Cognitive architecture

1. Gray et al., The Soft Constraints Hypothesis: A Rational Analysis Approach to Resource Allocation for Interactive Behavior

2. Cognitive architecture A cognitive architecture is a computational system that specifies a particular way of representing information and a fixed set of cognitive mechanisms for processing that information. (Lovett, Daily, and Reder, 2000, p. 101) Basic resources and operations that constitute the cognitive system.

3. Cognitive processing is significantly affected by the specific task environment. Gray et al. set out to describe “a functional mechanism that adjusts the mixture of low-level cognitive, perceptual, and motor resources to produce the observed higher-level changes in behavior” (p. 462).

4. Memory costs v. temporal costs Minimum memory hypothesis (MMH): Other things being equal, the cognitive system favors external storage over internal storage. Temporal cost hypothesis: “When the expected utility (i.e., the cost-benefit tradeoff) of alternative interactive behaviors can be quantified in terms of time, those that minimize milliseconds are selected over those that minimize the use of cognitive resources” (p. 462).

5. The soft constraints hypothesis The soft-constraints hypothesis: “at the 1/3 to 3 sec level of analysis, the control system selects sequences of interactive routines that tend to minimize performance costs measured in time while achieving expected benefits. Cost-benefit considerations provide a soft constraint on selection as they may be overridden by factors such as training or by deliberately adopted top-down strategies” (p. 463).

6. Criticism of Ballard et al.’s indirect evidence for the MMH Ballard et al. say that subjects, suprisingly, use external memory when internal would save time (3 sec v. 1.5); but they get their estimate for the use of the internal strategy (1.5 sec.) from later in the trials, when subjects frequently use internal memory. Those estimates do not take into consideration the possibility that earlier in the trials, time was spent memorizing the pattern of blocks.

7. Contrast with Gray and Fu’s earlier results When programming VCR’s, subjects who could, with some effort, get perfect information from the world often relied on much faster access to imperfect information in the head. Does not seem to gibe with MMH.

8. Elements common to the three experiments Blocks-world task, similar to Ballard et al.’s design Counted the mean number of times during a trial that the target window was uncovered Tracked the length of first look Counted blocks correctly placed following the first look

9. Experiment 1: Design Three access conditions to target window: --e1-low: ctrl key opens target window --e1-med: opens when cursor enters window --e1-high: time lag of 1 sec from time cursor enters window

10. Experiment 1: Results Number of target window access-events: e1-low and e1-med did not differ, but each significantly higher than e1-high Duration of first-looking event: Same pattern as above (e1-low and e1-med have shorter looking-times; each differs significantly from e1-high) Blocks placed correctly after first look: Same pattern as above (more blocks placed correctly after e1-high)

11. Experiment 2: Design Uses a variably sized radio button in the target window --e2-low: 260x260 pixels --e2-med: 60x60 --e2-high: 8x8

12. Experiment 2: Results Number of target window access-events: e2-medium and e2-high did not differ, but e2-low differed significantly from each of the others (e2-low has greater number of events) Duration of first-looking event: same pattern as above (shorter for e2-low) Blocks placed correctly after first look: all pairwise comparisons differ significantly

13. Experiment 3: Design There is a delay between the cursor’s entering the occlusion area and the appearance of the information Experimental manipulation: Between subject variations in lockout times (0, 200, 400, 800, 1600, and 3200 ms)

14. Experiment 3: Results Number of target window access-events: significant (inverse) linear correlation between length of lockout time and number of access-events Duration of first-looking event: same pattern as above, but not inverse Blocks placed correctly after first look: same pattern as above, but not inverse

15. Conclusion regarding situated cognition (p. 478) “we off-load cognitive work onto the environment. For this claim the soft constraints hypothesis implies that the control system is indifferent to information source; resources are allocated to knowledge in-the-world versus in-the-head not based on source, but based on the cost of accessing the source.” “the environment is part of the cognitive system. The soft constraints hypothesis offers the same comment on this claim as to the first—that the human information processing system is indifferent to the source of its information.”

16. Impartiality and Extended Cognition P1. The human cognitive system does not treat the distinction between organismically internal and external storage locations as in itself relevant to the choice of routines to use to solve a problem. P2. The decision to use any resource beyond the boundary of the cognitive system should be driven by explicit consideration of the external location of that resource. Therefore, some external resources are part of the human cognitive system.

17. Why accept P2? If we reject P2, how else can we define the cognitive system? In terms of the persisting architecture?