New paper by Carmen Saldana on (ir)rational decision making strategies across human and non-human primates

Our species is often acclaimed for its rationality, and apparently irrational cases of human decision-making— where we make choices that do not maximise expected utility—are therefore of perennial interest. A prime example of such a putatively irrational behaviour is probability matching. In simple prediction tasks where participants are asked to select between two variants with different (but fixed) probabilities of reward, adult human participants tend to select variants proportionally to those reward probabilities (e.g., selecting the variant that is rewarded 70% of the time approximately 70% of the time) even though the optimal strategy would be to always select the more-frequently-rewarded variant (a strategy known as maximising). This is particularly puzzling given that children, monkeys and rats often maximise in similar tasks. Why would an adult human be outperformed by a monkey or a six year-old child on such a simple task?

Recent studies suggest that probability matching behaviour in adult humans might be less robust than previously thought and very much dependent on task-design; further, the lack of systematic comparisons across species makes it hard to determine the nature and uniqueness of the adult human decision-making strategy. In our study we take a cross-species comparative approach and test adult humans and Guinea baboons (with a sample size unmatched in the literature, N = 20) on a series of closely-matched (and pre-registered) probability learning tasks, manipulating task complexity (two alternative binary or three-alternative ternary prediction tasks) and reinforcement procedures (with or without corrective feedback after unrewarded trials). We show that probability matching is restricted to humans and even then only occurs in the simplest possible binary prediction tasks; utility-maximising is instead seen in more complex tasks (such as ternary prediction tasks) as pattern-search becomes more effortful. In baboons, we observe utility-maximising everywhere, except in versions of the task where an even simpler strategy of random responding leads to the same reward, in which case baboons respond randomly. These results provide further evidence that neither human nor non-human primates default to probability matching in the presence of an alternative maximising strategy. They further suggest that probability matching is the most cognitively demanding strategy, with maximising being less demanding but still more challenging than random responding; unlike other primates, adult humans probability match when the cost of pattern search is low, and can maximise on tasks where baboons will fall back on random responding strategies.


Probability matching is not the default decision making strategy in human and non-human primates