OTT streaming services like Netflix use machine learning and AI algorithms to learn about their audiences. With this info, OTT service providers can better understand what users like to watch. Factors such as how long a user watches a specific type of content and what type of films they typically watch contribute to more accurate content suggestions. While recommendation engines have come a long way, there is still lots of room for improvement.
The challenge of creating content discovery environments
The sheer amount of content available to viewers is overwhelming. It is said that a user can spend more than a thousand years watching all of the videos on YouTube alone. Other OTT services have less content to offer, and while recommendation engines do well in suggesting movie picks, people still spend too much time browsing.
Today, users want instant results. No one likes to waste half an hour choosing a movie or TV series, only to realize halfway through that it isn’t in line with their tastes -- yet in reality, we do it quite often. As the OTT market has become highly competitive, the solution to keeping users around lies in providing them with the ability to instantly select something they will enjoy.
Often times we talk to our colleagues and friends and we recommend content to each other, discussing the pros and cons, while also keeping in mind their preferences. A future scenario could be one where users ask Alexa or Google Assistant what TV series or movies they should watch, and then get an accurate recommendation.
Adding value to users through personalization
Families that are all connected on the same account naturally have different tastes. A TV show might be of interest only to one family member. Yet, recommendation engines of today are not able to differentiate between users unless they have multiple accounts. The next generation recommendation engine would need to be able to decipher who is using the account. Are they watching the content alone or together as a family? The difference lies in finding a movie or TV series which would be appreciated by one or more individuals.
To add to this level of personalization, OTT providers would also need to consider other data sources such as social media activity, more info about our interests and personality. The pages and profiles a person follows on social media say a lot about their likes and dislikes. Browsing history can lead to recommendations that follow a person’s interest at a certain time in their lives, like suggesting travel documentaries if they browse travel websites frequently. The more personal data we share, the better the assistance we’ll get.
Ethical considerations for accurate suggestions
As AI increasingly makes its way into our lives, the concept of having algorithms without bias is becoming a central focus. Algorithms for AI and machine learning software need to develop ethically, in a way that incorporates data from a broad range of users. If the software is built with data that is limited by geography, race, or gender, it will invariably produce an experience that excludes certain users.
But there are also ethical concerns when it comes to the intended outcomes of the recommendations that are produced. As an example, Google Jigsaw has been working on a project that learns if a user is watching too many morally questionable videos. If the user searches lots of videos about terrorism or racism towards a certain ethnic group, then the recommendation engine redirects them to videos which de-radicalize them or teach them about empathy and tolerance.
The future of recommendation engines is one that can deliver a dynamic experience on OTT platforms with an unprecedented level of personalization and user experience. If the proper environments for content discovery are created, then OTT providers have a golden opportunity to keep their audiences around for the long haul. Furthermore, today’s intelligent solutions companies will be able to act responsibly, fighting bias and promoting ethical social attitude.