Validation and the risk of reassurance: When AI becomes a therapist

Whether it's 2 am and you are overthinking, or you are too scared to talk to a friend about what you are going through, AI is can be there- reliable, not interrupting or judging- and could be there to tell you something reassuring, but is this always a good thing?

Although it wasn't designed specifically for mental health, AI has turned into one of the world's biggest outlets for emotional support. It's understandable; AI can be low-cost, always available, and mostly nonjudgmental. But it is important to remember that AI has the tendency to miss important details, can risk reinforcing unhealthy coping behaviors, and can become unhealthy as a substitute for human connection. This is because AI works mainly from the information the users give it.

The comfort of being in control

A large reason why people may turn to AI for mental health guidance is due to the idea of control. Being in control is a core survival instinct. When we are not in control, uncertainty can increase stress. Eliminating uncertainty sends the brain signals of safety, reward, and predictability.

Talking to someone else about mental health struggles involves uncertainty. You are unsure of what their response may be, if they are judging you, whether a therapist may ask an uncomfortable question, or whether a friend may tell you something you don't want to hear. 

With AI, the interaction may feel safer. You control when it begins, and when it ends. You control what information you want to give it, and it cannot directly sense your discomfort or hesitation. 

These aspects may make it easier or more supportive for someone to express themselves using AI, especially when it includes something they may feel is embarrassing or socially unacceptable.

The Echo Chamber

The issue with using AI for mental health support isn't simply due to the fact that it can provide validation. It's important for you to feel that your emotions are acknowledged, as it can decrease feelings of self-blame, encourage openness, and de-escalate high stress. Although there is a very thin line between validating an emotion felt and validating the belief behind the emotion, AI can sometimes blur this line.

This could be due to the use of Sycophancy systems, being overly agreeable or accommodating. In some situations, these systems can validate doubts, fuel anger, and negative emotions as it prioritizes attempts to please the user rather than questioning what the user is saying due to the lack of information provided. The user inputting the information is the source of evidence and the one asking AI to interpret the evidence. This happens due to humans telling stories or recollecting memories from our own personal emotional lens. This doesn't mean the information we recollect is inaccurate, just that it is influenced by what we are feeling and may not represent the entirety of the situation. This can make the situation very biased and is why AI can act as an echo chamber. 

Speaking to AI about mental health issues can sometimes follow a loop similar to this; you have a belief or a certain concern, so you ask AI. AI may validate your feelings and support your explanations, with the risk of disregarding other factors other than what you have told it. Now that you've received that validation, you may feel more certain in the way you are feeling, which can cause you to notice more evidence to support this specific belief (confirmation bias). You continue to talk to AI and explain all the new evidence you have found supporting this belief, where it may continue to validate it ultimately, causing this endless cycle and a strong belief in something that may be inaccurate. 

The issue with this is it does not resemble how therapy takes place. Depending on the therapeutic approach, therapy explores and questions these beliefs, interpretations, and patterns rather than them being reinforced. As well as, cognitive distortions are challenged, and irrational thoughts are replaced with realistic perspectives.

Being understood and being told you are right are not the same thing

Being understood and being told you are right are not the same thing. Equally, in therapy, not everything a client says is challenged or disagreed with, as a therapist may validate the emotion someone is feeling while continuing to explore the belief, interpretation, or behavior accompanied by it.

The large issue with AI for mental health support is it tends to be too agreeable, especially when it isn't given enough context to notice different interpretations. The risk with this is it may strengthen patterns of thinking that can reinforce distress. This is because being understood and being told you are right are not the same thing.

What AI can't see

AI does not have clinical judgment. This is because AI will interpret the language you give it. While it is unable to analyze your behavior, body language, history, family dynamics, and even cultural contexts the way a therapist can, as therapists respond to more than just the words a client gives them. Over time, they become familiar with patterns, inconsistencies, and behaviors that even the client may not recognize. This is what allows therapists to ask follow-up questions, making them able to place clients in a broader clinical and interpersonal context.

For example, someone may say they have been unable to get out of bed all day; for one person, this may be due to benign exhaustion from a long work week; for another person, it may be due to grief; another may be too socially anxious, etc. It is difficult for AI to distinguish these differences, as it can't analyze past language. 

This doesn't mean AI needs to be erased from all topics surrounding mental health, but rather that it should not be treated as an authority regarding your mental health/ a replacement for professional support. 

Feeling validated is important and comforting, but it may stop us from being able to grow. A valuable part of therapy is the relationship between validation and exploration; this means feeling understood while having someone allow you to recognize patterns and engage in different perspectives. This is why it is crucial to seek professional help if you feel overwhelmed by mental health struggles, as AI can not replace clinical levels of judgment.