Figure 1. A screenshot taken from The Sims (source: Wikipedia.org)

Figure 2. Screenshot taken from Star Wars: Knights of the Old Republic. (source: LucasArts.com)

Figure 3. User interacting with Marie in Portrait One. (source: http://www.din.umontreal.ca/courchesne/)

The experimental game that is being used to explore generative gameplay elements on narrative is what can be considered to be a “conversational role playing game” (or RPG for short).  The game world consists of a bar that has a set of patrons.  The player is asked to create a character by selecting a set of personality traits that govern what the character is able to say in a conversation.  When the player is finished creating his or her character, the game randomly generates the bar patrons, giving them unique traits of their own.  The object of the game is for the player to find the love of his or her life before the bar closes.  The player has the opportunity to find his or her match by engaging each of the patrons in “small talk”. 

When the player has found the patron that seems most suitable to his or her character, the player can engage that patron in a full conversation.  The conversation consists of 3 possible statements that the player can choose from.  Each is chosen by the game based upon the traits of the character with which the player is playing.  Once a statement has been chosen by the player, the patron will respond to the player with a statement based upon an evaluation of the player’s statement against his or her own traits.  A full round of conversation contains 4 of these statement and response pairs.  As each round occurs, the player will get feedback from the patron about how successful the player is at achieving his or her affections.  The game ends after six rounds of full conversation.  At this point, the game will evaluate the character that has been most affected by the player and will display an epilogue involving the player’s character and the patron based on how well or poorly the evening went.

Figure 4. Player can click on 3 statements on the bottom of the screen, with the patron’s response appearing at the top area of the screen.

            In an effort to make the game as accessible and appealing as possible to everyone, the interface for this game is simply a mouse and  one button.  Since the game will be made in Macromedia Flash, the graphics will be 2-dimensional with a stylized animated look with people represented as silhouettes.

When the game starts, the player will be given the option of choosing from pre-made characters with traits already assigned to them, or creating his or her own character.  If the player chooses to create a character, he or she will be given a list of 18 personality traits and will be asked to pick 5 of them.  Each trait is paired up with another so that there are 9 pairs in all.

            When the player has his or her character created, the game will randomly generate the rest of the characters, giving them their own 5 personality traits.  When the game itself begins, the player’s character will appear at the doorway and there will be bar patrons placed randomly alone and in groups all along the bar counter and seated at some of the tables in the room.  In order for the player to interact with any of the characters, all he or she has to do is click on that particular person.  The view will change, giving the player the opportunity to click on one of a set of options below the person in question in order to interact with him or her.

To the left of the bar counter will be a band that will be playing the music that the player will hear in the background as the game progresses.  The player will have the ability to walk over to the band and select one of a set of songs.  This can have an effect on a conversation depending on the type of music that is being played and how it corresponds to a particular character’s personality traits.  Additionally, the player will be able to go over to the bartender and order drinks, which in turn can be used to affect, either positively or negatively, the outcome of a conversation.  Unfortunately due to time and budgetary concerns, this feature did not make it into the final game.

            The player will also be able to engage the bar patrons in some form of dialogue.  When the player clicks on one of the patrons, his or her character will walk up to that patron and a window will pop up giving the player the choice of engaging him or her in either “Small Talk” or “Full Conversation”.  If the player selects “Small Talk”, the patron will give the player a one-line statement about him or herself or one of the other patrons present at the bar.  The player can do this to learn as much as possible about the patrons at the bar without sacrificing any rounds of conversation.  This will help the player pick the patron that either seems the most compatible or the most interesting.

If the player selects “Full Conversation”, the view will change so that the player can see him or her in a close-up camera angle.  There will be three possible statements that the player can select from below the patron’s face.  From here, the player will go through 4 statement/response pairs that will make up one full round of conversation.  Depending on how the patron likes or dislikes the statement, the player will know if he or she is being successful based upon the dialogue response and a score represented by a meter on the left-hand side of the screen.

The statements that are picked for both the player and the patrons come from an extremely large database of statements.  Each statement is actually a movie quote taken from a long list of movies on the Internet Movie Database (IMDB) [12].  Of course, each movie that is quoted is properly credited in the game.  Additionally, each of these statements has various traits associated with it.  The first trait associated with a statement is a topic keyword.  These keywords are manually assigned to each based upon a potential subject that the statement relates to.  The keywords are used by the game to maintain topic consistency from one statement/response pair and conversation round to the next.  Despite this, it is possible for the patron or the player to randomly change the subject.  This is to ensure that a conversation isn’t stuck on one topic for the entire game. 

Figure 6. A small portion of the statement database.

The next set of attributes assigned to each statement is whether it can be used in the beginning, middle, or end of a conversation round.  Some statements can only belong in one phase of a conversation.  For instance, “By the way, what's your name?” can only be appropriately used at the beginning of a conversation.  Whereas, “Dry your eyes, baby.  It's out of character.” can be used either in the middle or at the end of a conversation. 

Another attribute assigned to each statement is a response value.  The response value indicates that the statement is either positive or negative.  This is done to ensure that the player will get a statement that fits the positive or negative score attained by the statement that was chosen.

The last set of attributes assigned to each statement is the full list of personality traits that can potentially be assigned to a character.  This is done to check to see if a particular statement matches the personality of a character who could potentially say it.  So for example, it would be confusing to have a character that is associated with the Warm personality trait to say, “We are protected by the enormity of your stupidity, for a time.”  Whereas, it would make sense for such a person to say, “It seems to me you could make some girl wonderfully happy.”

The game will evaluate the progress of the player’s character by giving each patron a score.  All patrons will start off with a score of 0.  A patron’s score can only change if the player engages him or her in a full conversation.  If the patron thinks upon the player’s character favorably, he or she will have a positive score.  If the player’s character messes up when conversing with a patron, the patron’s negative opinion will result in having a negative score. 

The process of evaluating the player’s score is done by first taking the statement that the player clicked on and comparing its attributes to those of the patron.  This is done by first calculating a Success Ratio.  This ratio starts off at 50 (out of a possible value of 100).  For each trait (up to 5) that matches the statement’s traits, the ratio increases by a value of 10.  For each of the statement’s traits that opposes the patron’s traits, the ratio is reduced by a value of 10.  Next there is a virtual 100-sided die that is rolled.  If the number that is randomly generated is at or below the Success Ratio, the player is rewarded +1 to his or her score relative to the patron.  If the number that is randomly generated is above the Success Ratio, the player is given -1 to his or her score.

After 6 rounds of conversation are completed and the bar officially closes, the patron with the score that is farthest from 0 in either the positive or negative direction will appear, saying a statement of either extreme love or extreme rejection, depending on whether the score is positive or negative.  In the case of a tie, the game will provide an epilogue based on only characters with which the player has conversed.

5. EFFECT OF GENERATIVE ELEMENTS ON NARRATIVE

It would seem that the narratives generated by Barfly are rather static in nature. This could possibly be due to the strict limitations imposed on the player’s character and bar patrons. There are only allowed 6 rounds of conversation in which a character has the given opportunity to change. And in this sense, the only character(s) allowed to change is the bar patron. It is accepted by most writers that a strong narrative has dynamic characters that have motivations and grow as people. While the motivation of the player’s character is clear, it is not clear on the part of the bar patron. Additionally, the characters themselves are all static. Other than how a bar patron perceives the player’s character, none of the characters display any growth as people. It is not immediately clear as to how to allow such growth to manifest itself in this experiment. One possibility is to allow characters to gain new traits and lose others. Another possibility is to give the characters motivations and fears, much like those seen in The Sims. When used in conjunction with the personality traits, this could give the user a better sense of the people he or she is interacting with. It will also give the player a goal of perhaps ridding a character of a particular fear. It can also allow the player to attempt to satisfy a particular motivational need. A good narrative allows the player to have a direct emotional investment in the events and the characters that play a role them. While the player will probably care about what happens to the characters in Barfly, he or she most likely will not have a direct emotional investment in the characters themselves as people.

Another difficult thing to keep under control in a generative game such as this one is the events as they make up the narrative arc. Barfly makes no real attempt at making sure that the story that is generated by its events contribute to the maintenance of a well-rounded arc. The best that it can do (in its current state) is keep conversations topical. In some respects, it will be up to the player to insinuate his or her own arc upon the events as they unfold. This practice is considered questionable by many as it requires the player to make up the story as the events in the game take place. This is the approach that The Sims takes. While many people find the stories they can tell with The Sims entertaining, it is questionable if such stories are of the same caliber as those written by professional writers who are skilled in that particular art form.

Façade takes a more direct approach toward keeping the dialogue and events that occur within a tightly woven narrative arc. The conversation that the player can have with its characters will always stay within the exact same topic, namely the relationship between Grace and Trip. The only possible changes that occur in their characters are whether or not they will hate each other or go back to loving each other at the end of the story. While this works on the level of experience a lovers’ quarrel, it does not show the depth of change necessary for a compelling narrative. It seems, however, that Barfly could expand upon this by allowing more variance in how the player can affect bar patrons, yet maintain a more tight-knit narrative structure by guiding the player into what the next event would most likely be. Unfortunately, this runs the risk of making the player feel that he or she lacks control over the outcome of the events in the game, which is what happened to the player in Façade.

Figure 7. A screenshot taken from Façade. (source: interactivestory.net)

            Notwithstanding the shortcomings of Barfly, people who have played it seem to have thoroughly enjoyed themselves.  Their encounters, though brief, appear to be memorable.  People like the movie lines and they feel that they have some control over the outcome of their characters.

Through the course of completing this project, I have noted that generative gameplay elements make it more challenging for a game designer to maintain a consistent and compelling narrative.  This is due to the unpredictability of the possible events that can take place due to such elements.  Design is further made complicated for games in which characters are themselves, generative.  This makes it unclear to the designer how such characters can have motivations and personality traits that grow and change over the course of the game.  Despite these difficulties, a story that is generated can be made compelling by the specific limitations enforced upon the characters and the events that take place by the designer.  It would seem that the true narrative and message communicated to the player within a game built upon generative gameplay elements comes from the designer’s rules and boundaries that define the game itself.

ACKNOWLEDGEMENTS

First I would like to thank my thesis advisory committee consisting of Peter Brinson, Erik Loyer, and Andreas Kratky.  Their experience and wisdom have helped guide me through the process of undertaking this project.  I would also like to thank Mark Bolas and Michael Naimark.  These two professors cracked the whip and made sure I stayed on time and focused.  Lastly, I would like to thank Phillip Gregorchuk and Uel “Jack” McMahon.  Both of these students have sacrificed much of their free time to help me get this project off of the ground.  I could not have done it without them.

Other people who contributed include:
Nayan Savla
Mariantonieta Madrid
Emmanuel J. Mitsinikos
Michael Rodgers
Jamie Green
Robert Livingston
John Brennan
Kyle Fujita