Abstract
Probability matching has long been taken as a prime example of irrational behaviour in human decision making; however, its nature and uniqueness in the animal world is still much debated. In this paper we report a set of four preregistered experiments testing adult humans and Guinea baboons on matched probability learning tasks, manipulating task complexity (binary or ternary prediction tasks) and reinforcement procedures (with and without corrective feedback). Our findings suggest that probability matching behaviour within primate species is restricted to humans and the simplest possible binary prediction tasks; utility-maximising is seen in more complex tasks for humans as pattern-search becomes more effortful, and we observe it across the board in baboons, altogether suggesting that it is a cognitively less demanding strategy. These results provide further evidence that neither human nor non-human primates default to probability matching; however, unlike other primates, adult humans probability match when the cost of pattern search is low.
Introduction
In the study of decision-making, "rationality" is often defined by an agent's ability to maximize expected utility. In a simple game of chance—for example, a task where one button yields a reward 70% of the time and another only 30%—the optimal strategy is to always select the 70% option. This strategy, known as maximizing, ensures the highest possible success rate.
However, decades of psychological research have shown that human adults often engage in probability matching, where they distribute their choices to match the reward frequencies (e.g., picking the 70% option 70% of the time). This behavior is technically sub-optimal, as it drops the success rate from 70% to approximately 58%. This paper investigates whether probability matching is a fundamental primate flaw or a specific byproduct of the human drive to find patterns in random data.
Methods
The researchers conducted four preregistered experiments comparing 238 adult humans (recruited via Amazon Mechanical Turk) and a social group of 25 Guinea baboons (Papio papio) at the CNRS Primate Center in France. The baboons utilized automated, touch-screen testing booths accessible voluntarily within their social environment.
As shown in Figure 1, the study manipulated two primary variables:
- Task Complexity: Subjects faced either Binary (2 choices, 70/30 distribution) or Ternary (3 choices, 70/15/15 distribution) prediction tasks.
- Reinforcement Procedures:
- Standard: Subjects received a reward for correct guesses and nothing for incorrect ones.
- Guidance: Incorrect guesses were followed by a corrective prompt, requiring the subject to select the correct answer to receive a partial reward. This guaranteed a reward eventually but added "effort" to errors.
Results
The findings revealed a striking contrast between species. Under standard conditions, baboons reached near-perfect maximizing behavior—choosing the 70% option over 95% of the time—by the midway point of the experiment for both binary and ternary tasks. They did not exhibit the human bias toward sequence-searching; they simply identified the best-paying option and stuck to it.
Human behavior, however, was highly sensitive to task complexity. In simple binary tasks, only 25% of humans maximized. When the task was expanded to three choices (ternary), this jumped to 60%. As illustrated in Figure 4, the proportion of optimal responses was consistently higher in the ternary (skewed 3) condition than in the binary (skewed 2) condition across Experiments 1 and 2. Even doubling the number of trials in Experiment 1B did not lead humans in the binary task to converge on the optimal strategy.
Discussion
These results strongly support the "Pattern Search" hypothesis. Humans appear to probability match because we are obsessed with finding sequences (e.g., "Left, Left, Right") even in random noise. In simple binary tasks, the cognitive "cost" of searching for a pattern is low, so we continue to try to "crack the code." In ternary tasks, the number of possible patterns explodes, making the search too taxing. When pattern-hunting becomes too difficult, humans "give up" and switch to the simpler, more effective strategy: maximizing.
Interestingly, the study found that maximizing itself requires some cognitive effort. In the "Guidance" condition (Experiment 2), where rewards were guaranteed regardless of the initial choice, baboons stopped maximizing and began hitting random locations. This suggests that primates will abandon the effort of maximizing if the stakes (the cost of error) are removed.
Conclusion and Limitations
This research demonstrates that probability matching is not the default primate strategy. Instead, it is a specific human behavior triggered when the perceived cost of searching for patterns is low. While baboons appear more "rational" in simple tasks because they lack the human drive for sequence-searching, both species default to maximizing when the cognitive effort of other strategies becomes too high.
Limitations of the study include the differing nature of rewards (money for humans vs. wheat pellets for baboons) and the testing environments. Future research could explore whether explicitly interrupting pattern-search behavior can force humans to maximize in even the simplest tasks.
Summary updated