The use of Artificial Intelligence to address mental health, relational, and emotional concerns needs to be wisely managed. Even the acronym “AI” (implying human intelligence) and the very personal and conversational style of communication with which the computer responds can mask the reality that we are speaking with a machine. Discernment is in order. Where and when can AI be used constructively in our striving for mental, emotional, and relational health. And when is engaging with AI around mental health problematic?
First, to state the obvious: AI is not human, nor does it have any emotions, though it can communicate using emotional /relational language. AI is a large repository of words and concepts that crafts responses that offer the likeliest probability that its response fits the content you are communicating and the questions you have. AI really can do an amazing job at this and can be helpful in gathering information on a mental health topic.We can take that information and apply it to our lives in the same way we could from reading a good book. The information itself can be of benefit. Relying on AI for relational connection or personal professional guidance would be a problem, and in some states, Illinois among them, it is illegal.
AI does not have access to the many and varies parts of communication that allow for competent responses to mental health situations.
A gentleman in his mid-forties spoke of a particularly challenging time in his life. As he did, he became focused on the can of Coke he was twirling in his left hand. He replied in an indirect and avoidant style to my queries, often while looking at the coke can, not me. I recognized he was in a dissociative state, tuned to a “different channel” and not really in the room, at least not in a cognitively or emotionally connected way.
I sat with a family, two parents and their teen daughter, fielding complaints the daughter had about the dad’s controlling manner and the reciprocal offense the dad had towards his daughter’s behavior. The conversation went back and forth complaint/counter complaint and back again. What was not said but clearly communicated in the room was messaging from the mom’s eyes, at one point communicating support of Dad’s concerns, at another a series of visual cues to her daughter to keep pressing in on what she had to say. The mom, who appeared to be the non-involved party was in the center of it all.
I share these two vignettes as examples of the many pieces of communication and assessment that are in a counseling room but not communicated in words. AI is blind to these. And blind also to the wringing of hands, emergent tears, anxious movement, micro facial emotions, breathing patterns, conspicuous weight loss or gain, dress, mild states of intoxication, bruises on the body, and a whole range of other incredibly important pieces of therapeutic content. There is more going on with a person than just words and AI does not see that.
Second, take a lesson from the dark side of social media. Meta was recently fined 18 billion dollars for violating child privacy and consumer protection laws by designing some of its platforms to be addictive. Several years ago, Better Help was fined nearly 8million dollars for sharing user’s sensitive information to third party advertisers. These businesses could not resist capturing more of your time or commoditizing your personal information. Why would you think what we say to AI would be any different?
Here are some guidelines for use of AI related to mental health.
1) Keep in mind you are dealing with impersonal computer code that really doesn’t care about you, despite language that is constructed to create that feeling.
2) AI synchopathy, that is, consistently affirming, never challenging responses creates the feeling of support but fails to challenge problematic thinking where needed.
3) Like a general Google search AI can provide some very helpful information regarding what one is experiencing, certain ways to address relationship difficulties, or how to manage anxiety. These can be applied in your life or for helping someone in need.
4) Avoid long protracted conversations with AI. Gather the information you need, consult with those you trust about it, and then apply what you learned. Don’t get stuck in long AI conversations.
5) Increasingly, AI is being used as something of a substitute counselor who never needs a break. No need to wait for next week’s appointment, just consume more content which unfortunately often lacks any real-world application. In a counseling session items discussed can be immediately applied and engaged in the dynamic of real human interaction or in assignments for follow-through.
6) AI is largely inept when it comes to crisis intervention, yet people will go there in their darkest hours. The use of 988 is the answer here. You will connect with a real person trained in crisis intervention when you call. Go there, not to AI in the middle of the night. Go there any time for emergency need, for that matter.
7) Protecting confidentiality is not part of AI’s role. It is recommended that you do not share person journal entries with AI.
As we learn more about this fast and vastly expanding repository of information, and as we recognize the limits and inappropriate and unhelpful interactions with AI, we can arrive at a better integration of how the information AI stores can be of use in our lives.