As we continue to look ahead toward the next generation of wireless technologies, the notion of intelligence has taken center stage. Artificial intelligence, and more specifically machine learning, has been proposed for use in nearly every aspect of wireless communication systems. Semantic communication is a recently-revived paradigm shift toward meaning-oriented communications that can benefit greatly from these intelligent technologies. One approach to semantic communication that leverages this intelligence is based on the theory of "conceptual spaces." In this work, we present two major innovations for semantic communications with conceptual spaces. First, we introduce the use of the variational autoencoder as well as the 𝑝-Wasserstein distance metric from optimal transport theory for learning the domains of the conceptual space model of semantics. Second, we present for the first time a mechanism for carrying out causal reasoning over semantic information modeled by a conceptual space with respect to some communication goal. We simulate the proposed methods and confirm their benefits as compared to some baseline communication techniques, demonstrating accuracy improvements of up to 40% on classification tasks and providing similar performance to a more traditional communication technique with a 99.9% reduction in rate. These innovations represent significant steps forward in the pursuit of truly intelligent semantic communication systems.