Suggested Approach to AI Usage: A Framework Comparison

Safety assessment before therapeutic work begins. Is the person in immediate danger? Stabilize before anything else.

Check for hard limits before anything happens. What’s absolutely off the table? What would cause real harm? Non-negotiable, established upfront.

Is your kid eating? Sleeping? Going to school? Maintaining any human contact? If the basics are intact, you have time to be thoughtful. If not, that’s a crisis regardless of AI.

Motivational Interviewing. Open-ended questions. No leading. “What does this look like for you?” Meet the person where they are, not where you think they should be.

Check in with genuine curiosity. “Where are you right now? What are you feeling? What do you want?” Approach their experience without assuming you already understand it.

“Show me what you do with this. I’m curious.” Not “why are you wasting time with a chatbot?” Ask what they get from it that they’re not getting somewhere else. Actually listen to the answer.

Reflect back what you heard. Ask “is that right?” Don’t proceed until the person confirms. If they say “no, that’s not what I mean,” don’t argue, ask again.

State what you think you heard and wait for confirmation. The conversation doesn’t advance until both people agree on what’s actually being discussed. Negotiation doesn’t start until understanding is confirmed.

“So you’re talking to something that reminds you of Grandma because you miss her?” Then wait. They might say “no, it’s about having someone who listens the way she did.” Those are two different things. You don’t know which one until they confirm for you.

Collaborative treatment planning. Both parties state goals and concerns. The plan is agreed on, not prescribed.

Both people state wants, needs, and limits. The agreement is built from the overlap. Concrete and specific, not vague good intentions. Negotiation is expected and encouraged to ensure both sides are fulfilled.

“Here’s what I need: to know you’re still seeing friends, to see one conversation a week, and for you to tell me if it says something that upsets you. What do you need from me?” That’s a negotiation, not an ultimatum.

Treatment goals reassessed regularly. “Is this still working?” at every session, not just intake. The plan evolves with the client.

Consent is ongoing, not one-time. Regular renegotiation is normal and expected. Circumstances change. Feelings change. The agreement updates.

The conversation you had three months ago doesn’t cover what’s happening now. The AI has changed over time, or may have been replaced by another. Your kid has changed. “Is our agreement still working for both of us?” is a question that never stops being relevant.