Wednesday, March 19, 2008

User-generated content, incentives, reputation, and factual accuracy

Not all user-generated content is factually accurate and it does not have to be that way. I don't expect Wikipedia to be completely accurate and some how many people have a problem with that. Traditionally the knowledge-bases that upfront requires high factual accuracy have been subjected to slow growth due to high barrier to entry. Wikipedia's prior stint, Nupedia, had a long upfront peer review process that hindered the growth and eventually led to the current Wikipedia model as we all know. Google Knol is trying to solve this problem by introducing the incentives to promote the quality of the thoughtocracy. I haven't seen Knol beyond this mockup and I am a little skeptical of a system that can bring in accuracy and wild growth at the same time. I would happy to be proven wrong here.

For an incentive-based system it is important not to confuse factual accuracy with popularity of the content. If content is popular it does not necessarily have to be accurate. If we do believe that incentives can bring in the accuracy, we need to be careful in associating incentives to the accuracy and not to the popularity and that is much harder to accomplish since the incentive scheme needs to rate the content and the author based on the sources, up-front fact-checking, and not just the traffic which could indicate popularity. Mahalo is trying to solve the popularity problem and not the accuracy problem. There have been some attempts to try out the reputation model for Wikipedia but the success has been somewhat underwhelming. I see many opportunities and potential in this area, especially if you can cleverly combine the reputation with the accuracy.

In reality what we need is a combination of restriction free content creation, fact-checking, incentives, and reputation. These models are not mutually exclusive and not necessarily required at all the times and should not be enforced to all the content across all the users. Guided or informative content tend to be more popular irrespective of the factual accuracy since it is positioned as a guide and not as a fact. The people who are in the business of working off the facts such as media reporters, students working on a thesis etc. should watch for the content that is useful, looks reputable, current, and may be factual but is pure wrong and should go through a systematic due diligence fact-checking process.

No comments: