Don’t be afraid of new technologies: be curious!

Using machine learning to identify relevant content for users

Machine learning technology underpins Twitter’s content selection and discovery features – features that lie at the heart of a platform that enables its users to stay informed about what’s happening in the world. “Close to 500 million tweets are created each day on Twitter. They range in subject from the mundane, like someone’s ride to work on the subway, to substantial news updates from professional publishers like Le Monde, to a topical alert about an earthquake in Indonesia from a local resident,” JP explains. “One of these tweets could be critical to you, so the question for us is: how do we make sure that you see it?” Over late 2016 and 2017, Twitter introduced a number of new features to ensure users would see the content most relevant to them, all underpinned by machine learning technology. For example, tweets always appeared in the timeline based on recency: the most recent tweet created would appear at the top of a user’s timeline. A year ago, Twitter introduced “In case you missed it” which displays in your feed a set of tweets selected algorithmically based on user interest and behavior, so users don’t miss the tweets most relevant to them even if they don’t check the application frequently. More recently, Twitter upgraded its “Explore” tab with “Today’s Moments & Trends” which also uses machine learning technology to surface tweets for you around specific topics. “Machine learning algorithms have an internal feedback loop that enables them to get better over time at deciding which tweets are relevant to you, based on multiple signals that you send as a user: who you follow and what interests you have, i.e., what tweets you engage with in terms of likes, shares, and replies. These algorithms also take into account the quantity and velocity of attention around a tweet from other users,” reports JP. “Over the past 3-4 years, we’ve seen these technologies develop rapidly, and we’re investing more and more into them.”

Turning negatives into positives with help from customer service bots

“People often use Twitter for customer service,” explains JP. “You contact a company on Twitter, and the company tweets back at you in a private direct message. Even when a customer service interaction starts out negatively on Twitter, with a complaint, we find that very often the user will finish the interaction by publicly tweeting ‘thank you’ to the brand, because of the successful resolution of their problem via direct message. From a brand perspective, these interactions are actually great, because it gives brands the chance to publicly transform a negative into a positive.” Twitter is investing in using bots to automate and enable these types of customer service interactions on their platform. “Bots have the potential to significantly increase the scale of customer service activities on our platform, enabling companies to respond to a far larger number of inquiries,” says JP, who confirms that customers also often prefer chatbots thanks to their superior speed and accuracy. “Bots are still nascent. We’re still in the testing phase,” JP reports. “But one day in the not too distant future they will be able to answer even complex customer inquiries very effectively.”