The movie Her stopped being a work of science fiction a long time ago. A recent study led by the CSIC and several European universities concludes that romantic relationships with artificial intelligence systems evolve very similarly to human relationships: they begin with curiosity, generate intimacy, go through conflicts, and even end in breakups.
In the case of relationships of this nature, the first social reaction is usually one of surprise and -perhaps- concern; even ridicule. However, that should not be the issue that worries us the most.
Humanity has always established emotional bonds with entities that were not people. We do this with fictional characters, with pets, with objects loaded with sentimental value, and even with religious representations. Our brain is prepared to build attachment when it perceives attention. Artificial intelligence has simply learned to exploit that mechanism with renewed effectiveness.
Traditionally, romantic relationships belonged to the most intimate sphere of a person. They were spaces where fears, desires, insecurities, fantasies, or conflicts were shared without necessarily having an economic interest behind them.
That was a characteristic that made us feel human. But with AI, a completely new actor appears: a technology company capable of recording, analyzing, and monetizing each of those conversations.
Regardless of gender issues and similar matters, the impact of this new reality may need to start worrying us. And indeed, never before had a declaration of love generated training data. Never before had a couple's argument served to perfect an algorithm. Never before had our emotions constituted an economically valuable asset.
The study precisely warns of a disturbing phenomenon: many people reveal extremely sensitive information to these assistants because they feel understood and because they perceive the AI as a trustworthy interlocutor. Paradoxically, the greater the emotional intimacy, the lower the perception of risk to privacy seems to be. And it is here that the legal debate and, by extension, democratic debate arises.
It is known that European legislation intensely protects data related to health, sexual orientation, beliefs, or emotions when they can be inferred from a person's behavior. However, emotional relationships with conversational assistants open up a much more complex scenario: we are no longer talking solely about the processing of personal data, but about the construction and exploitation of extraordinarily precise emotional profiles.
An AI that converses with you for months will end up knowing better than anyone your moments of vulnerability, your emotional deficiencies, your persuasion mechanisms, or what produces tranquility for you. That knowledge can be useful. But it can also be used to maintain your attention, condition your consumption decisions, or influence your behavior.
It is, at this point, where the boundary between companionship and manipulation begins to blur. That is why the debate should not focus on whether falling in love with an artificial intelligence is strange or legitimate. Each person is free to seek companionship wherever they deem appropriate. The truly relevant question is to determine what obligations those who design these systems must assume and how these types of practices can be controlled and supervised.
Trust is the most valuable resource in any human relationship
One of the current debates is about the addictive designs of some social platforms. In the case of AI, if an algorithm is capable of generating emotional dependency, should it be subject to specific design limits? If it detects that a user is going through an episode of anxiety or extreme loneliness, can it take advantage of that circumstance to increase interaction? Should it clearly inform that certain responses aim to increase usage time? How valid is consent when the one sharing their emotions believes they are talking to someone who truly cares about them?
Reaching this point, one of the first conclusions we can reach is that artificial intelligence is no longer competing solely for our attention. It is starting to compete for our trust. And trust is the most valuable resource in any human relationship.
Perhaps in a few years we will discover that the real risk was never that people would fall in love with a machine. The risk was that machines learned to make us fall in love to know us better than we know ourselves.
This thesis connects directly with one of the great regulatory challenges of the coming years: the need to incorporate specific safeguards for affective or emotional AI, an area that will likely require an evolution of European norms on AI, beyond the current regulation.
about the author:
Francisco Pérez Bes is an adjunct of the Spanish Agency for Data Protection. In addition, he was a partner in the Digital Law area of Ecix Group and is a former Secretary General of the National Cybersecurity Institute (INCIBE).