One of the things I have had trouble explaining when defining the concept of Collective Intelligence is the preceding headline. That is, there are situations that can lead to Collective Intelligence (CI) in which individuals do not interact directly or are even conscious of the fact they form part of the collective.
The MIT Centre for Collective Intelligence and many well-known authors in the field admit both the results of non-conscious aggregations and the consequences of active collaboration as manifestations of CI. In James Surowiecki’s “The Wisdom of Crowds”, the book that popularized CI, there are plenty of examples based on pure statistical aggregation, that is, without any direct interaction among participants.
To make the differences between modalities clearer I use the terms “Collected” and “Collaborative” Collective Intelligence. Let me explain both.
“COLLECTED” Collective Intelligence:
In this case the collective results are emergent properties of the sum and combination of independent acts. Someone collects the data that is later aggregated. There is no direct social, nor conscious, interaction among individuals. The result emerges from the aggregation of individual behavior of individuals acting strictly in their own self-interest.
Amazon’s recommendation system or Google’s search engine are well known examples. In both cases there is “collective aggregation”. The metadata of thousands or millions of users is processed to generate a product recommendation or rank the links to some content or other.
In this case the participatory architecture is mainly tech based because the design only needs to find an ingenious way of aggregating individual contributions. And what has more relevance for this analysis: the technological solution, normally some type of statistic algorithm, is in the hands of an external aggregator, beyond any collective control. The “Crowd” hands the data over, and the external agent (Amazon, Google, Facebook, etc.) processes and feeds back the collective results, that individuals will make use of.
James Surowiecki actually insists that the “independence” of the individual is a pre-requisite in order to obtain Collective Intelligence. Following his reasoning, if we avoid individuals’ direct interaction we can guaranty there will no mutual influence, and that effect of statistical aggregation will be free of biases or manipulation. Here individuals act as if there were no structure connecting them as a group, leading some authors to call these “PASSIVE systems”.
“COLLABORATIVE” Collective Intelligence:
In contrast to preceding modality, in this case interaction is deliberate and conscious, implying the effort of group members socializing. There is intentional dialogue among individuals, leading some authors to call these “ACTIVE systems” of Collective Intelligence.
Individuals collaborate to obtain their personal and joint objectives. The iconic example of this mode of interaction is Wikipedia. Here intelligence emerges from conscious connections, an interest in sharing, giving, receiving and socializing. This CI behaves as a “collaborative intelligence”, and the participatory architecture emphasizes reinforcing the social aspect of these conscious links.
Pierre Lévy is the main inspiration for this way of seeing Collective Intelligence, as an interactive process among those that take part that implies feedback and the explicit will to search for a shared solution to a problem. The great Henry Jenkins is another author that has highlighted the advantages of this modality of CI.
“COLLECTED” vs. “COLLABORATIVE”:
I prefer to be a realist and accept that both modalities of aggregation are viable and useful. In both cases individual data can be aggregated to generate a result that makes us more intelligent as a collective. Who can deny that Google’s search engine (even with its darker aspects) is very useful for filtering information based on its capacity to aggregate thousands of individual preferences?
The final choice depends to some great measure on the type of problem we want to solve. If we need to process large volumes of data to obtain a statistically significant result, “collected” mode can (under certain circumstances) work quite well. But if the probability of reaching a collective solution to a challenge improves with interactivity, meaning, if one can foresee that the search for the solution to a qualitatively complex problem can be enriched through conversation and collective dialogue, then “collaborative” is the best option.
The advantage of “collected” CI is that it is easy to scale up to manage large volumes of information; this is a lot harder when collaborative mechanisms, with considerable coordination costs, are used. On the other hand, the problem with the “collected” model is in the use (and abuse) of the data by the external aggregator. This is a serious drawback for, e.g., Big Data. The only way to attenuate this risk is to demand more transparency in the aggregators’ machinery insuring that the participants’ rights are respected.
To round up, what happens when data collection is neither conscious nor participatory? According to Elizabeth F. Churchill a “collected” model implies a certain inconvenience: “it is not that people don’t get to learn the solution, they don’t get to see the problem either”. This is why I am interested in calibrating the degree of intentionality. Individuals should not just be conscious that they a working together, they should know why they are doing it. Investing Social Capital, in the process of building social intelligence through direct interaction itself, is extremely valuable. A the same time, I tend to distrust situations in which individuals act as mere passive data providers for others that collect and obtain value, as often happens with Big Data.