IEEE Technology and Society Magazine - December 2016 - 42

something linked to the representations they have
already consumed. Cinematch is engineered to reveal to
subscribers something they would want to see, but they
didn't know existed. For de Lauretis, there is revolutionary potential in collectively turning our attention to
whatever is off the radar, because the space-off is composed of material that will help people better
understand their world - their filter bubble, as it were -
and what it is filtering out.

A high-quality video is not just a video
that gets the most views, but a video
that has a high proportion of complete
views or repeat views.
Determining Quality
Regardless of the social implications, there are clear
business imperatives to develop systems that will recommend appealing new content to audiences. YouTube's goal is to get visitors to watch YouTube like
they watch TV, rather than watching one short video
and moving on to another site. It is in YouTube's best
interest to introduce calculated serendipity into their
recommendations so that viewers will watch lengthy
blocks of varied content that have been personalized
for them.
Look at the parallel in successful TV programming
blocks. A network would never schedule three
extremely similar comedies together: the topics, settings, characters, and locations must be different. The
unifying feature is more often about pace and tone,
which can be very difficult to codify and make keyword
searchable. The successful ABC programming block,
Thank God It's Thursday (#TGIT), features three shows
by Shonda Rhimes: a political thriller, Scandal, a medical drama, Grey's Anatomy, and a legal drama, How
To Get Away With Murder. All of them exhibit the signature characteristics of a Shonda Rhimes' show, but
in very different contexts with diverse ensemble casts.
One thing that YouTube has tried to do is determine what the indicators are for quality - a serious
problem for a company that trades in user-generated
content. They have determined that their best indicators for quality come from the analysis of human
behavior. A high-quality video is not just a video that
gets the most views, but a video that has a high proportion of complete views or repeat views. More specifically, the videos that they want to recommend to a

visitor at any given moment are the videos that many
other people watched all the way through, or repeatedly, after watching the video that a visitor just
watched. By mining data about real human consumption habits, YouTube can discover patterns in preference that do not boil down to simple characteristics
like affection for a certain actor, franchise, or genre,
which in many ways, still drive motion picture and traditional TV development.
Netflix looks at similar data points, of course, paying
very close attention to how long people watch, where
they pause, what scenes they play over and over, and
which videos are played all the way through. This is all
in addition to the ranking feedback that viewers offer
them, which can be skewed by the desire to demonstrate highbrow taste. For instance, a subscriber may
give best picture Oscar-winner Spotlight four stars, and
only watch it once, and then give three stars to Melissa
McCarthy's comedy Spy and watch it four times.
Since Netflix is a subscription service, Cinematch is
crucial to their business model because new members
usually only have a couple dozen movies that they
know they want to see. In order to keep these people
as subscribers Cinematch must tell them about titles
they have never heard of and that they subsequently
like. Various reports say that between 50 and 75% of
Netflix viewing is driven by recommendation engine
suggestions, while Hulu reported in 2014 that it
accounted for 75% of their views.

Nostalgia vs Novelty
No matter how much data there is, it is no easy task
predicting what someone might like. Netflix, YouTube,
Hulu, Spotify - all of these companies face the same
challenge people face when they need to buy a gift for
someone they have known for some time: what can be
offered that is different from what was given before, but
still be a safe bet? It appears that novelty is not always
the key to success. One research study on maximizing
profits for movie recommender systems found that
movie consumers prefer to watch, and even pay for,
movies they have already seen rather than movies that
are new to them. One of Netflix's delicate challenges is
figuring out when appealing to nostalgia - offering an
oldie but a goodie - is better than suggesting something shiny and new. Indeed, the vast majority of their
rentals are from the backlist - older movies, old TV
series, classic films, and documentaries, as well as
small indie films that probably never made it into theaters in most towns.

Creating Cosmopolitans
For Netflix to be successful it is in their best interest
to have a cosmopolitan user base who appreciates a

IEEE TEchnology and SocIETy MagazInE

∕

DECEmbEr 2016

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2016