Swarm intelligence and the wisdom of crowds
What is the difference between the wisdom of crowds and swarm intelligence?
· Janne Ruohisto
Editor's note:
This article was originally published in 2014, at a time when the concept of swarm intelligence was beginning to enter the Finnish public conversation.
It was also cited by Yle (2015).
The concepts derived from swarm intelligence — swarm work and swarm learning — have been steps towards what has since become SwarmMind, a learning platform where people and AI learn together.
The original post has stayed with us over the years as part of our history, and it remains one of the most searched and most read pieces on our site.
The terms wisdom of crowds, crowdsourcing and swarm intelligence are often used as synonyms, even though they do not quite mean the same thing.
Broadly speaking, collective intelligence refers to gathering ideas and feedback from a large group of people. In crowdsourcing, an organisation or institution hands a task or a problem to an undefined audience — a crowd. That task might be an idea competition, a call for solutions to a difficult problem, or something like the gamified digitisation of old newspapers.
In the wisdom of crowds as James Surowiecki describes it, participants generally do not collaborate with each other. They give independent personal opinions, ideas or proposals, often anonymously, and the party coordinating the effort aggregates them. Anonymity and the absence of interaction between members of the crowd support the independence of judgement that crowdsourcing depends on. By minimising the influence of social power relations on ideas and opinions, the method sets out to solve the problem of groupthink — captured in the old saying that a crowd concentrates stupidity.
Don Tapscott and Anthony D. Williams, however, argue that collective intelligence can also take the form of mass collaboration. They use the umbrella term wikinomics for the phenomenon: openness, peering, sharing and acting globally, enabled by technology. Wikipedia and the Linux community are examples of this kind of collaboration. Similar work can also be described using Kari A. Hintikka's concept of network power (verkkovoima), which refers to collaboration in which an arbitrary group of people self-organises over networks to achieve a shared goal, without being directed or controlled from the outside by any organisation or institution.
Swarm intelligence, in turn, is a term first introduced in 1989 in the field of robotics, in the study of the collective behaviour of self-organising agents. Although the behaviour of the agents is not centrally controlled, global patterns of behaviour emerge from local interactions between them.
Since then, applications based on swarm intelligence have been developed in many areas: data mining, network design, and the control of unmanned military vehicles. In the study of human behaviour, swarm intelligence has been applied to things like simulating how airline passengers board a plane.
The definition of swarm intelligence is not, however, especially clear. The common denominator across its various applications appears to be little more than that they take inspiration from swarms in nature — the swarming behaviour of bees, ants or birds.
The definition usually includes the assumption that individual members of the swarm follow only a few very simple predetermined rules, without any broader awareness of the whole. A bird in a flock knows how to evade a hawk and how to avoid colliding with the bird beside it, but not necessarily much more than that. And yet the flock's collective behaviour looks distinctly intelligent.
Applying swarm-like behaviour in this sense to human collaboration has been criticised, because people are more intelligent than birds or insects. People are not programmed to follow simple rules; collaboration is driven by motivation, by the will — or the lack of will — to work together. This is why Intunex's swarm model takes motivation into account alongside skills.
In the swarm work model developed by Intunex, a swarm is defined as a working group that organises itself around a shared social object — an idea, a problem or a project — on the basis of skills and interest. The swarm works collaboratively to develop that shared object, creating new knowledge or new ways of working as it goes.
For a swarm to form, someone has to share at least a rough outline of the shared object, so that interested people can join and contribute their own thinking. By this definition, self-organising community events can also be called swarms — Community Manager Appreciation Day, say, or kaljakellunta, Finland's improvised mass beer-floating gatherings. Intunex's model, however, was developed primarily as a way of working inside an organisation, for collaboration and skill sharing, and it accounts for the fact that the work is steered by the organisation's goals.
A swarm is worth distinguishing from traditional groups such as project groups, working groups or teams, which are typically formed administratively. A traditional group carefully defines how many people take part, how much each of them contributes, who reports to whom, and when the group starts and finishes. In a swarm, essentially all of these are different. Participants are not defined administratively or in advance. The number of participants is not capped. The level of participation can vary flexibly according to skill and interest.
Because swarms form according to skills and interest rather than position, title or hierarchy, they do not suffer from groupthink to the same extent as traditional groups. The way a swarm forms adds diversity to the group — which is also one of Surowiecki's criteria for successful collective intelligence.
In Surowiecki's crowds, diversity is achieved by collecting proposed solutions from as wide an audience as possible, whose members offer independent proposals, often anonymously and without interacting with one another. While this undeniably produces excellent results much of the time, the method can be criticised as inefficient. For it to work, every problem has to be put to a large audience in order to get enough variation in the answers. A company could hardly poll its entire staff — let alone a wider audience — on every difficult problem it faces.
Swarms make more efficient use of existing skills by directing the problem to the people who have relevant expertise or interest. Diversity is achieved by breaking the traditional hierarchical ways of organising, which brings people with different backgrounds and from different positions into the swarm.