In a previous post, I mentioned that AI was changing relationships. In this multi-part blog, I want to delve further on this topic to describe how AI is changing marriages. I will focus on algorithmic matchmaking which is present in dating sites like e-Harmony. Because these online dating sites have been around for a while, the idea of using technology for matchmaking has gained some social acceptance. However, from a historic perspective, this is nothing short of revolutionary. For centuries we have resorted to family members, traditions and even our own whims for finding our mate. Now this is being done through Artificial Intelligence. While e-Harmony arranged marriages account for a small portion of marriages (only about 4% at the moment), given our dismal record on divorces, I see adoption of this method to continue to grow in the coming years.
Before criticizing this trend as another example of technological over-reach, it is important to understand how these applications work. To enter the company’s database, the spouse-to-be needs to answer over 150 questions about themselves. This includes personal preferences, demographic (gender, age, place or residence), and behavioral data based on website click-data. This data is then aggregated with other users and historical data. The data is fed into machine learning algorithm that then assign different weights to each factors making it possible to quantify how compatible two users are to each other. So for example, if the historical data shows that male Nascar fans tend to stay together with females that like Harley-Davidson bikes, then users with these traits score higher in compatibility and are more likely to be paired together. The model analyzes hundreds of traits in this manner producing a final compatibility score between pairs of users.
What is different about e-Harmony is that they actually track their relationships success and are able to feed that knowledge into their models. In other words, they are able to match people based on how successful marriages were for similar past versions of these new users. With 16 years of data, the models should have a solid baseline for compatibility. Also, unlike other dating sites, e-Harmony does not allow the user to look for dates. They are limited to the matched produced by the algorithm. That is, they trust that their model is better than the user him/herself in looking for long-lasting relationships. If their numbers are to be believed, divorce rate of e-Harmony couples is around 3.8%. Now, I would sure love to see whether these numbers hold out over a period of decades.
While I am not privy to all the variables and methodology used by the site, their approach seems sound. They are basically quantifying data available from both their user base and psychological research on the topic and making that wisdom available to relationship seekers. It is as if you could evaluate a large number of marriages at once to try to figure out what works and what doesn’t. Certainly, a computer would do that job better than just relying on one person’s experience. While we are limited to our own experience and others around it, models can aggregate results from thousands of people. I see this approach expanding beyond dating site to becoming an enhancement for matchmaking in the offline world.
This does not mean all marriage partners should be chosen through this process. The model’s foundational idea is that marriages live or die because of compatibility. That may not be the deciding factor for all marriages. Also, people change in ways that questionnaires cannot capture. Moreover, even if the approach proves out effective over-time, I don’t see entire societies changing century-old traditions to adopt this new way of choosing mates. Technology changes fast, people not so much.
This is the first part of this blog where we explore how AI is impacting relationships. In this case, we have an example of AI altering relationship between humans. The next level of impact is how AI is reinventing machines interfaces in a way that may displace human relationships. This is AI not enhancing but actually replacing human interaction.
This is when things get complicated.
