In [Gru07], Tom Gruber states that:

The web, as a community, is not yet a “collective” intelligence, rather a “collected” intelligence. This comes from the fact that there is no new level of understanding. User-generated content is being shared, gathered and collected in domain-specific sites. We can find what things are more popular or what are the current fads. However, while popularity is one measure of quality, it is not a measure of veracity. Mass authoring is not the same thing as mass authority.

This classification of collected vs. collective intelligence of the web renders a definition of what a Collective Knowledge System can be and its key properties are summarised below:

• User-generated content. The bulk of the information is provided by humans participating in a social process. A traditional database or expert system, in contrast, gets the bulk of its information from a systematic data gathering or knowledge modelling process.
• Human-machine synergy. The combination of human and machine provides a capacity to provide useful information that could not be obtained otherwise. These systems provide more domain coverage, diversity of perspective, and sheer volume of information than what it could be achieved by searching official literature or talking to experts.
• Increasing returns with scale. As more people contribute, the system becomes more useful. The system of rewards that attracts contributors and the computation over their contributions is stable as the volume increases. In contrast, a text corpus and simple keyword search engine does not get more useful when the volume of content overwhelms the value of keywords to discriminate among documents. Similarly, if the reward system encourages fraud or fails to bubble up the best quality content, the system will get less useful as it grows.
• Emergent knowledge. The system enables computation and inference over the collected information, leading to answers, discoveries, or other results that are not found in the human contributions. This fourth property is what differentiates a collective from a collected knowledge system.

Conveying these key properties, a Collective Knowledge System can be composed of the following elements, depicted in Figure 1:

• Community of motivated people with problems and solutions. These contributors share their expertise and knowledge on the specific domain.
• Larger population of intelligent people with similar problems. Who actively search for personalised solutions to their problems.
• Computer mediated social communication. Whether through tagging, blogging or commenting, the social process is augmented and nurtured.
• Semi-structured information repository. Acting like a storage facility for a more long-term memory and where the solutions are collected and shared.
• Socially clustered data knowledge-base. Where the solutions are catalogued and clustered according to the social interaction and multidimensional analysis.
• Faceted search engine. So that the solutions seekers can look for personalised solutions, through contextual browsing.
• Recommendation engine. To keep the users in perspective and assisting in obtaining more rapid and effective answers to their specific issues.

It might be relevant to say that not all of these elements need to be present for a system to be considered of Collective Knowledge. At least, the knowledge quality evolution through social interaction and its access need to be enforced.

< back to start page