- people are probably interested in others who are similar to them
- we want to respect our user’s privacy
- the suggested connections must have added value over random encounters
- we’re not a dating agency
Birds of same feather flock together: modelling Serendipity
Article by Tom Rijntjes
When you’re the host of a convention, who do you introduce to a person you’ve just met? Are the person’s knowledge and experience the main predictors of a valuable connection? How do you leverage technology to facilitate serendipitous encounters?
These are the questions we are facing. In a small trilogy of blogs I’ll shed light on the process of translating a grand vision into actual physical encounters using artificial intelligence. In the first blog I want to talk about why it is interesting to model valuable connections and which assumptions lie at the basis of our approach.
Matchmaking: it’s not optional
Why match people in the first place? Creating serendipitous encounters at Seats2meet is not optional; it’s the core business. If there’s no interaction, there’s no point in bringing together skilled people in a pleasant working atmosphere. Introducing people to each other is a traditional task of the host, but a human host is limited in her in depth knowledge of each person’s skills and current interests. I think this is true for one location, but this insight is much more important if the availability of people transcends the boundaries of one physical space to the vast international network Seats2meet is building. You can’t go about meeting 10.000 people to solve a problem you might have. But you can meet three.
Reducing the scope
It’s a huge challenge to prioritize the thousands of candidates to a workable quantity in a meaningful way. Meaningful, in this sense, implies that the proposed candidate is interesting to you. If you stop for a moment and think about the great variety of reasons you might have to be interested in meeting someone, you’ll begin to see why this is a rather complex task.
We established a few rules to reduce the scope of this problem:
Birds of same feather?
Machines typically aren’t very good at understanding interpersonal relations, but it’s becoming increasingly viable to create a workable model of one’s social tendencies based on online interactions like social media. I think this is invasive and not the point of a professional environment like Seats2meet. The first design choice is probably the most dramatic. To see why, consider the following story. Sometimes I have some beers with my friend Max, who is a philosopher. In many ways Max is the opposite of me in terms of skills and background but he helps me put my work and life in perspective. Although our friendship is truly serendipitous, I also believe our connection is of a rare breed, precisely because people don’t typically stick around without common ground. It turns out that connections based on non-similarity are very difficult to model without violating the third rule. If there is no common ground to engage in constructive conversation, the connection will lead nowhere most of the time.
In the next installment I’ll discuss how we model knowledge to predict valuable connections. In the mean time, please share your take on our approach to modelling serendipity in the comment section.