Inference methods play a critical role in cognitive architectures. They support high-level cognitive capabilities such as decision-making, problem-solving, and learning by transforming low-level observations of the environment into high-level, actionable knowledge. However, most modern infernece methods rely on a combination of extensive knowledge engineering, vast databases, and domain constraints to succeed. This work makes an initial effort at combining results from artificial intelligence and psychology into a more pragmatic and scalable computational reasoning system. Our approach uses a combination of first-order logic and plausibility-based uncertainty consistent with methods first described by Polya [3]. Importantly, concerns with optimality and provability are dropped in favor of guidance heuristics derived from the psychological literature. In particular, these heuristics implement cognitive biases such as primacy/recency [1], confirmation [2], and coherence [4]. The talk illustrates core ideas with examples and discusses the advantages of the approach with respect to cognitive systems.