The Effect of Context and Individual Differences in Human‐Generated Randomness

Abstract

Many psychological studies have shown that human-generated sequences are hardly ever random in the strict mathematical sense. However, what remains an open question is the degree to which this (in)ability varies between people and is affected by contextual factors. Herein, we investigated this problem. In two studies, we used a modern, robust measure of randomness based on algorithmic information theory to assess human-generated series. In Study 1 (N = 183), in a factorial design with task description as a between-subjects variable, we tested the effects of context and mental fatigue on human-generated randomness. In Study 2 (N = 266), in online research, in experimental design, we further investigated the effect of mental fatigue on the randomness of human-generated series and the relationship between the need for cognition (NFC) and the ability to produce random-like series. Results of Study 1 show that the activation of the ability to produce random-like series depends on the relevance of the contextual cues (χ 2 (2) = 7.9828, p = .0192), whether they activate known representations of a random series generator and consequently help to avoid the production of trivial sequences. Our findings from both studies on the effect of mental fatigue (Study 1t (47, 529.5568) = −18.62, p < .001; Study 2 -F (edf = 3.587, Re f .df = 3.587) = 11.863, p < .0001) and cognitive motivation (t (180) = 2.66, p = .009) demonstrate that regardless of the context or task's novelty people quickly lose interest in the random series generation. Therefore, their performance decreases over time. However, people high in the NFC can maintain the cognitive motivation for a longer period and consequently on average generate more random series. In general, our results suggest that when Correspondence should be sent to Mikołaj Biesaga,

The role of the working memory storage component in the random-likes series generation

“…The model proposed by Biesaga et al (2021) accounts for the locality of the generation process and relies on the account that when producing a random series people never consider each element separately. Instead, when generating series, first, they propose a new element to be added to the already generated sequence.…”

“…Based on this theoretical account, Biesaga et al (2021) argue that the process of random series generation involves both storing and processing components of working memory. While the former is used for maintaining active the already generated sequence the latter is employed for the evaluation of whether the proposed element increases the randomness of past choices (compare Figure 1).…”

“…For example, Warren et al (2018) showed that when the working memory limitation is considered the overall frequency of elements in human-generated series compared to the outcomes of the Bernoulli process is similar. However, Biesaga et al (2021) pointed out that the generation of random series is not a static process and over time the randomness of human-generated series decreases. Furthermore, they demonstrated that this effect is consistent regardless of whether people are asked to produce a long series or an array of short ones.…”

“…Consequently, researchers not only struggled with the task definition but also with the measurement of the randomness for relatively short sequences. Therefore, it was very difficult to compare the results inbetween studies (see Biesaga et al, 2021, for the review of used methods).…”

“…The article is organized as follows. In the next section, we describe a simple theoretical model of random-like series generation proposed by Biesaga et al (2021) that accounts for the involvement of two components of working memory. Then, in two studies using modern methods for quantifying randomness grounded in AIT and based on the notion of algorithmic (Kolmogorov) complexity, we test the hypothesis regarding the role of working memory storage component (H1), the effects of individual differences in the randomness judgments (H2) and working memory capacity (H3), on the complexity of human-generated series.…”