The Interpretability of Codebooks in Model-Based Reinforcement Learning is Limited
This work addresses the problem of interpretability in deep reinforcement learning for operators, showing that a commonly suggested method is insufficient, which is incremental as it challenges existing assumptions.
The study investigated whether vector quantization methods provide interpretability in model-based reinforcement learning, finding that the codes are inconsistent, lack uniqueness, and have limited impact on concept disentanglement, which are necessary for interpretability.
Interpretability of deep reinforcement learning systems could assist operators with understanding how they interact with their environment. Vector quantization methods -- also called codebook methods -- discretize a neural network's latent space that is often suggested to yield emergent interpretability. We investigate whether vector quantization in fact provides interpretability in model-based reinforcement learning. Our experiments, conducted in the reinforcement learning environment Crafter, show that the codes of vector quantization models are inconsistent, have no guarantee of uniqueness, and have a limited impact on concept disentanglement, all of which are necessary traits for interpretability. We share insights on why vector quantization may be fundamentally insufficient for model interpretability.