Abstract:Transfer learning, the reuse of newly acquired knowledge under novel circumstances, is a critical hallmark of human intelligence that has frequently been pitted against the capacities of artificial learning agents. Yet, the computations relevant to transfer learning have been little investigated in humans. The benefit of efficient inductive biases (meta-level constraints that shape learning, often referred as priors in the Bayesian learning approach), has been both theoretically and experimentally established. Efficiency of inductive biases depends on their capacity to generalize earlier experiences. We argue that successful transfer learning upon task acquisition is ensured by updating inductive biases and transfer of knowledge hinges upon capturing the structure of the task in the inductive bias that can be reused in novel tasks. To explore this, we trained participants on a non-trivial visual stimulus sequence task (Alternating Serial Response Times, ASRT); during the Training phase, participants were exposed to one specific sequence for multiple days, then on the Transfer phase, the sequence changed, while the underlying structure of the task remained the same. Our results show that beyond the acquisition of the stimulus sequence, our participants were also able to update their inductive biases. Acquisition of the new sequence was considerably sped up by earlier exposure but this enhancement was specific to individuals showing signatures of abandoning initial inductive biases. Enhancement of learning was reflected in the development of a new internal model. Additionally, our findings highlight the ability of participants to construct an inventory of internal models and alternate between them based on environmental demands. Further, investigation of the behavior during transfer revealed that it is the subjective internal model of individuals that can predict the transfer across tasks. Our results demonstrate that even imperfect learning in a challenging environment helps learning in a new context by reusing the subjective and partial knowledge about environmental regularities.

Transfer of experience with a class-schema to identification-learning of patterns and shapes.

Mirror-image relations in category learning

High-variability training does not enhance generalization in the prototype-distortion paradigm

FUNCTIONAL CLASSES AND EQUIVALENCE RELATIONS

Cross-modal Transfer in Man

Playing on patterns: is it a case of analogical transfer?

Learning in rich networks involves both positive and negative associations.

Recognition of Shapes across Modalities by Infants

Cross-Modal Transfer of Training in Monkeys

Stimulus equivalence and transfer of function: Teaching categorization skills to children

Identifying Transfer Learning in the Reshaping of Inductive Biases

Prior familiarity enhances recognition memory of faces, not just images of faces, when accompanied by conceptual information

Incidental learning of temporal and spatial associations in hybrid search

Prototype or Exemplar Representations in the 5/5 Category Learning Task

The Effects of Multiple Exemplar Instruction on the Relation Between Listener and Intraverbal Categorization Repertoires

Where Hypotheses Come From: Learning New Relations by Structural Alignment

Infants can use temporary or scant categorical information to individuate objects

The role of category valence in prototype preference

Test performance in optional shift and configural acquired equivalence are positively correlated.

Asymmetrical Transfer of Learning Effects Between Signal-Based and Criterion-Based Task Uncertainties in Perceptual Decision