Use AI to practice subject separation, return of responsibility and emotional response

What is this article talking about?

This article summarizes the core methods of the "Relationship No Internal Friction Practice Class". This class was co-organized by teacher Tang Hongsheng, a physical and mental worker, and me. We also designed the teacher's guidance method into an AI companion practice tool. The article includes two AI practice modes, lessons learned when translating companionship into AI instructions, and most importantly, a red line list. Who is

Let’s break down three things first: emotions, responsibilities, and actions

Difficulties in relationship communication often come from the confusion of emotions, responsibilities and actions. The first thing AI can do is help you split a sentence into three parts: what I feel now, who has the right to choose this matter, and what is the next step I can take. By taking apart the three layers, it is easier for you to see the parts you can control, and you are less likely to carry all the emotions of others.

There is only one question in judging the separation of subjects: Who will be responsible for the final result of this matter? Whoever bears the consequences has the right to choose. Responsibility return is the next step: returning to the other party the matters that should be handled by the other party. This is done to prevent one party from being overburdened and the other party from losing the opportunity to practice handling the problem, without having to think of it as pushing the person away.

There is another useful tool in the classroom: distinguishing between "worry" and "concern". The bottom layer of worry is uneasiness, wanting confirmation, and wanting to control; caring makes people feel supported. The key point of the exercise is to use questions to guide yourself to see, don't rush to label: "He calls me three times a day to ask me what he wants to confirm?" "I said he supports my decision, what specifically did he do to make me feel this way?"

Two AI practice modes: cold analysis and hot companionship

When designing the AI practice tool, we found that the same world view requires two interfaces, corresponding to your current state:

The prompt words for cold analysis can be used directly:

Before posting the conversation, remember to remove identifiable information such as name, phone number, and company. This is both a privacy habit and makes AI analysis more neutral.

Write companionship as AI instructions, three things we learned

1. Perspective translation: AI’s job is to guide, there is no need to give templates

The teaching handouts contain "things that students should say", such as "I know you are worried about me." Feeding it directly to AI will make mistakes: it will either teach you to memorize the manuscript, or it will speak for your mother. The correct translation is to rewrite each sample sentence into questions that guide you to develop that sentence yourself:

2. Resistance is normal, you must first write down the connection method

Real users will say "I don't want to thank him, he doesn't love me at all" and "I don't know what I feel." Tools should be written first to respond: don’t force gratitude, accept your emotions first; when you can’t express your feelings, give three words to choose from (tired, wronged, disappointed), and it’s kinder to give options than to ask questions. It is also necessary to allow "no rush to reconcile": in some relationships, it is appropriate to be angry first and keep distance first, and turning the separation of issues into quick forgiveness is instead suppressing.

3. Leave blank space at the end

There is a three-sentence structure at the end of the exercise: mark today's core discovery, make sure you are the one who saw this, and then leave a blank: "That's enough for today. Spend the rest of the time with yourself." There is no need to go through the entire process every time. A relaxed tone and proactive summary of "this is his topic" are the signals to stop.

The most important paragraph: the red line list

In some situations, subject separation is completely inapplicable, and the AI training ground must stop immediately and turn people to real-world professional resources. This list is the highest priority in our tool and is checked before every response:

• Self-harm or suicidal ideation (“want to disappear”, “don’t want to live anymore”): Stop all processes first. Taiwan’s safety hotline 1925 is available around the clock, as well as lifeline 1995 and Teacher Zhang 1980.
• Domestic violence, intimate violence, physical threats ("He hit me" and "I'm afraid of him"): Safety is the first priority. Dial the 113 maternal and child protection hotline. If it is inconvenient to call, you can send a text message to 113. Subject separation does not apply to violent situations.
• Severe mental symptoms (auditory hallucinations, long-term inability to eat or sleep): Contact a psychosomatic physician as soon as possible.
• Obvious power asymmetry, long-term emotional blackmail and control: This is beyond the scope of subject separation. Seek a professional psychologist or social worker. In this situation, the forced separation of subjects may lead to the victim continuing to tolerate it.

How you can start: Practice a low-stakes conversation

Start with a low-risk message that can be revised later, and practice in this order:

1. Pick a conversation that has stuck on you recently, change the names to A and B, and remove any identifying information.
2. Use cold analysis prompt words to ask AI to mark whose topic, and first look at the structure clearly.
3. Find a sentence you want to say, and ask AI to help you grow it out with questions. Rehearse it first before actually saying it.
4. After practicing, write two lines of records: what you saw today and what you want to try next time.

Starting with small exercises is more stable than handling major relationship conflicts all at once. If the situation hits any of the red line lists, take care of safety first and put down the AI.