The creative endeavour of a story-based game is about communicating. It is the act of storytelling that communicates the experience and underlying narrative to the player. The quality of that experience, however, is subjective. How do we align a player’s expectations of what they believe is a “good story” in a video game? Good storytellers draw in the audience by going through a continuous loop of observation and orientation that adjusts the delivery of the story. At Evodant that system is the Toska Engine, and the process is adaptive storytelling.

How far can a system such as this take a virtual story experience? Our goal with Toska is to deliver a personalized story within a game world that aligns with someone’s narrative expectations. It does this by running through the following (high-level) loop:

  • Observe player and world states
  • Factor in current narrative plan
  • Project possible narrative paths that align with player psychology and character action
  • Factor in non-player character (NPC) goals and behaviours
  • Adjust narrative plan and NPC goals
  • Present the narrative

This is a generalized outline, with the story adapting to a continuous change in the world state and character actions. At a deeper, implicit level causation weaves throughout this narrative model.

Aristotle presented “efficient” and “final” causes. Without being too academic, “efficient cause” roughly means what we normally think of when X causes Y. We perceive some reasonable sequence of events, or story, connecting X and Y. Such as, “My son picked up a ball and threw it towards me, so I caught the ball.” If we also apply a meta- context for this sequence (playing catch) then the story is more easily parsed by the audience and it may also elicit emotions depending on the individual’s own experiences.

“Final cause” is the idea of fate. Using our example, my son throwing the ball would always happen because I catch it. While this is primarily a physical example, there is also a psychological version of final cause. It is the justification for why something happened. Again, using this example, “Why did my son want to play catch?—Because my son loves throwing—or loves spending time with me—or…”, the reasons are as many as one can reasonably justify, and typically they are reasons with a high degree of personal connection.

When you watch a film or TV show, these fixed narratives are a form of final cause in the telling of the story. The end was always going to be the end it was because of everything that happened before and it could be no other way. That’s the script.

Note that with causation, especially under final causes, we can make judgments with incomplete knowledge. Something we think is likely given a series of facts, beliefs or desires.

So how does this relate to adaptive storytelling and Toska? Without a pre-scripted story and its subsequent ending, the player’s narrative is free from a pre-defined resolution. We focus on the “efficient” form of the storytelling. The narrative plan does, however, serve the function of “final cause” validator in some respects. To find what a player would experience as a good story, we need to allow the system to test paths that deliver a satisfying resolution.

This “polynarrative space” is where Toska lives. These are just some ways that the artificial intelligence in the hybrid framework work to deliver an adaptive story in real-time that morphs along with the player.