Once perceived as an odd and somewhat dangerous way to meet people, the internet has become the go-to way to meet like-minded people in our own areas. Whether that be through dating apps, in the gig-economy where we can hire strangers to do jobs for us, or staying around empty houses while they are away, our acceptance of this medium has grown leaps and bounds in recent years.
However, this acceptance has been down to the increasingly more sophisticated recommendation engines which offer safer, more nuanced methods of interactions. One such company exploring the limitations of what and who we can find online is Meetup. Meetup is a service which allows like-minded groups with similar interests find each other online but meet each in real life.
Ahead of this year's Machine Learning Innovation Summit at this December's DATAx New York festival on December 12–13, we spoke to Meetup principal machine learning engineer Zachary Cohn to find out how he is using machine learning (ML) to better bring people together.
DATAx: How do you see ML further transforming the way we meet new people online?
Zachary Cohn: People are increasingly digitizing all aspects of their lives. As we instrument more and more of our interactions, that data makes ML increasingly valuable. One of the areas where I expect to see some change is explainability; seeing those machine-driven decisions become more transparent and interpretable by end users. For example, when Meetup recommends the group NYC Machine Learning to me it makes sense. It's even better if it explains that the recommendation is because of my topic preferences and it says that the next event is hosted by a speaker from a previous event I liked.
DATAx: As an ML specialist, what are the greatest challenges you face trying to create what is essentially a recommendation engine for people?
ZC: Most of the ML challenges are the same, almost regardless of the domain; getting the right data in the right place, monitoring performance and scaling everything (industry tools are getting better but a lot of this remains bespoke). At Meetup, we try to be especially mindful about the human elements of our recommendations. For instance we take into account the societal biases that are in our data and make sure the reasons we might recommend a group or event are consistent with our values. For instance, joining a men's trail running group should result in recommendations for running or hiking rather than groups that are disproportionately men.
DATAx: How have your team's priorities changed as Meetup and social media in general has grown in popularity?
ZC: At Meetup, we're trying to make it easier for people to connect and build in-person communities to have those richer, face-to-face interactions. Social media's rise and the increase in screen time around looser communities has highlighted the need for enabling in-person interactions. If we can do a better job of connecting people by their interests, activities, concerns or passions, we can help create a world with more real community. Now that people are connecting more online, we are trying to solve how to get those connections to take place offline.
DATAx: How far do you envision the possibilities of what a recommendation engine is capable of?
ZC: Recommendation engines today already add a lot of utility, but there's still a long way to go to make the technology feel like the experience of a conversation with a trusted, empathetic expert. And we shouldn't lose sight of that fact. "People who liked this liked that" is a powerful and effective paradigm but its ubiquity also comes from the fact that it's computationally efficient. In time, we'll see more and more complex, conversational search and discovery experiences.
DATAx: Have you begun leveraging blockchain technology in your enterprise? If so, in what ways?
ZC: Not yet, but we are always looking at ways to innovate Meetup.
DATAx: What will you be discussing in your presentation at DATAx New York?
ZC: Matching newly created Meetup groups with interested members is a key part of Meetup's platform. I will be sharing the practices we followed to modernize Meetup’s approach to recommendations: How we build and operate a robust and scalable system that notifies millions of members each day, how we measure and monitor that system, and how we apply an iterative product development cycle to best improve user experience driven by machine learning.
Zachary Cohn will be speaking on Day One of Innovation Enterprise's Machine Learning Innovation Summit, part of DATAx New York, on December 12–13, 2018. To attend and hear more great insights from AI and ML professionals across some of the biggest and most influential organizations, register here.