This is not to say fully automated recommendation systems will not become more clever, and I do use some of the smarter ones but I think this is a different topic on personal v. cold recommendations.

So, question. Do you go straight from netflix recommendation to queue or will you google the film to get a second opinion?

If you’ve ever done the latter then you dont -trust- the recommendation. I would go as far to say if you trusted their recommendations 100% you would just have them automatically added to your queue and never log in to Netflix, right?

Anyway, the beauty of recommendation systems is that it’s just that: a recommendation. It’s not a final decision, it’s a suggestion of something you might be interested in. The thing is, the better the recommendations become, the more trust the user places in the results.

And specific to Recommendations. If a user were to trust the recommended results 100% then that user accepts the machine knows best. Why isnt everyone using ‘Im feeling lucky’? Users have genuine questions about the quality of the recommendation, who has had a vested interest in the results you’ve received. And quite frankly, a lot of automatic systems suck, pulling down the value of the better ones.

We can take this in to email but I feel its still within topic and of value to Bokardo. Let me know - pauric

Why will automated recommendation systems never have the trust of users?

I’d love to continue this conversation, if you want to do it via email drop me a line.

As it is they dont even have their ‘recommended’ ads up there.

From your post I would say that Yahoo tried to build the ‘internet menu’ by hand, allowing users to search their menu. Google used a system that could be viewed as recommendations to push more popular items to the forefront of the menu.

However, personally I consider a recommendation system to be something with an element of what UIE call Inukshuk. Otherwise you have something like Walmart ‘recommending’ Planet of the Apes to buyers of a Martin Luther King video.
http://www.rashmisinha.com/archives/06_01/walmart-recommender.html

Automated recommendation systems, google search included as perceived automated, will(?) never have the trust of users. So my question then is, leading on from your post, should pagerank be called a recommendation system if users dont think of it as such?

Or does it really matter?

“You see, Yahoo! was building an internet directory/search engine where everything was categorized by hand. It’s easy to see now that this is a near-impossible feat with the millions of new pages that pop up daily. What Google figured out is that using the hyperlink as a kind of vote of attention, you could start to organize the world’s information. Essentially what Google did was seize the opportunity to build a smarter search engine by enlisting the attention data of web-users in the form of hyperlinks.”

Just thought I’d try to clear that up.

If a system had 10 pages would you build search functionality to aid the user’s navigation? Scale up to 100, I would focus effort on high level menus and search. Up again to Amazon, there’s a menu but the site is so large most people just use the search. Now scale up to the internet ’system’. Is there any point trying to build a tradional menu for the internet?

Google is a pure search/navigation interface to the internet. Thats a fairly obvious statement, but think of it from a user interface perspective. They’re ’start’ button for the net with recomendation functionality that dynamically builds a menu list to information available on the system based on your search criterea.

If users of google dont perceive it to be a recommendation engine, then its results arent diegested as recommendations. People arent feeling all that ‘inukshuk’ goodness. Perception is 9/10ths of the law in UI world. (I just made that up but it sounds good, from ‘possesion is 9/10th..)

When you think of google as the user interface to the internet you see why they avoid being a portal, why http://www.google.com is simple and an insight in to the thinking behind a lot of their ui design decisions. Could you infer this from the Googlemendations standpoint?