Making Experts: Optimizing Perceptual Learning in Complex, Real-World Learning Domains
How do we accelerate the process of gaining expertise? Recent research suggests that advanced pattern recognition and fluency can be developed in a short period of time using adaptive and perceptual learning technology (e.g., Kellman & Kaiser, 1994; Kellman, Massey, and Son, 2009). Much is still unknown, however, about the connections between perceptual learning and adaptive learning technology that allow for the efficient development of such expertise effects.
In six experiments, I examined a number of learning principles that bridge perceptual and adaptive learning and explored the generalizability of these learning principles across domains. In particular, I evaluated how different types of learning trial formats and feedback may bring about fluent structure recognition while improving training efficiency. To ensure that principles and experimental results are not confined to a single learning domain, I carried out these studies in two separate domains: mathematics and medical learning. Experiments 1, 3 and 5 trained undergraduates to interpret electrocardiogram recordings; Experiments 2, 4, and 6 replicated the design of the other three experiments, but trained participants to map between graphical and symbolic representations of trigonometric and exponential functions. Experiments 1 and 2 showed that the combination of passive exposure to the correct classifications and active classification practice enhanced fluency in pattern recognition while improving training efficiency. Experiments 3 and 4 explored the benefit of comparisons among contrastive examples and revealed that training with only comparisons can be detrimental, but that having some comparison practice can facilitate far transfer. Experiments 5 and 6 evaluated and demonstrated the effectiveness of a new paradigm that adaptively triggers paired-comparisons based on learners’ error patterns to maximize training efficiency. Positive effects on learning were found in both learning domains.
These findings help to illuminate basic questions about the processes by which expert information extraction advances, and they inform our understanding of the general mechanisms that operate across learning domains. The results also lend themselves to applications in which learning interventions maximize the ease with which students pick up relevant structural relations in novel situations while minimizing training time.