Understanding the Digital Swarm

Those familiar with the to-and-fro of social media messaging may well recognise the concept of "digital swarming," the observation that traffic on these platforms isn’t constant, but there are periods with large amounts of messaging activity interspersed with periods of relative quiet. Often, an issue arises that generates many messages and conversations, and then these die away again very quickly. In some ways, the digital swarm resembles the assembly of large numbers of independent creatures who join together to herd, flock, or swarm. In fact, studies of "biological swarming" behaviour can give insight into the reasons behind digital swarming. These insights may allow the social media user to recognise and understand the behaviours of those surrounding them in the digital swarm.

In nature, a swarm (or flock, or herd) occurs for many reasons and can serve many functions for the gathering beasts. It is often suggested that a key reason for a swarm is the avoidance, or confusion, of predators. Indeed, the "selfish herd" idea¹, although now some 50 years old, remains a well-researched theory used to explain swarming behaviour. Basically, the imperative is to get somebody else between you and the lion. Other theories of swarming include factors such as resource protection, increased chances of mate selection, and joint problem-solving². The root of this latter idea is that if you hang around with other birds and one of them finds food, then you’ll have a better chance of getting some yourself.

Thus, creatures swarm for self-protection and/or increased intelligence-gathering (access to resources). The question for humans in the 21^st century is: How does all this relate to the digital swarm? In fact, this question has been the focus of several investigations into the uses of digital swarming³. Many AI enthusiasts have anxiously sought evidence that mass digital interactions are a good thing, which will enhance human intelligence by allowing real-time cooperation in information-seeking behaviours. Indeed, there is evidence that this suggestion may have some utility under limited circumstances, but there are also some big buts to be considered in this field.

One study3 that suggested reasons for optimism regarding the digital swarm enhancing group intelligence, devised a platform in which people could cooperate to reach a prediction about the results of an event. If true, this clearly has some adaptive advantage. The prediction was arrived at by group members acting simultaneously, not one at a time as in a successive poll (which increases the chance of "groupthink" developing). The swarm predictions were about 30 percent more accurate than those made through successive polling. However, there are two things to note: firstly, the task was about something entirely trivial (who will win the Oscar?), and of no real individual value to any of the swarm members; and secondly, we do not know how the accuracy of these swarm predictions compare with those of a well-informed single individual.

Psychology^2,4 and other AI⁵ studies suggest that the contingencies required for the stability of swarm systems do not point to strong chances for their long-term usefulness as information-gathering processes. If anything, there is more evidence that the digital swarm may have a greater benefit to small-group protection—which comes at its own costs to advancing knowledge and coherence across large populations and also brings the debate back to the selfish herd¹.

A recent study of the structure of swarms⁵ revealed two possible kinds of aggregation of individuals. The first structure was an equally-weighted set of organisms who all talked to one another quite a lot. The second structure was more hierarchical, where one organism did a lot of information-gathering, which was relayed to the rest of the swarm, which did the receiving and talking to one another. The first type of swarm had high levels of social coherence but low levels of good knowledge. The second type of swarm had higher levels of good knowledge, but a mixed degree of social cohesion (one disconnected leader and many connected followers).

The reason that one or other of these two structures develops is premised on the suggestion that any individual can do one of two things at any given time: they can observe and gain information directly (which is resource-costly, but accurate); or talk and gain information indirectly (which is less costly, but also less accurate). The degree of talking together increases swarm cohesion while the degree of individual observation decreases group cohesion, but vice versa for knowledge. There are other factors, like suggestibility, to consider, but this is the basic idea.

There is reason to suppose that the latter form of digital swarm may allow members more and better information and is useful in gaining intelligence. However, think of the contingencies acting on the swarm "leader," and ask: what’s in it for them? They transmit a lot of hard-earned information to others and do not receive much back in the way of communication. Indeed, when they do, it is of dubious informational quality. Unless the leader receives some other kind of benefit from this process, they will quickly extinguish this behaviour, the swarm will have no good information, and it will perish. If the leader receives some other, non-informational benefit, then their motives around information-gathering are questionable (especially if that benefit comes from giving some types of information and not others—think, here, of social media influencers). Essentially, this reverts to the problem of social learning, which is only useful if at least one member of the social group knows what they are doing⁴.

The other form of swarm may fare a little better, but perhaps not much. In this swarm, all members are rewarded by mutual communication. All feel valued, and are likely to stay in the swarm. The swarm may then offer benefit in terms of the selfish herd hypothesis—there’s safety in numbers. This may even begin to be reinforced by the swarm fending off attacks from other groups². In this way, the development of collective narcissism⁶ can be explained, where groups of like-minded people find strength purely by saying how great they each are, and how awful and threatening everybody else is (think of political parties and extremists). However, while this may protect the group from predation, and perhaps prolong its existence, the swarm is not good at finding new accurate information; eventually, it will fail to adapt to new circumstances and will perish.

Social Media Essential Reads

Social Media Scams Continue to Dominate Losses

We Know Social Media Use Stirs Up Anxiety and Depression for Many

All of this combined theorising from psychology, AI, and evolutionary research suggests that neither version of digital swarming has great long-term usefulness as a means to further knowledge acquisition for humans. Perhaps it may be better to simply learn what you need for yourself, or from a few trusted others, and not to follow the digital herd.