Swarm intelligence in animals and humans


Group living enables animals to solve problems that are difficult or impossible for single individuals to resolve. For instance, individuals in groups can catch larger prey more easily or better protect themselves against predators. The recent exploitation of self-organisation theory (see Glossary) in animal behaviour and the resulting study of collective behaviour has prompted the realisation that group living can also facilitate solving cognitive problems that go beyond the capacity of single animals: a phenomenon known as swarm intelligence (SI), a term coined by Gerardo Beni. For instance, there are many examples of SI in social insects and slime moulds which demonstrate the ability to make speed versus accuracy tradeoffs in decision making and to negotiate mazes, respectively. The underlying perception in most biological (non-human) case studies (particularly in invertebrates) of SI has been that the individual animal is cognitively relatively simple and restricted in what it can achieve, whereas the group collectively is capable of astonishing feats. In humans, however, there is not only interindividual variation in cognitive abilities, but there are also some individuals that are high performers by any standard. As a result, one emphasis in psychology has focused on assessing if a group can outperform high-performing individuals and on trying to find the limits of what a group of a given size and composition can collectively achieve. In contrast to the recent arrival of SI in the biological sciences, experimental work by psychologists on decision making in human groups stretches back approximately seven decades. One of the first to point out that individuals can benefit from collective decision making was an 18th-century French mathematician, Nicolas de Condorcet. He assumed that each individual can be either right or wrong with a certain probability p. Provided that p > 50%, the chance of a correct collective decision of the group will increase as a function of group size. However, Condorcet’s theorem requires that the individuals that have the correct information are in the majority. Therefore, the theorem has more to do with the means by which consensus decisions are made when a majority already has correct information. In 1907, Galton recognised that SI does not require a majority of people who already know the correct answer (in fact, SI does not require anybody to know the correct answer) and was the first researcher to empirically demonstrate SI in humans.
Previous reviews of SI have focused mainly on animal groups because the literature on humans usually does not use the term ‘SI’ (but does use terms such as ‘group decision making’). However, the human literature has much to offer on this topic and an integration of the SI work on animal and human groups would help further advance this rapidly developing field. To facilitate integration, we propose a general definition of SI. The term ‘SI’ is often used very loosely (especially in the media) and a clear definition is missing from the literature. In this context, there has been much speculation on the possibilities of SI but little critical assessment to put it into perspective. Our review has three main aims: (i) to provide a general definition of SI; (ii) to address not only the possibilities but also the limitations of SI; and (iii) to integrate SI work on animal groups and human groups.