Abstraction promotes creative problem-solving in rhesus monkeys

Abstract

Abstraction allows us to discern regularities beyond the specific instances we encounter. It also promotes creative problem-solving by enabling us to consider unconventional problem solutions. However, the mechanisms by which this occurs are not well understood. Because it is often difficult to isolate human high-level cognitive processes, we utilized a nonhuman primate model, in which rhesus monkeys appear to use similar processes to consider an unconventional solution to the difficult reverse-reward problem: i.e., given the choice between a better and worse food option they must select the worse one to receive the better one. After solving this problem with only one specific example-one vs. four half-peanuts-three of four monkeys immediately transferred to novel cases: novel quantities, food items, non-food items, and to the choice between a larger, but inferior vegetable and a smaller, but superior food item (either grape or marshmallow), in which they selected the inferior vegetable to receive the superior option. Thus, we show that nonhuman animals have the capacity to comprehend abstract non-perceptual features, to infer them from one specific case, and to use them to override the natural preference to select the superior option. Critically, we also found that three monkeys had a large learning and performance advantage over the fourth monkey who showed less generalization from the original one and four half-peanuts. This difference suggests that abstraction promoted problem-solving via cascading activation from the two food item options to the relation between them, thus providing access to an initially nonapparent problem solution.

Neural representational geometries correlate with behavioral differences in monkeys and recurrent neural networks

“…Future studies on abstraction should be planned to test whether the task variables encoded in an abstract form, as opposed to those that are not, would facilitate the generalization of the rules to new items or conditions. The ability of generalization has been reported by several studies on macaques [37, 38, 39, 35]. For example, Falcone et al [39] have shown that monkeys can transfer the nonmatch-to-goal rule from the object domain to the spatial domain in a single session, and Sampson et el.…”

“…The ability of generalization has been reported by several studies on macaques [46, 47, 48, 44]. For example, Falcone et al [48] have shown that monkeys can transfer the nonmatch-to-goal rule from the object domain to the spatial domain in a single session, and Sampson et el.…”

“…For example, Falcone et al [39] have shown that monkeys can transfer the nonmatch-to-goal rule from the object domain to the spatial domain in a single session, and Sampson et el. [35] have shown that abstraction can allow generalizing to new conditions, such as new foods, of the rule to choose the worst between two options.…”

“…We know from machine learning studies on curriculum learning that artificial neural networks can solve the same task in different ways depending on the order of presentation of the samples and more generally, on the details of the learning process [31, 32]. Differences in strategies have been described in experimental studies, in particular in the information representations of the reward [33], in the strategies adopted by two monkeys to solve the same task [34], and in some abstraction tests [35]. A recent study, using a more complex task as the well know pac-man game, has even shown that different strategies can be flexibly switched based on different task demands [36].…”

“…Preliminary results on the RNNs indicate that an unbalanced distribution of trial types during the training phase can introduce biases in the resulting policies (data not included). Differences in strategies have been described in experimental studies, in particular in the information representations of the reward [42], in the strategies adopted by two monkeys to solve the same task [43], and in some abstraction tests [44]. A recent study, using a more complex task as the well know pac-man game, has even shown that different strategies can be flexibly switched based on different task demands [45].…”