Over the weekend I was reminded of how, in the early days of the Net, Google supplanted Yahoo as the place to find websites. And that made me think about the current e-discovery predictive coding discussions.
The web exploded around 1995. Yahoo! was among the first success stories helping users locate useful websites. And I mean “locate”, not “search”. Yahoo!’s approach was a directory, a human-built, nested set of menus.
Good idea – until search engines came along. Google invented an algorithm that used the number of inbound links as a “voting engine” to rank sites. That blew away the directory idea. The automated approach beat the human approach.
A similar story will likely play out in e-discovery. Published studies support what many of us have long known: humans are neither particularly accurate nor consistent. See, for example, Predictive Coding: Reading the Judicial Tea Leaves in today’s Law Technology News.
Human review of discovery documents is like the Yahoo! approach. Fine for low volume. Not so for high volume. We already see today the challenge of scaling.
This does not mean humans cannot sometimes pick better websites / documents than computers. It just means that as volumes grow, the latter are more reliable / consistent than the former.
Ordinary internet users figured this out in a year or two. How long will it take lawyers and judges to figure out that a skilled humans using smart software are better than the brute force of an army of humans?