Human vs. Algorithmic Recommendations

This morning, I was reading a really interested post about a couple of startups that are building algorithmic engines that look at past and current views and clicks on a given site in hopes of serving you better content/offerings/ads. Amazon has done this for years, though their recommendations are certainly far from perfect. These startups also claim to have predictive value by determining what the user will buy in the future.

This got me wondering about whether recommendations from humans are still inherently more accurate and useful vs. algorithmic results? Perhaps it depends on what web category we're considering.

For example, let's supposed I'm looking for a job that is most likely contains some form of the title "Director of Product Management." And, I'm really picky about who I work for and want someone who is a clear communicator, strategist, and who would be a good mentor. I sign up for Jobster alerts and peruse CareerBuilder, Monster, Linkedin, etc for jobs that might be interesting.

In this case, there's no way an algorithmic recommendation will be of much value to me. For starters, the description itself is often inaccurate but you don't know that until after you've clicked and read the actual description (and read between the lines). That data point is captured by an algorithmic engine and presumed to be valid, and it will be used to refine its result set. How many times does a recruiter from Microsoft enter an entry level product management position but tag it with descriptors like "Director of Product Management?" The answer is: all the time! How many times is the description not accurate and very aspirational (but not very realistic)? All the time. Finally, how often does the information you really need to find the right job not even exist on the web? All the time. Yet, with an algorithmic engine, they think they understand what you want by analyzing your surfing when in reality that surfing hasn't produced anything of value. In this case, knowing somebody that knows the company, the position, and the hiring manager (potentially, your next boss) can offer far more relevant results. There are probably lots of other examples where human intelligence provides far greater insight than anything devised by an algorithm.

The problem is, how do you tap into this human intelligence, find people that have the right information, get them motivated to share that information, and then actually have them share the information? Social networks, email, blogging are all obvious tools that can help unleash this information to those who would benefit.

We still have a long way to go before this algorithmic recommendation system works properly. I, for one, look forward to its perfection because I am tired of Amazon recommending book after book on road racing based on a book I purchased for a friend's birthday two years ago.