I did the initial 20 questions, plus 20 more. And what surprised me was the US corporate mainstreaming of advertising that it spit out. There was no cultural recognition that I was a Canadian, hence all I got were suggestions for American Magazines, books, toys (Lincoln Logs).
So I cynically concluded that this is nothing more than an attempt at targeted US advertising. They must have some really interesting algorithms running in the background to conclude that I should be a Science Fiction geek, yet interested in Freakanomics, National Geographic, etc. Since I work in semantic projects and we do our own interesting correlation algorithms to instantiate instances into the correct classes (Like hunch is doing, taking my answers and pigeon holing me into a "class" of complex makeup with "rules" to suggest appropriate books, magazines, travel destinations, food/recipes, cars, and TV shows.
Yet, with all that great algorithms, it missed my nationality, the fact that I am interested in Article level subjects, not journals, that I actually have two jobs, one in Canadian Military history, the other in knowledge management, etc. So in essence, their algorithms did a very poor job of really giving me the recommendations I wanted. That is why I do not use Hunch, rather I use my personal networks I have established on Academia.Edu, use Google Alerts, Linked in, Facebook, and Twitter to find out the information I am interested in.
By the way, I did subscribe to National Geographic this Christmas for the first time in 30 years, I have been meaning to look at Freakanomics (but it is about book 200 in priority), and I would love a Tesla Roadster (but would prefer a 1986 Porsche 928 Turbo) but both would not do well in New Brunswick, Canada winters.
SOOOOOOOO.... data interpretation/recommendations is only as good as the questions that are asked, the responses received and the assumptions black box algorithms and rules are based on (i.e. cultural/age).
I have the same opinion regarding lack of cultural recognition.
What I was wondering is that a machine can effectively recognize who you are and what you may be interested in. I was about to conclude that the problem was that a machine can't do this, and then I remembered GoogleAds, that are AMAZINGLY accurate almost all the time.
And stumble upon... I almost always like the websites it suggests.
In sum, I don't know anything about algorithms to provide suggestions, for me Huntch wasn't accurate at all [I didn't click on the suggestions, maybe I would have liked some of them] but I think the rejection on our part is more on a cultural bases than in any other reason.
@Mary, I'm with you on this one, the site is a marketing tool pure and simple which bothers me more than a little!
What surprised me was, silly me, I kept getting questions and I kept giving answers and I didn't see much change or improvement. I can't say that I answered 20 more and I was bothered that I had no indicator of progress while all these questions were coming my way.
I'm not likely to use hunch. IMHO my tastes are so eclectic that it is never going to be a big help. Some of its suggestions were also just plain "duh" such as recommending the Daily Show when I had selected Jon Stewart as a personality choice along the way.
I think your point about not seeing "much change or improvement," which I found to be fairly true if when only cycling through the home page.
However, when I went about two clicks into a particular category, where there is a top ten list of recommendations on the right side of the screen and the question box in the upper-left corner of the screen, what was interesting was watching some of the results change in real time as I answered more questions.
I don't know if you experienced anything like that when you were tinkering with it, but I found it kind of fascinating to see things change as I continued answering questions.
I too have pretty eclectic tastes and it can be pretty hit and miss, but when the recommendations are on target, they are surprisingly decent.