Informing a BDI Player Model for an Interactive Narrative

Abstract

This work focuses on studying players behaviour in interactive narratives with the aim to simulate their choices. Besides sub-optimal player behaviour due to limited knowledge about the environment, the difference in each player's style and preferences represents a challenge when trying to make an intelligent system mimic their actions. Based on observations from players interactions with an extract from the interactive fiction Anchorhead, we created a player profile to guide the behaviour of a generic player model based on the BDI (Belief-Desire-Intention) model of agency. We evaluated our approach using qualitative and quantitative methods and found that the player profile can improve the performance of the BDI player model. However, we found that players self-assessment did not yield accurate data to populate their player profile under our current approach.

Exploring Apprenticeship Learning for Player Modelling in Interactive Narratives

“…Findings such as the drop in performance of the RHIRL model with a horizon of 5 are worth further investigation. Another goal is to develop a hybrid model using RHIRL and more declarative models like the Belief-Desire-Intention approach in our previous work to mitigate the learning difficulties due to the noise in player traces [20,19].…”

“…Using player traces collected in our previous work [20], our overall methodology consists of the following steps: 1)Convert player traces traces to demonstrations for RHIRL, 2)Extract policies resulting from RHIRL, 3)Execute policies in the Anchorhead engine. The Anchorhead engine is based on the work in [22], using their code with permission.…”

“…• By player profile as defined in our previous work [20]. We created two groups for each player profile factor, for low and high scores respectively.…”

“…Consists of four variables defined with the aim to differentiate between player styles in an IN [20]:…”

“…These policies were run on the Anchorhead engine. Following the same approach as in our previous work, we measured the similarity between traces using the Jaccard index of the sets of plot points discovered in each player log and those discovered in each RHIRL policy [20]. These trace comparisons were made only between the policy and all the player traces available in each group.…”