Personal Recommendations, Digitized

by Adam on October 20, 2012

One of the ongoing themes within the online world is the idea of personalized recommendations for ___________ (insert your interest area on the previous line). Everything from food to travel to book buying to pet care is being catalogued and classified such that an algorithm or piece of software can more easily get you what you want, and quicker than you could ever expect. Amid this dizzying array of widgets that are constantly recommending, honing profiles, and re-recommending, is the nagging question of whether we aren’t narrowing our field of focus so much that this personalization is actually causing us to miss out on a wide variety of content that we would otherwise discover in an unautomated world.

Take, for instance, this newly launched company, though in pointing them out I’m not trying to bring particular attention to them or their business model:

http://www.digitalbookworld.com/2012/booksai-wants-to-help-you-discover-new-books-through-artificial-intelligence/?et_mid=585289&rid=2642229

I have many issues with a model that espouses the use of technology as the basis for recommendation, primarily because my sense is that any such model fails to take into account specific use cases. For instance, on Amazon, if you wanted to buy a gift for a friend who happened to like bird-watching and you went particularly crazy and bought ten bird-watching books, wouldn’t Amazon’s algorithm rightly think that you enjoyed bird-watching yourself and as such would begin recommending products to you in that vein? And Amazon’s algorithm is particularly lackluster, so I imagine the issue would correct itself over time, but what about a more advanced system? A system that catalogued that book buying information and used it to inform the development of your profile on the site?

Herein lies the issue, because for all their work getting you to what you want faster, programmed algorithms are surprisingly inflexible in dealing with situations where you are buying for someone else, cooking for a large group, planning travel for business rather than family, etc. There are myriad situations where you need to be able to have one set of criteria that govern the decision, and then have a completely different set of criteria govern another decision.

When walking into a bookstore, you can say to the clerk there, “I’ve just read Confederacy of Dunces and want something similar” or “I’m in the mood for a smashing romantic comedy” or any number of things. What you wouldn’t do is walk into that store and have the clerk come up to you with a handful of books that are similar to the book you bought last time. That wouldn’t happen. Because it’s an inefficient and presumptuous way to sell books.

This model also says nothing of the lack of serendipity (which is a theme I’ll pick up on in a later post, influenced by a good friend of mine and someone well-versed in publishing) inherent in this model. When a machine recommends something to you based on things you’ve bought or read or consumed before, isn’t it inherently creating a walled garden of information that is relevant only to specific data points? How then are you supposed to be able to pull out different perspectives or experiences when everything you are bound to get in the future is based on all of those things you’ve already gotten in the past.

In the drive for instant gratification (whether in content delivery, or products at your door) we must remember there is always something we are missing by moving too quickly. If we don’t take the time to discover the content that truly matters to us, instead relying on machines to do it for us, aren’t we giving something up in not undertaking to understand why this content is important to us in a wider context and why it should matter?

 

{ 0 comments… add one now }

Leave a Comment

Previous post:

Next post: