The research proposes a deterministic framework for compiling VGDL games into SCSMs. This methodology translates game components – sprite dynamics, interaction rules, and termination conditions – into explicit structural equations. Each game tick represents a causal transition between state variables. The resulting SCSMs offer transparent causal pathways. This framework provides a direct mapping from symbolic game descriptions to causally grounded game AI, addressing issues of spurious correlations and hallucination prevalent in standard reinforcement learning and large language model approaches.
Source: https://arxiv.org/abs/2609.05459