| dc.description.abstracteng | Decision makers are often faced with different options and must make decisions
without having perfect knowledge about the outcomes. They have only samples of
data at hand, extracted from past and present states of their environment, that they
must use to generalise and make predictions.
To help them in this process, humans developed in the 19th century an application
of probability theory commonly referred to as inferential statistics, which
deals with uncertainty inherent to inferences drawn from samples to populations.
However, it is possible that humans were making statistical inferences way before the formalization of this branch of mathematics and that they possess, as part of their biological heritage, a set of probability intuitions that help them handle uncertain situations from infancy onward. It is also possible that such intuitions have deeper evolutionary roots and that humans’ closest relatives, the nonhuman primates, also share such reasoning abilities.
These questions have fascinated scientists for several decades and have led to
numerous studies with human adults, human children of various age categories, and recently, with nonhuman great apes. Findings suggested that human adults are
statistical reasoners depending on how information is presented to them (e.g., natural frequencies vs. percentages). Research with human children yielded ambiguous results. Children from 5 years onward and infants seem to incorporate aspects of probability theory in their inferences. However, research with 3- and 4-years-olds suggests that they are devoid of probability intuitions. Research with nonhuman primates suggested that apes also share some probability intuitions that they use to make predictions from populations to samples.
The aim of my PhD thesis was twofold. First, I aimed to further investigate statistical reasoning abilities in nonhuman primates to shed more light on the evolutionary origins of this ability and to assess how different presentation formats affect nonhuman primates’ reasoning abilities. Second, I aimed to shed more light on the development of probability intuitions during human childhood. In particular, I wanted to try and uncover why 3- and 4-years-olds were outsmarted by infants in the same kind of tasks.
In my first study, I tested whether long-tailed macaques, similar to human children and nonhuman apes, can make predictions from populations of preferred and dis-preferred food items to samples. I used different conditions to rule out that subjects rely on quantity heuristics (i.e., comparing absolute quantities of food)rather than on proportions of food to make predictions. Results showed that longtailed macaques, in contrast to children, apes, and capuchin monkeys (that were tested later on), do not rely on proportions but rather on heuristics to make inferences from populations to samples. These findings might be taken to suggest that the ability to draw this kind of inferences evolved in a convergent fashion in New World monkeys and apes. However, there are still open questions regarding why long-tailed macaques failed and regarding whether the other tested species really relied on proportions.
In my second study, I investigated whether long-tailed macaques can extract statistical information from repeated types of events to make rational decisions under uncertainty. Subjects were presented with different options associated with different probabilities of rewards. Subjects could interact several times with each option separately to learn the different probabilistic patterns of rewards. In a subsequent test trial, subjects could choose between both options presented simultaneously. Results revealed that long-tailed macaques extract statistical information from repeated types of events to make inferences about predictive factors and rational decisions. These findings suggest that, depending on how information is presented, long-tailed macaques are statistical reasoners.
In my last study, I tested whether human infants and preschoolers make inferences from populations of preferred and dis-preferred objects to samples, based on a paradigm already used with both age groups. In my study, however, all objects in the populations were visible to make sure that subjects could rely on the entire proportional information. Results showed that neither preschoolers not infants relied on proportions to make inferences. However, it is not clear why infants failed, as they also failed in a baseline condition in which the outcomes were certain. One
possibility is that infants had no strong preference for any object and therefore chose at chance. Another possibility is that the procedure confused them even if we kept it as close as possible to the procedure of previous studies. A last possibility is
that infant’s performance has been over-estimated in previous studies. In fact, it
is possible that in previous studies, parents, who were not blind to the conditions, interfered with their child’s decisions.
In conclusion, these findings suggest that statistical reasoning is evolutionary ancient, as different primates species (and bird species) engage in it. However,
there might be more than one cognitive mechanism at play, as long-tailed macaques succeeded when statistical information could be experienced through repeated types of events but not when it had to be extracted from the present environment. These findings also suggest that the ability of infants to rely on proportions of objects to make inferences about random sampling might have been over-estimated in the past. In general, more research is necessary to disentangle between different kinds of probability intuitions and their respective evolutionary origins and to understand how these intuitions develop during human ontogeny. | de |