- It questions fundamental differences in how we can define people and predict future behavior: self-reports and preferences vs. behavior.
- It highlights the paradox of how social seemingly individualistic things are: The social nature of decision-making; not to mention, the very collaborative style of the Netflix' competition.
- It touches on the limitations of computers in understanding nuanced human decision-making.
My favorite things!
2 comments:
I loved the piece too, but mostly because it helps to undermine the black box approach to recommendation engines (and, by analogy, to relevance) that holds back progress in information retrieval.
My take: The Napoleon Dynamite Problem
hey daniel- thanks for stopping by. The black box approach only resonated with me in so much as it highlights the common problem of factor uninterpretability (via SVD), or "categorization obscurity" because of inexplicable human choices. This is the exact reason I'm such a fan of Endeca-- you guys are wizards at capturing those deep, complex relationships. After reading your post, I realize there's another element to the black-box approach from the algorithm side. In this case, I too err on the side of transparency, precisely because that metadata of 'process' (among other social cues) is far more compelling than a rating per se.
Post a Comment