“…The most widelyused cognitive architectures, ACT-R (Anderson, 2007) and Soar (Laird & Congdon, 2009), were originally designed to model problem solving and symbolic processing rather than adaptive mechanisms. It might thus appear more natural to use cognitive architectures that are designed to support implicit knowledge, for example CLARION (Sun, Peterson, & Merrill, 1999), MicroPsi (Bach, 2008), CLA (Chaput, 2004), ICARUS (Langley & Choi, 2006), or ADAPT (Benjamin, Lyons, & Lonsdale, 2004). These later architectures, however, already implement specific learning algorithms, and we found that only Soar offered enough flexibility for our research on a new learning algorithm.…”