Overcoming learning traps with summative feedback

Abstract

Learning traps arise when early experience leads to a false belief about the reward structure of the environment which, in turn, leads to avoidance of rewarding options. Previous work on the negative effects of such traps have focused on static learning environments. The current work examines an additional negative effect of learning traps in dynamic environments – blindness to change in the features that predict decision outcomes. In two experiments (N = 416), participants had to decide whether to approach members of two different categories, respectively associated with either gains or losses. Early in learning, a category rule involving two feature dimensions predicted category membership. Subsequently, there was a change in the feature composition of this rule. When outcome feedback was only provided when an item was approached, a substantial proportion of participants fell into the trap of using a simple one-dimensional rule to guide approach decisions. Most of these participants did not notice the subsequent rule change and never learned the new rule. Signaling the possibility of rule change (Experiment 2) had no effect on change blindness for those in the learning trap, but did improve learning of the new rule for those who initially avoided the trap.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 0 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.