Online Catalogs and Users
In 1999, Barnes & Noble decided to accept customer reviews of the books it sold on its website. I had to manage that. Customers would come to the site and post comments on the books they were reading - the VPs at the time thought it was a great way to get content onto the site without editorial staff to create it.What a monster.
Yes, we set up filters to catch posts with offensive or hate-based language. But you can't catch everything. And it wasn't long before B&N found itself on the receiving end of letters from various lawyers, citing various posts as defamatory, etc. "So-and-so stole my idea" is not going to trip any filters.
Machines don't catch everything. You need humans to look over your content.
These days, it's tag clouds. Look at any tag cloud and you'll see the results of users managing their own data with little thought as to how other users are managing THEIR data. On Technorati today, for example, there are tags for both "John McCain" and "McCain" - each of which gives you different search results. Shouldn't they be combined into a single tag? The tag cloud on I Can Has Cheezburger is a right royal mess - there are 3 or 4 tags for every concept (something I find particularly irksome because I use them in presentations). The tag cloud on LibraryThing is nearly unusable.
You need human intervention for standardized, meaningful data. And yes, humans are costly. And when you're starting something new, you don't want to shell out for humans when machines are "good enough". But machines will never be entirely good enough. And you will always, always, ALWAYS wind up paying more later - in lost sales/clicks, in frustrated customers, in hiring someone to come in and clean up your data when it finally gets out of control, in hiring someone else to come in and set up a new system that institutes business rules in a data-entry form so that users can't screw up as much.
In other words, the best correction for human error is...other humans.
