytelling than other types of agents. Obviously a major factor in this scenario is accessibility. So, goal (1) can be achieved even with the inclusion of agents other than simple-reflex agents. The next stage then is to continue with our categories of bots and story types.
Mark, yes, I like the ones you’ve added. There are a few different ways that bots are being utilised in storytelling. So the question goes back to how should/do we want to categorise bots and story?
Either one or a mix of all of the above?
There are some bots that aren’t really fictions but can be experienced as an entertainment of some form (this is probably going to be a discussion like that of differentiating between games and narrative?!). I like Marie-Laure Ryan’s combinations of the 3 features of literature: literary, narrative + fiction.
So, subserviant chicken would go under (5); Ractor’s output would go under (1) but the bot would be further categorised under supplementary function perhaps (shared role with human?); Agent Ruby is a character from a film so it would be perhaps ’supplementary’ or in a ’satellite’ node (in terms of multi-platforms) and maybe 3,6 or 8…
But these are all more general categories. We also need specific sub-types of botfiction or bots and story as I started before.
That’s all from me for now, I’ve gone in circles. Over to you Dirk and Mark.
Hi Christy,
Thanks for your Eccky analysis… If you have any questions, let me know.
Yme
Business Manager Eccky
What if we imagined tamagotchi-style botmastering (”raising”) within a graphical environment as the grey edge between game and authorware?
I realize that is somewhat perverse - most of the Eccky users are going to think of it as a game, the moreso if their creations “leave home” at the end. Still, you are raising / creating a potential bot, out of millions of possible configurations, by using simple controls to excercise choices.
So, if we were to draw new lines, lumping botmastering in with authorware, then a (very rough) analogy to IF:
AIML : Eccky :: Inform : Adrift
The analogy doesn’t preserve the division between authors and player/readers, but it does emphasize different classes of participants who want either a lower barrier to entry (within more predefined structure) or more control (with a high barrier to entry).
And they *are* classes - I don’t know how it is in the botworld, but there have been some very strong feelings across the Inform-TADS / Adrift fence, an argument of the “talentless hacks vs. pompus elitists” variety that I also remember in HTML authoring in the mid-90s when the first HTML GUI authoring environments were coming out to mixed excitement and rasberries….
I agree Jeremy, botmastering as game can be ‘lumped’ with authorware — but on a continuum. I think there is a big difference between Dreamweaver and Aurora Neverwinter Toolset. The later creates a storyworld within which users/players/designers/botmasters can play. Obviously the designer can subvert the environment — that is the desire of many — but the choices of equipment and functionality are pretty much pre-defined. So too with botmastering as game, the rules of the world and mode of interaction are pre-defined. We could then tip-toe a bit further along the authorware continuum and see all botmastering as having a prescribed mode of interaction. I think the level of freedom is greater in authorware than game botmastering. But we should always remember that in the end the level of invention is reliant on the individual…whether they have a game, authorware or pure code environment as tool…
Christy,
I agree with you that invention springs from the individual - and a continuum sounds like a very good model for authorware-gameware.
I’d like to complicate your idea of the “level of freedom,” however:
On the authorware end of the spectrum is the “freedom to design” with greater choice and less constraint. Many poets might argue that less constraint is not necc. desireable, but there it is freedom nonetheless.
On the gameware end of the spectrum is the “freedom to participate” with greater access - literally, you are not required to read a manual or spend 20-100 hours of prep and debug time. You may freely enter.
Perhaps I’m framing this in part due to my exposure to Free Software Foundation folks and Richard Stallman’s “Free as in Freedom / Free as in Beer” distinction. But while the freedom on the authorware end of the spectrum is rights / abilities, and the freedom on the gameware end is in some sense economic (literally, participation has a low “opportunity cost”) it strikes me that these both have importance in this context.
More generally, perhaps software like Dreamweaver and projects like Mystery House Taken Over are interesting because they are dual-mode - attempting to cover more of the spectrum, giving us easy access to rip-mix-burn while also appealing to our Freedom To Tinker.
ik wil eccky spelen.
Sorry Quincy, I don’t know what this means. What language is this, anyone, so that I can use a translator?
Christy - it is Dutch: “I want to play eccky.” A child passing through, perhaps…?