Thanks to New World Notes, we get a little more detail about Philip Rosedale’s LoveMachine system, the reputation-based closed economy we mentioned yesterday.

Cory Ondrejka was Linden Lab’s CTO until 2007, and he was instrumental in developing the LoveMachine system as it operated within the company; here he is explaining a little more about how it came together:

One of my tasks was to invent a new system for employees to give each other feedback, one that would be fun, so easy everyone would use it, and that would generate interesting aggregate information about how individuals and the company were doing.

The design that emerged was tipping.

Tipping — via an internal web tool — would be a positive-sum, transparent game, a way to publicly thank a fellow Linden for going above and beyond. Finding a crucial bug, crunching some extra numbers, helping you figure out the right person to take a question to. Think “Twitter plus $1.” The key was to make it a small amount of money, as a payment makes it real but you don’t want to distort behavior with meaningful payouts.

Tipping was designed to solve three problems: help Lindens know what their fellow employees were doing, generate aggregate data on connections within the company, and identify extreme outliers. It wasn’t clear to me if your tipping rank would be important, but it might be meaningful data if you were generally at the top or the bottom of the list.

He also has suggestions on the problems that LoveMachine – or similar systems – may need to overcome if they’re to be of genuine utility:

The challenges that emerge, of course, fall into three broad categories. First, we optimize for what we measure, so unless you know what you are measuring exactly aligns with business goals, there are going to be misalignments. At Linden, people wrote tools to make it easier to use The Love Machine by irc, chat, email, and the web. This created “pile-on voting”, where an employee would thank someone and other employees would also deliver love to the recipient. This made the amount of love received a function of the time of initial delivery and the communication channel used, which may or may not have been desired. Second, people don’t like just being numbers, they want to understand what they can do to improve, so while The Love Machine should provide additional context for peer and manager feedback, it clearly can’t replace those conversations. Finally, with a transparent system like the Love Machine, are those ranked at the top retained? Are employees who leave or who are fired near the bottom? If not, you may introduce more communication and management overhead rather than reduce it.

If there’s one thing that can be said for certain about reputation engines, it’s that they’re not going to be an easy fix. I guess we’ll only really find out if they work once someone builds one successfully…