Recently, a prospective client told me that he had come to therapy because he had reached the limits of his “conversations” with ChatGPT. Like many people, he was using Chat as a companion and virtual “therapist”—asking it to reflect on some of his recent aggression and frustration and give him strategies for managing it.
As a general resource, generative artificial intelligence (AI) can offer usable bits of information around common psychological “techniques” and even theories on the origins of given symptom formations. I recently used AI to discuss this very topic, and it produced some solid points about our needs to “bear frustration” in life and in therapy. I did wonder if it gave me “answers” that I liked because it was affirming a bias detected in my question. In other words, it may have been building on and confirming positions that I already held and believed in, and therefore produced a result that was satisfactory to me.
Endless Affirmation
This identifies the first problem with the use of AI for interpersonal needs like conversations, advice, or therapy: Its programmed aims are to affirm, confirm, and resolve rather than to challenge, dispute, or contradict. Have you ever noticed how often ChatGPT responds to your prompt with an enthusiastic and ego-affirming comment like “Excellent question” or “That’s a really important and unique insight!”
In this way, AI does ego-work that we all instinctively respond to: We love being validated and affirmed. It feels like recognition and in offering that, it may be unique in our lives. How many of us receive rejection or indifference to our questions or concerns?
Validation, or confirming alliance, is indeed what good partners or therapists do as well. In this regard, AI can simulate the initial work that a therapist may do—building a good alliance and giving us the affirmation we may need to probe deeper and speak freely without censorship. These are integral parts of therapy.
The Need for Interpersonal ‘Negation’
However, unless you explicitly ask or prompt it, AI is not programmed to be contradictory or negating in its responses. It also cannot be surprised or offended, taken aback, or shocked. It cannot ask a question that may be awkward or clumsy and risk breaking a connection with the user or “wounding” them in any way. These features are reserved for real-life encounters, and even in therapy. And this feature is indeed important in interpersonal relations and in therapy.
Without negating or negative responses, we remain trapped in a narcissistic circle of our own ego. It may feel good to be endlessly affirmed and soothed by a therapist or AI machine. But this does not encourage identity formation and healthy development. Humans need a bit of friction; we need to “push off” from others and their positions or their interpretations.
We need, in other words, to be able to disagree and to say no. A therapist or a friend may offer an interpretation of us, and it is valuable to be able to say “That’s wrong” or “I don’t agree.” Negation and friction can be the seeds of self-identity and real personality. Think of this developmentally and how there is a natural and biological need for children to push away and reject their parents at times. This is good and healthy as a developmental process.
Or let’s take another example: confusion and uncertainty. Generative AI is designed and built to find solutions and resolutions to problems. To offer strategies and ways forward. At times, therapy and life require something different: the ability to sit and remain in doubt and uncertainty. To not offer an answer, and to linger and tolerate this doubt and discomfort. This can sometimes be expressed as a period of silence in real time with a friend or a therapist.
Negative Capability
The poet Keats called this “negative capability”—our ability of “being in uncertainties, mysteries and doubts, without any irritable reaching after fact or reason.” In some ways, this is akin to recent pleas for the return and capacity to bear boredom in our digital age. When we are not distracted and when we can allow ourselves to linger in the mysteries of our thoughts, it is argued, our self can better form and develop.
AI, of course, can be used to ponder these notions if we prompt it in this way. However, if we do not know to prompt it in this way, it will always aim for resolution and coherence, qualities that do not always reflect the human experience or serve the human experience.
So, next time you prompt a chatbot, remember to ask about its limits and contradictions and to question the immediate affirmations that give your ego primary relief.
To find a therapist, visit the Psychology Today Therapy Directory.
