Modeling parallelization and flexibility improvements in skill acquisition: from dual tasks to complex dynamic...

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Artificial Intelligence

Background:

• Rule-based models traditionally exhibit brittle behavior, contrasting with observed flexibility in cognitive skill acquisition.
• Emerging parallel processing capabilities challenge existing models of learning and cognitive skill development.

Purpose of the Study:

• To reconcile rule-based models with observed parallel processing and flexibility in cognitive skill acquisition.
• To explain how rule-based systems can learn and adapt, avoiding brittle behavior.

Main Methods:

• Investigated two principles for rule-based models: gradual learning of task-specific rules and prioritizing bottom-up processing.
• Applied these principles to a dual-task time-sharing model (Schumacher et al., 2001) and a complex dynamic task model (Carnegie Mellon University Aegis Simulation Program [CMU-ASP], Anderson et al., 2004).

Main Results:

• In dual-task learning, speedup learning and bottom-up instruction activation explained parallel behavior.
• In the CMU-ASP model, parallel behavior arose from the shift from serial to flexibly activated rules (top-down and bottom-up), enabling opportunistic instruction reordering and strategy emergence.

Conclusions:

• Rule-based models can exhibit flexibility and parallel processing by learning rules and utilizing bottom-up processing.
• These models provide a framework for understanding the gradual emergence of new strategies in complex cognitive tasks.