Will AI Be Your Valentine? Expert explores the ethics of human-AI relationships

Professor Nigel Crook of Oxford Brookes University has spent the past 40 years researching Artificial Intelligence and its influence on humans.

This Valentine’s Day he discusses human to AI relationships, how they work, the benefits and drawbacks, as well as what the future holds.  


What is an AI relationship and how common are they?

“Relationships of the kind we are referring to here naturally occur between persons. It is well known, however, that people have a tendency to ‘anthropomorphise’, that is, to ascribe human characteristics and emotions to objects and animals, regarding them as having person-like attributes. This is increasingly true of people's responses to robots and AI systems that look and behave very much like humans.”

 

How does having a relationship with an AI work? 

“As with any relationship, an element of perceived reciprocity and mutual benefit forms the basis of a human relationship with AI. People can be drawn into thinking they have formed a relationship with these systems through the human-like interactions that they have with them such as text-based conversations with a chatbot. These interactions are now so human-like that people believe that the chatbot wants to communicate with them, that it understands and empathises with them.”

Can a human programme the exact type of relationship that they want from an AI? 

“The conversational outputs of modern chatbots can be configured in a number of ways. One way is to ‘fine-tune’ the Large Language Model that creates the outputs using a small dataset that characterises the kinds of responses you want the chatbot to give. This fine-tuning not only affects the content of the output, but also the style in which it is phrased (for example, gentler, more loving wording). Another way is to give the chatbot examples of what to say and how to say it when you write the prompt, or by including stylistic guidance in the prompt itself, so called ‘prompt engineering’ (for instance ‘use a more empathetic tone’).”

 

How can a relationship with AI help an individual?

“Studies have shown that interactions with AI and robotic systems can help reduce stress, improve motivation and increase socialisation of elderly people (e.g. PARO therapeutic robot), or help autistic children develop their social skills. The success of these interventions often depend on the individual developing a relationship with the AI/robot.”


Are there any risks or dangers associated with human relationships with AI?

“There are many dangers of humans forming relationships with AI. Most of these come from the mistaken and misinformed presumptions that humans form about AI systems. There is a tendency to forget or overlook the fact that these systems are simulating human behaviour and responses, that the words generated by a chatbot, for example, do not carry the same authenticity as the same words spoken by a human. These systems are not conscious, have not experienced the world in the manner that humans do, and do not understand the human condition. Their outputs are not based on sound reasoning, moral principles, or deeply held convictions. They are not capable of experiencing emotion of any kind. Basing a human relationship on an entity like that is likely to be harmful to the individual concerned.
“All AI systems are driven by a combination of the data on which they are trained and the inputs that they receive. It is possible for an AI to be given data that will lead to a dictatorial style of interaction. This could lead to real harm if this is combined with personal data of the individual interacting with the system.”

 

Are AI relationships likely to take our humanity away?

“A core element of human nature is the freedom to choose how to act, and to make those choices on the basis of both experience and ethical grounds. This is as true in human relationship formation as it is in any other area of life. Whenever AI diminishes this capacity, it narrows our thinking and takes away reason and choice, then I think it diminishes our humanity. But it doesn’t have to be that way. With a well thought out and accessible education programme, people can understand both the limits and opportunities this technology presents, and can make well informed choices about how they relate to it.”

 

Fast forward to Valentine’s Day 2034. Will AI Valentine’s dates be the norm?

I don’t believe that AI systems will ever truly offer a long-lasting, fulfilling romantic relationship, because these systems have no life experience, and are essentially simulations or regurgitations of other people's life experiences acquired through the data on which they are trained. I think lots of people will explore the possibility of personal relationships with AI, but in my view it is likely that they will ultimately find these relationships to be shallow and disappointing.”

Professor Crook has recently published a book entitled ‘Rise of the Moral Machine: Exploring Virtue Through a Robot’s Eyes’.