MAC/FAC: A model of similarity-based retrieval

Curb Your Embodiment

“…For example, that a particular experience of a conversation with a friend happened in the park has little relevance to the concept of friendship . Still, literature on problem solving and other effortful cognitive tasks shows that even when a task requires processing abstract principles, people often rely on irrelevant surface features of the problem (Brooks, Norman, & Allen, ; Forbus, Gentner, & Law, ; Goldstone, ; Goldstone, Medin, & Gentner, ; Goldstone & Sakamoto, ; Goldstone & Son, ; Haryu, Imai, & Okada, ; Landy & Goldstone, ; Ross, ; Ross & Kennedy, ; Ross, Perkins, & Tenpenny, ). For example, when participants solved probability problems, their solution was influenced by irrelevant features of earlier problems, such as whether the problem was presented as the assignment of cars to mechanics or as the assignment of athletics teams to teachers.…”

Computational models of analogy

“…Structure‐mapping theory9,17–20 proposes that the following constraints govern the mapping process: Structural consistency : Structural consistency is defined by two constraints:o 1:1 constraint : Each item in the base maps to at most one item in the target, and vice‐versa.o Parallel connectivity : If a correspondence between two statements is included in a mapping, then so must correspondences between its arguments. Systematicity: Mappings that place systems of relations—especially those governed by higher order constraining relations—into correspondence are preferred. Tiered identicality: Identical matches between predicates and functions are preferred. By default, relations must match identically, but non‐identical functions can be aligned if such alignments would support a larger overlapping structure.…”

“…ARCS47 for example, used a two‐stage connectionist network, which first filtered candidate memory items in parallel and then used ACME to match the best. Likewise, when MAC/FAC20 has a case in working memory, it computes a simple feature vector from the structural representation and uses that in a parallel search of long‐term memory. It generates up to three candidate memory items.…”

“…Over the long term, gains in expertise appear to lead to changes in encoding,71 as well as to greater likelihood of relational retrieval 72. Forbus et al20 proposed that uniform relational encoding is a characteristic of expert encoding, and that this promotes relational retrieval. Finlayson and Winston73 argue that encodings that lead to intermediate‐sized representations provide more expert‐like retrieval.…”

Psychology in Cognitive Science: 1978-2038

“…Analogy research has had a long‐standing role in the cognitive science of problem‐solving and reasoning, but it is now also making contributions to accounts of learning in both children and adults. Work on analogy mostly continues the representational tradition of Cognitive Science; most accounts of analogical processing posit processes that operate over structured symbolic representations, although some processing may be assumed to be parallel (Forbus, Gentner & Law, 1995) or distributed (Hummel & Holyoak, 1997).…”