Gesellschaft | Technology

The moral of the recommender

The critical thought behind the machine. A conversation on the ethical and social implications of recommender systems with Joseph Kostan.
Hinweis: Dies ist ein Partner-Artikel und spiegelt nicht notwendigerweise die Meinung der SALTO-Redaktion wider.

What is a recommender system? Probably the vast majority of readers do not have the slightest clue what this word could stand for. In reality, it's a very common technology allowing users to find personalized content in a website. Each one of us uses it day by day surfing for example on youtube or Amazon without even being aware of it. We asked Joseph A. Konstan who teaches Human-Computer Interaction and User Interface Technology at the University of Minnesota what implications could such technology have on our societies. Konstan is one of the most renowned professionals in the field of recommender system andrecently invited to hold a lecture at the Faculty of Computer Science of unibz.

 

What are the newest trends in recommender systems?

I think that the biggest trend that affects user experience right now in the field of recommender systems is in the area of context-sensitive recommendation. People are still delighted and amazed by the ways in which a recommender maybe aware of where they are, with whom they are, what they are doing. Sometimes this can also be intimidating because of privacy issues. It seems that the recommender system knows too much about you. At the same time, it can be a wonderful positive experience when you get recommendations that really match the things you are interested in right now. It has this double effect. 

Recommender systems could also have important social and even political effects:  they can connect me with issues and products which I would not normally look for.

I think this is a huge issue, particularly in countries where politics is extremely polarized. The question is: Can a recommender system move broader than just recommending things that are similar with what you have already liked? Would that be possible if requested? Can you say, “I really want to be aware of what all parts are saying on this issue”?  Does a recommender have a responsibility in some contexts to present that information if it is not what you like? Is there a societal good that says, “We understand you’re interested in this, but you should be aware that there are opinions- and news that say the contrary”? It can be scary in and politically sensitive, whether it is Holocaust denial or Climate Change denial.

Does this problem have a solution?                    

Technically speaking it is very easy to solve. The social problem is larger. I do not believe that at this point it makes sense to outlaw somebody forming a collection of content that has one point of view. I do not know a better solution than education and hopefully good reputation from sources that make a point of exposing multiple points of view.

Are there recommender systems in non-profit contexts?

There are many opportunities to find nonprofit uses of recommender. Some provide information content: Wikipedia is a perfect example. If you are an educational non-profit organization there are many cases where recommender are being used in technology-enhanced learning, helping people find online courses, Coursera for instance. But the bigger challenge is to do this in a very small context.

Bild
Profil für Benutzer Roland Kofler
Roland Kofler Mi., 09.12.2015 - 15:23

"Welcome Back" sagt Konstan bei jedem Video im Coursera Kurs zu "Recommender Systeme". Wird vom TF_IDF Filter ignoriert.... insider Joke.

Toll das mit Prof. Ricci und Prof. Konstan zwei Koryphähen auf dem Gebiet sich treffen.
Ich schreibe gerade einen (kommerziellen) Artikel wie man Elasticsearch als Recsys verwenden kann. Vielleicht kann mir die UniBZ da helfen ;-)

Mi., 09.12.2015 - 15:23 Permalink