There’s no doubt that dating is evolving. With the advent of new technology, the way we connect and communicate with potential partners is changing. And, as our lives become increasingly busy, it’s no surprise that more and more people are turning to online dating. So, what does the future hold for dating? Well, we can expect to see more and more people using online dating sites and apps. This is likely to be driven by the continued growth of the mobile market, as well as the increasing popularity of social media. We can also expect to see more niche dating sites and apps appearing. This is in response to the growing demand for more specific and targeted matches. And, as our lives become ever more connected, we can expect to see more opportunities for meeting potential partners online.
Since the first lonely hearts advert in 1695, the first computer-based matchmaking service in 1965, the launch of the website Match.com in 1995, and apps Grindr in 2009 and Tinder in 2012, technology has been an integral part of the assortative process of dating and mating. Algorithms, often kept top secret by the apps, are used to match people, but nothing so far has been able to predict chemistry in real life.
Algorithms can’t predict the unique interaction of two people in real life
A study by Frost et al (2008) makes an important distinction between the experiential and searchable. Experiential attributes (such as humour, warmth and empathy) – which can’t be measured from a dating profile but must be experienced on a date – were three times more important to daters than searchable attributes (such as education, politics and music tastes), which can be categorised and used by the algorithms. There is a difference between the recipe and eating the dish, and when it comes to attraction and love, we need the full multi-sensory experience.
There are plenty of ways to mislead before you meet too, that the algorithm can’t fix. Daters tend to display a carefully curated ideal self on dating apps and in messaging, which is much less spontaneous than real life conversation. This is despite the fact that the ideal self is seen as less attractive and authentic by daters. The consequence is that the gap between the online and the offline self can lead to disappointment over your date’s appearance, conversation, or emotional availability.
Joel et al (2017), using a form of AI called machine learning, found that their algorithm could predict choosiness and how attractive you appear to others from self-reported traits and preferences, but not attraction towards specific others. As Joel explains, you can’t (yet) predict whether someone will laugh at your jokes using algorithms. However, Facebook dating matches you with people who have liked similar posts and expressed an interest in similar events, so it is conceivable that soon Facebook’s algorithm will be able to predict which comedy nights you might both enjoy. The quite startling demonstration at the end of this paragraph of an AI with an avatar shows its ability to understand and explain quite subtle jokes that rely on inferential chains, such as wordplay and analogies, and is another indication that soon algorithms will be able to predict whether we’ll appreciate each other’s jokes. (This uses PaLM, Google’s Pathways Language Model, which is a natural language predictor that uses deep learning to produce human-like text, trained on data from multiple languages.) However, it is at the time of writing better at understanding than creating jokes.
What data do the algorithms on dating apps use and how do they work?
The dating site Eharmony uses the Big Five personality assessment (openness, conscientiousness, extraversion, agreeableness, neuroticism – easily remembered using the acroymn OCEAN), demographics, and preferences (including, controversially, ethnicity). Eharmony’s claims that their psychological assessments used to match people were “scientifically proven” were banned by the Advertising Standards Agency because the studies they quoted did not show this. Although the studies showed that those who met on the app were happier than meeting through any other offline or online source, they did not reveal what proportion of their users had actually met anyone on there. So it could be that only 1% of users found love through eHarmony compared to 10% elsewhere. If you find love, it might be better, but it doesn’t make you more likely to find it in the first place.
Tinder – owned by the Match Group that includes Hinge, OKCupid, Plenty of Fish, Our Time, Meetic and Match – is the most popular of the dating apps worldwide, and keeps its algorithm a closely-guarded secret. Judith Duportail, a French journalist and Tinder user, spent two years investigating Tinder’s algorithm and the data it had accumulated on her. Journalist Austin Carr was granted access to Tinder’s system which uses a desirability score for each user, based on how many people have swiped right on them. The Hinge algorithm uses who you went on dates with (and whether either of you want a second date) to learn more about you and improve your matches. This behavioural data is more valid than our stated preferences, but if we’re still at an exploratory stage of dating a wide range of people casually, the data we provide might not be so useful for us when we later want to look for a relationship. It’s up to you whether you provide that information to Hinge, so you might want to decide on a case by case basis whether to do so – a chance for you to hack the algorithm.
Algorithms can be biased
Algorithms may use clustering to find similar others to those you have liked/swiped right on but this may lead to or reinforce bias. An example – given in this video – might be that when there is a small pool of Asian men and you swipe left on one, you may not see another for a while, because the algorithm decides the attribute of that person you didn’t like was their ethnicity, rather than because, say, that they like pineapple on their pizza. In 2016, daters from the Coffee Meets Bagel dating app reported being shown partners solely of the same race as themselves, even though they selected “no preference” when it came to partner ethnicity. This can also happen if the algorithm is trained on data from biased daters or if it rates attractiveness based on frequency, such as in the AI beauty contest, which was biased towards white people because the data provided was mainly of white people. (You might also want to listen to Joy Buolamwini’s TED talk on fighting racial bias in algorithms).
You can try to hack the algorithm as Chris Finlay describes in this very funny video, or you can “finish Tinder” using a script like this one from Matt Taylor, who after 25,000 swipes got 9 matches and 1 date (whom he then married – it only takes one!).
Algorithms may not find long-term compatible partners
Could it be that the algorithms are used to find people you are attracted to rather than those you are compatible with in the long-term, as this it is financially beneficial for the apps to keep you on them? The algorithm will also be using data that tells it who you are attracted to rather than our stated preferences for a relationship, because it uses your behaviour on the apps – who you match and message with – which is often just who you are superficially and initially attracted to. Unless you have done some work on intentional dating (perhaps with a dating coach like me!) you may not be aware of those unconscious processes going on, where we are using old blueprints for attraction that we could grow out of as we move towards being ready for a relationship rather than short-term dating.
It is also good to be aware that those preferences are often abandoned as soon as someone is actually interested in us – especially in real life rather than hypothetically – and that includes abandoning our ethnic preferences (on those apps where this is still an option). This means that we may be easily swayed by the excitement of a connection, even if the person doesn’t fit our dream long-term, and this could be good (such as when we broaden our horizons) or bad (such as compromising on our desire to have children).
In fact, it is very difficult to predict from a set of measurable data items whether a relationship will be successful, even using machine learning and longitudinal data. More effective is the extensive observational and physiological data collected by The Gottman Institute, which is able to very accurately predict the future of your relationship within a few minutes while you have an argument!
How else can technology influence our dating?
Apart from algorithms, there are other forms of technology that are becoming an integral part of dating. For example, you might end up messaging a bot. With a little practice, it’s easy to spot a fake profile and carry out tests on it. Can you spot the red flags in this example?
A video-date as a preface to a real-life date has become popular since the pandemic, and looks set to stay: more and more dating apps are providing this facility within the app (Bumble, Tinder and Our Time, for example); other daters use a platform like Zoom; and virtual speed-dating apps – some without photos to reduce superficial swiping – are emerging. A short video date before deciding whether to meet means you waste less time dating people that aren’t right for you.
Cognitive enhancement is another controversial area of technology for our love lives that philosophers Brian Earp and Julian Savulescu discuss in their book Love Drugs:The Chemical Future of Relationships: in the future, there could be doses of oxytocin to re-ignite love or some Ritalin or beta-blockers to focus the mind on a date. (Meanwhile, there’s always a glass of wine.)
As well as dating apps, there are apps to generate conversation starters for dates and in relationships as well as apps to help couples organise their lives. And AI therapists have been shown to be successful, so it’s likely there will one day be artificial dating coaches too. In the novel Lab-Grown Meat Bites Back by Christian Darkin there is a rather too convincing chatbot avatar matchmaker and international virtual reality dating. Apps could also guide your relationship and help you make decisions, like in my short stories Predictive Text and Compliments And Appreciative Platitudes Season Two. Personally, I think it’s more healthy to interact with a human one, but then I would, wouldn’t I?
Artificial friends and lovers are also part of the future of relationships. Already you can clone a voice after five seconds, generate a grief bot as a deep fake of someone who has died; and sex robots continue to develop (read Sex Robots and Vegan Meat by Jenny Kleeman or Science, Sex and Robots by Kate Devlin). There are of course pros and cons to these relationships: they could give isolated people opportunities to improve their social interaction and enjoy the physiological and emotional benefits of social connection. But they could also make people lazy about working on real life relationships or not challenge them when they treat their artificial friends badly. (Do you shout at your Alexa or Google Home?)
In my next blog post, I’ll be sharing how I used AI to generate original human-like text that could be used for your online messaging. AI language models (like PaLM described above) are so convincing that they are being used to write articles and even fiction. In fact, can you spot which bits of this blog post have been generated by AI? Find out in the next blog post here!