(*** Insert Figure 4 ***) (*** Insert Figure 2 and 3 ***) (*** Insert Figure 1 ***) the Study of Learning Mechanisms in Unified Theories of Cognition 1

1. Behave flexibly as a function of the environment. 2. Exhibit adaptive (rational, goal-oriented) behavior. 3. Operate in real time. 4. Operate in a rich, complex, detailed environment. a. Perceive an immense amount of changing detail. b. Use vast amounts of knowledge. c. Control a motor system of many degrees of freedom. 5. Use symbols and abstractions. 6. Use language, both natural and artificial. 7. Learn from the environment. 8. Acquire capabilities through development. 9. Live autonomously within a social community. 10. Exhibit awareness and a sense of self. 11. Be realizable as a neural system. 12. Arise through evolution. Table 1 7 Captions Table 1: Multiple constraints that shape mind according to Newell (1991). Figure 4: Schematic representation of problem spaces for the Go-problem. Triangles represent problem spaces, while circles represent succesive states within a problem space. A line downwards to a new triangle means an impasse has occured, resulting in a subgoal with a new problem space. 6 legal moves, and rules that describe the consequences of a move. Also we may suppose there are some basic search mechanisms, implementing some sort of look-ahead search. We will call the problem space containing the basic Go-rules and mechanisms to implement look-ahead search Go-basic. Soar is capable of creating new problem spaces. However, these new problem spaces are initially empty. Empty problem spaces can be filled by chunking from subgoals. So the learning process runs as follows: 1. An instance of a Go problem is offered to Soar. 2. The Go-top problem-space is instantiated, which tries to find a suitable strategy. Since there are no strategies yet, an impasse is reached, resulting in the creation of a new, empty problem-space, called, say Go-S1. 3. Go-S1 is instantiated as a sub-goal of Go-top. Since there are no rules yet in this problem space, a No Change impasse is reached. As a sub-goal for this impasse Go-basic is instantiated. 4. Go-basic selects through look-ahead search some move, and copies this move to its super-goal: Go-S1. After this, Go-basic terminates, and we are back in Go-S1. At this point, a chunk is built in Go-S1, which summarizes the search in Go-basic. The next time Go-S1 will be invoked, it will not be empty, but will contain one rule. If this rule than applies, it will fire, else an impasse is reached, and Go-basic used to select a move. Figure 4 shows the layout …