Most of us are familiar with the worst feeling in the world: being madly in love, or as science likes to call it, limerence. Feelings of insecurity and uncertainty, the longing for reciprocity and fear of abandonment characterize this state of mind (I admit, there are some nice things about it as well), and it’s probably with good reason that when Googling ‘limerence’, the top search suggestion is ‘limerence cure’.

So why do we fall in love with people, instead of more predictable and reliable agents such as robots? Well, the standard line of reasoning goes, you fall in love to make babies, and you can’t make babies with a robot. We are genetically predisposed, shaped by natural selection, to be attracted to others that we think we can have healthy babies with. But please join me in a thought experiment: what if robots would be indistinguishable from humans? Well, you could probably fall in love with a robot then, just as you can fall in love with an infertile person, for example. But how human-like should robots be in order for us to fall in love with them?

Very simplified, there seem to be two factors that are essential for limerence to sustain: (1) an attraction to someone’s appearance, and (2) positively appraising personal characteristics, such as being smart, funny and some amount of unpredictability.

This means that first of all, our object of affection should be attractive, meaning it should at least look human. This aspect is quite easy to get wrong: the uncanny valley effect shows that we are easily repulsed by agents that look like, but are still distinguishable from humans. However, rapid advances are being made in humanoid design that may overcome the uncanny valley in the next 15 years. Having a robot that is physically indistinguishable from a human may satisfy the first requirement of limerence, however, we all know that being in love is very different from just being sexually attracted to someone (really!).

The second requirement is possibly much harder to satisfy. Being smart or funny requires some sort of intelligence, and robots as we know them are not exactly known for their sense of humor. The challenge of recreating human cognition in digital systems (known as strong artificial intelligence, or AI) has been a topic of research since the 1950s. The idea is that only when you are able to recreate something, you really know how that something works, and we really want to know how human cognition works. Since, we have made great progress in areas we long thought were typically human, such as face and speech recognition, and more recently even autonomous cars requiring the integration of all sorts of information. But we don’t fall in love with good face recognizers, and we don’t call someone smart if they can merely understand what we say. If we want to build robots that are truly human-like, we need artificial intelligence to see humans as more than really fast information processing machines. We need AI to be creative, entertaining and assertive. Only recently have we begun to study creativity and humor in artificial systems, with algorithms now being able to recognize ‘knock knock’ jokes as well as the even funnier ‘that’s what she said’ jokes. In the area of creativity, computer programs can compose scores of music with some limitations, and have even demonstrated jazz improvisation.

With robots becoming more advanced, it is necessary to take a step back and take a critical look at the societal consequences of such advancement. Falling in love with a robot seems to be inevitable in the future. However, with more and more physical dependence on robotic systems, we need to ask ourselves if we want to become emotionally dependent on them as well. As history shows, what we want does not always seem to matter. Our relationship with robots will likely become a love-hate relationship. Just like relationships between real people.