Real friends understand each other, even if imperfectly...
Michael Meyer is a professor emeritus of philosophy at Santa Clara University and a faculty scholar with the Markkula Center for Applied Ethics. Views are his own.
For social creatures like us, friendship is almost universally seen as essential to human happiness. The friendless person is likely to be seen as being in need of help rather than admired, and much has been written about the level of loneliness in our society. In the age of generative artificial intelligence (AI), we are facing a new question: What are the possible effects of the use of chatbots on the crucial habit of human friendship?
Though friendship is widely seen as one key to happiness, not all beneficial relationships (like that with your empathetic Uber driver, or your dog) count as real friendships. A true friendship is a deep and important human relationship, potentially creating a kind of “second self.” Friends of this sort are a mirror in which you--through thoughtful support and emotional insight--can come to see and become your better self.
So, can chatbots be effective substitutes for human friends in this deep sense which is directly connected to human happiness?
At present, AI chatbots simulate human conversations without any reliance on a pre-programed set of responses, and they are often advertised as “understanding” the questions asked and thereby providing useful answers. As such, they may be suitable for customer service and more, but the word understand is in scare quotes above for obvious reasons. Chatbots do not have anything even close to a modest understanding of the questions we ask them or their own answers.
The notion of an understanding chatbot friend is at present an illusion because the ability of an LLM to generate the next plausible token or word and thereby a seemingly relevant conversational response (no matter how complex and impressive that process is) does not count as an even very basic case of understanding a human being. And real friends do understand each other, even if imperfectly.
Will chatbots become better at friendship as AI evolves? Suppose current models are laid aside (or grafted onto) a new generation of what are sometimes called World Model AIs, with the ability to grasp physical phenomena like the physics of cause and effect. Even then we will still be very far away from anything like artificial intelligence having an even rudimentary understanding of the lived complexities of human existence.
To begin to understand a human being like a friend, one must surely recognize and appreciate, perhaps intuitively, their emotions, the proper times for criticism and praise, sources of motivation, and so much more.
Of course, if AI boosters expect a great deal less of the notions of understanding, or friendship, or happiness, they might come to another conclusion. But watering down crucial ideas like friendship and human happiness hardly seems the path to real progress.
Speculating about a truly far off future is a bit of a parlor game, but what about Data of Star Trek fame or the character of Klara from Ishiguro’s beautiful novel Klara and the Sun? Perhaps someday we will get embodied, sentient, insightful, empathetic, sometimes wise robotic AIs who can love us like human friends do now–imperfectly but with some common sense and emotional insight.
Yet, it seems likely that to truly understand human emotions one must experience something very similar to human emotions. The urgent issue, for now, is to avoid distorting our ideas of friendship and human happiness in the way that present pornography can distort our ideas about and our experience of sexuality. This is especially important in the screen-saturated world of the young, wherein present-day chatbots are already being advertised as replacing flawed but truly human friendships.
Image: Reihaneh Golpayegani / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/