Learning traps and change blindness in dynamic environments.

Area of Science:

• Cognitive Psychology
• Behavioral Economics
• Machine Learning Theory

Background:

• Learning traps occur when initial experiences create false beliefs about reward structures, leading to avoidance of beneficial choices.
• Prior research primarily investigated learning traps in static environments.

Purpose of the Study:

• To investigate the impact of learning traps in dynamic environments, specifically focusing on "change blindness"—the failure to notice alterations in predictive features.
• To examine how learning traps affect adaptive decision-making when environmental rules change.

Main Methods:

• Two experiments (N=416) involved participants deciding to approach items from categories associated with gains or losses.
• A two-dimensional category rule was used initially, followed by a change in its feature composition.
• Outcome feedback was contingent on approach decisions, simulating real-world scenarios.

Main Results:

• A significant portion of participants adopted a simplified one-dimensional rule, falling into a "learning trap."
• These participants exhibited "change blindness," failing to detect or learn the updated two-dimensional rule.
• Signaling the possibility of rule change did not help trap-entrenched individuals but improved learning for those who avoided the trap.

Conclusions:

• Learning traps exacerbate decision-making deficits in dynamic environments by inducing "change blindness."
• The findings highlight the persistence of maladaptive strategies even when environmental dynamics shift.
• Interventions to prevent falling into learning traps may be crucial for promoting adaptability in changing environments.