Understanding randomness via the perception and prediction of binary sequences

Abstract

The thesis explored whether a common concept of randomness determines behaviour on two randomness cognition tasks; perceptions of randomness and predictions of future outcomes. The impact of a causal agent, presentation format, and sequence length, on randomness cognition was also investigated. Judgements of randomness were influenced by the use of causal information about sequences. Participants in a control condition, with no information about the sequence source, were not biased in their judgements of randomness, whilst those told that sequences represented either coin tosses or shots in basketball were biased. Predictions of future outcomes were shown to be strongly influenced by sequence alternation rate. However, over-alternation still occurred, particularly for mid-alternation rate sequences. Causal belief was found to influence predictions of future outcomes. People were more likely to continue the run when the cause was animate but to end it when the cause was inanimate. The effect of presentation mode was examined by comparing responses to symbol and random walk trend sequences, with mixed results. For the simplest tasks no effect was found for presentation mode. However, the trend sequence format was susceptible to influences from other variables such as a causal example. Sequence length influenced judgements of randomness, as people were more biased in their ratings of short sequences than long sequences. Two proposed mechanisms underlying randomness cognition were explored: limitations in working memory capacity (WMC) and sequence complexity. Results found no support for WMC as a general mechanism for randomness cognition, as performances on working memory and perception of randomness tasks were not correlated. Additionally, no support was found for sequence complexity as a mechanism for explaining perceptions of randomness. The results have important implications for the randomness literature as well as questioning some previously proposed underlying mechanisms. The commonly used causal example was shown to influence judgements of randomness, thus suggesting that previously reported biases towards rating higher alternation rate sequences as most random may be due to the use of a causal example. Taken together, the results raise doubts about the validity of a common concept of randomness underpinning performance on different types of randomness cognition.