Crowds go mad when going gets tough

Bad crowd behaviour is more likely when people face a difficult task or challenging circumstances, new research at the University of St Andrews has found.

Through a large-scale online experiment, the team of researchers examined why human groups sometimes exhibit collective intelligence but at other times demonstrate maladaptive herding.

They found that the more difficult the challenge faced, the more people are likely to conform; and that the greater the number of people exhibiting a behaviour the more others are likely to copy.

The research, published in Nature Human Behaviour, combined mathematical simulations with information gathered from 699 participants who took part in an online experiment.

Dr Wataru Toyokawa, of the School of Biology at the University, said: “Our study shows that maladaptive herding becomes more likely when both group size and task difficulty are high, because people rely heavily on copying the majority’s behaviour without learning themselves.

“Therefore, we should be more aware of the risk of collective madness when these conditions are met. Somehow stimulating individual independent thought may reduce the risk, but this requires further study.”

Instances of collective intelligence have been well documented in human societies. Collective (or group) intelligence can emerge when people work together, for example online shopping consumer behaviour is influenced by product reviews and what products other people buy, allowing ‘smarter’ purchasing decisions.

However, human crowds also suffer from collective madness, when ineffective or even harmful knowledge goes viral – a phenomenon called maladaptive herding – which can trigger financial bubbles and high volatility in stock markets.

Theoretical and experimental studies previously suggested that people are more likely to copy when a task is difficult and group size is large, but it is unclear how increased copying behaviour would affect collective human decisions.

This new research identifies two key factors, “conformity” (that is, the extent to which an individual follows a majority’s behaviour) and “copying tendency” (that is, the extent to which an individual ignores their private knowledge and relies solely on copying), and shows that maladaptive herding becomes prominent when both factors are heightened.

The research team, which included Professor Kevin Laland and Andrew Whalen, also of the School of Biology at the University, believe their results could guide how collective intelligence in real world situations is implemented, for example in consumer-generated web media and prediction markets, where people face challenging and dynamic collective decision problems.