Two more ACT-R papers

[ERIK M. ALTMANN]

Are at hfac.gmu.edu/people/altmann/cogsci99.sea.  The first presents a
closed-form model of serial attention that integrates base-level and
associative activation, showing that both are necessary to achieve
empirical performance accuracy given noise in memory.

The second presents a model of Tower of Hanoi data that stores goals
in memory instead of on the goal stack.  The model predicts average
number of moves per trial, in addition to fitting latencies on optimal
trials with R^2 = .99.

Comments appreciated.

Erik.

Altmann E. M. & Gray, W. D. (1999).  Serial attention as strategic
memory.

    Abstract - Serial attention involves focussing on a sequence of
    thoughts under deliberate control. As such, it has two phases,
    attention switching and attention maintenance. Attention switching
    involves rapidly building up the activation of the new thought so
    it can be temporarily dominant. Attention maintenance allows the
    current thought to gradually decay to prevent it from intruding on
    the next. SASM, a model based on this analysis, suggests that this
    balance of initial activation followed by gradual decay reflects a
    strategic adaptation to architectural constraints and task
    demands. SASM makes accurate predictions about error patterns and
    encoding time, and integrates attention maintenance and attention
    switching in one unified theory.

Altmann E. M. & Trafton, J. G. (1999).  Memory for goals: An
architectural perspective.

    Abstract - The notion that memory for goals is organized as a
    stack persists as a central feature of cognitive theory, in that
    stacks are primitive mechanisms in leading cognitive
    architectures. However, the stack construct over-predicts the
    strength of goal memory and the precision of goal selection
    order, while under-predicting the maintenance cost of both. A
    better approach to understanding cognitive goal management is to
    treat goals like any other kind of memory element. This makes
    accurate predictions and reveals the nature of goal encoding and
    retrieval processes in detail. The approach is demonstrated in an
    ACT-R model of human performance on a prototypically goal-based
    task, the Tower of Hanoi. The model and various theoretical and
    methodological considerations suggest that cognitive
    architectures should enforce a two-element constraint on the
    depth of the stack, to deter its use for storing task goals while
    preserving its use for basic cognitive processes like focussing
    attention and symbolic learning.