Heterogeneous Suppression of Sequential Effects in Random Sequence Generation, but Not in Operant Learning

Area of Science:

• Cognitive Psychology
• Behavioral Economics
• Computational Neuroscience

Background:

• Human random number generation is a classic experimental paradigm.
• Research has evolved from identifying heuristics causing non-randomness to predicting future choices.
• Understanding cognitive strategies in random sequence generation remains crucial.

Purpose of the Study:

• To analyze temporal sequences in binary random number generation using generalized linear regression and Reinforcement Learning.
• To investigate psychological principles and heuristics influencing deviations from randomness.
• To compare sequential effects in random sequence generation with operant learning tasks.

Main Methods:

• Logistic regression analysis to characterize temporal sequences of choices.
• Generalized linear regression and Reinforcement Learning framework.
• Population-level and individual-level analyses of sequential effects.

Main Results:

• Population analysis showed weak recency effects, contrary to standard models.
• Individual analyses revealed monotonous decreasing sequential effects with delay.
• Behavioral heterogeneity significantly averaged out individual effects in population analysis.
• A heterogeneous model effectively captured sequential dependencies in random sequence generation.
• Sequential effects in operant learning were more homogenous than in random sequence generation.

Conclusions:

• Substantial behavioral heterogeneity exists in random sequence generation tasks.
• Participants employ diverse cognitive strategies to suppress sequential dependencies.
• Reinforcement Learning framework provides insights into sequential effects across tasks.
• Individual differences are key to understanding deviations from randomness in human behavior.