Random behavior is stable across tasks and time

Whether it’s choosing a tennis serve or escaping a predator, the ability to behave randomly provides a range of adaptive benefits. Decades of work explores how people both produce and detect randomness, revealing profound non-random biases and heuristics in our mental representations of randomness. But how is randomness realized in the mind? Do individuals have a “one-size-fits-all” conception of randomness that they employ across different tasks and timepoints? Or do they instead use simple context-specific strategies? Here, we develop a model that reveals individual differences in how humans attempt to generate random sequences. Then, in three experiments, we reveal that random behavior is stable across both tasks and time. In Experiment 1, participants generated sequences of random numbers and one-dimensional random locations. Behavior was remarkably consistent across the two tasks. In Experiment 2, we gave participants both a random-number-generation and a two-dimensional random-location-generation task, such that the tasks diverged in structure. We again observed stable individual differences across tasks. Finally, in Experiment 3, we collected data from the same participants as in Experiment 2, but one year later; we found stable individual differences across that span. Across all experiments, we find idiosyncratic behaviors that are stable across tasks and time. Thus, we suggest that a trait-like randomness generator exists in the mind.