Choose your own manual annotation

My Team's implementation (Le Chiffre) concentrated on a gamified story telling mechanism using a "Choose your own adventure" structure. The game's objective was to get the player to follow and essentially "complete" a story which adheres to the "Hero's journey" template. The player is introduced to four different archetypes derived from Jung's theory: the Mentor, the Trickster, the Helper and the Shadow. The player is introduced to the mentor first, and is encouraged to interact with the mentor with the goal of eventually becoming inspired by them. By deriving various scenarios between the player and the current archetype from the action triples in the Scaletrix dataset, the player could choose from a number of actions in order to respond to the current interaction with the archetype. When the player reached a pivotal moment through choosing from a set of specific actions, the natural follow on could occur and the player would then be introduced to the next archetype. For example, when interacting with the Mentor, if the player chose an action which led to the protagonist being inspired by the mentor, the commencement of the journey becomes a logical follow on and the player is introduced to the next archetype. The ultimate goal and also the conclusion to the story is for the player to arrive at a situation where the Shadow is defeated (this could be a number of actions - the Shadow becoming imprisoned, or dying etc.)

The advantage of using such a system where the player chooses the actions which bring about the story are two-fold. Firstly, the way in which the game was implemented allows for both a human player and a computer to choose from the possible set of actions. Therefore, by allowing a computer to "play" the game, either through random decisions or indeed using a more informed approach (perhaps by harnessing expert systems?), a rudimentary form of entirely computational story telling is achieved.

However, the second advantage, and one I would argue has much more potential, is the possibility of recording and saving the decisions made by human players as the game unfolds. By taking advantage of the tremendous power behind crowd sourcing and collective intelligence, a huge database of human made decisions can be collected. These different decisions reveal much about story telling and what pleases the human mind when consuming narrative fiction. The frequencies of decisions made with respect to previous actions undertaken for example gives us a very powerful way of contextualizing the action triples: if the majority of people choose one action when having followed the same "path" that lead to this decision being made, then it is more than likely rational and coherent.

The system can also provide ways of filtering specific action triples with regards to archetypes. By analysing the frequency of different actions undertaken by human players when interacting with a specific archetype, the likelihood of what fits or makes sense when describing these interactions can be derived computationally. For instance, if very few players chose to "love" the Shadow archetype, this action will be chosen much less frequently in an automated story telling environment.

Essentially, by saving the decisions human players make when playing the game, a combination of manual and automated "coherency rank" is achieved. By manual/automated, I mean that through gamifying the story telling aspect, a "manual" coherency or contextual rank annotation is actually just the by-product or side-effect of game play. The task of saving the decisions made and their associated frequencies is trivial, but if enough data is collected, the potential uses are numerous and exciting.

Scholarly Lite is a free theme, contributed to the Drupal Community by More than Themes.