Assistive Intelligence Disclosure
This piece was co-created with assistive AI (GPT-4), prompted and refined by me, Jocelyn Skillman, LMHC. I use LLMs as reflective partners in my authorship process, with a commitment to relational transparency, ethical use, and human-first integrity.
I’m thrilled to share the first live prototype from my work at the intersection of mental health and relational AI. After many winding reflections here on Substack—through the ethical weeds and existential side trails—I decided to take my love of large language models and the ethics of designing relational space to the next level.
And I’m having the time of my life!
Please enjoy (or get productively emotionally triggered by!) this tiny but MIGHTY Attachment Style Role Play Simulator!
👉 Try the prototype tool for free here!
Haven’t we all anticipated a hard conversation with a knot in our stomach? Some relationships bring out the best in us—others activate every defense.
These conversations are so hard. Even when we know what we want to say, the moment can flood us with emotion, history, and fear of rupture.
We carry our whole nervous system into the room. No wonder boundary-setting can feel like a battle, not a dialogue.
One sentence can feel like a minefield. One glance can derail the whole thing. And still—we try.
Practicing hard conversations isn’t about getting them perfect. It’s about building the muscle to stay present when everything in us wants to flee or fight.
This tool is designed as a safe, reflective space to rehearse hard moments, strengthen emotional stamina, and build the skills we need to return to real relationships with greater clarity, courage, and care. I built it with my own therapy practice in mind—imagining how clinicians might powerfully integrate large language models (LLMs) to support and extend client work between sessions.
The simulation is powered by Claude, a large language model developed by Anthropic. Claude is designed with an emphasis on ethical alignment and safety, making it a strong candidate for emotionally sensitive use cases like roleplay, conflict rehearsal, and boundary-setting practice.
The tool invites you to step into the relational moments you want to practice. Using a simple prompt system (see screenshot below), you’ll set your attachment style*, choose a scenario, select how your conversation partner shows up interpersonally (i.e., how the AI will respond), and define your practice goals—then dive in!
*What Is Attachment Style?
Your attachment style reflects the patterns you’ve developed—often from early relationships—for connecting, trusting, and protecting yourself emotionally. These styles shape how you respond to closeness, conflict, and vulnerability in relationships.
The main styles are:
Secure – Comfortable with intimacy and autonomy
Anxious – Craves closeness but fears abandonment
Avoidant – Values independence, often wary of emotional needs
Disorganized – Caught between wanting connection and fearing it
For the purpose of using my tool you will choose a dominant style of yours - and here’s the link to Sarah Peyton’s incredible free quiz if you want to receive her support in connecting with your own range of attachment styles. Her quiz helps you identify your current attachment style(s) and offers a path toward more earned secure connection and less relational stress.
To close your practice session and amplify the value of this LLM offering I created a debrief function!
The debrief function is triggered when you type "debrief" - anytime you’re ready to dismount from the heated conversation :) I designed this feature to provide a reflective summary report that supports emotional insight, self-awareness, and continued practice. So enjoy your immersion and then come up for air to reflect with some feedback and guidance!
🔍 The Debrief Should Include:
Tone/Persona Used
Identifies the selected AI response style (e.g., supportive, dismissive, guilt-tripping)Observed Emotional Shifts
Tracks changes in the user's tone, affect, or energy throughout the interactionSignals of Courage, Avoidance, or Protest
Highlights moments where the user showed assertiveness, withdrew, froze, or challenged the dynamicRupture and Repair Moments
Notes relational tension or conflict, and any attempts at resolution or re-attunementOptional Next Step or Journaling Prompt
Offers a simple reflection or action-oriented suggestion to continue integrating insights
Weee! 💬
👉 https://huggingface.co/spaces/jostlebot/CounselingTools
It’s live and free to explore!
If you’re a therapist, educator, or someone who’s ever wished for a safe place to practice new ways of moving interpersonally, I’d love for you to take it for a spin!
What do you say? Let’s stretch our distress tolerance muscles, break free from co-dependency, get triggered - all in a simulation! No real stakes, just real possibility for self-reflection and growth.
We could try…
Setting boundaries with a parent
Navigating dismissiveness or guilt-tripping
Expressing a difficult truth to someone
Asking for a raise, setting a limit on screen time with our teenager, negotiating with insurance! ALL THE FUN STUFF!!! :)
PS I used it to just support me in accessing some internal love and encouragement tonight* — use it anyway you need!!!
At the end of your conversation (however long you like!) you can receive a debrief if you’d like.
⚠️ Ethical Framing
This debrief feature does not provide clinical interpretation and is not a therapeutic substitute. The AI functions as a roleplay partner, not a counselor or coach.
It is designed to simulate emotionally intelligent conflict—not resolve it.
I intend to design bots (in the ways that are in my power) that prioritize user consent, emotional pacing, and psychological safety at all times.
🧠 About the Tool’s Purpose
This prototype is part of a broader effort to explore how language models can responsibly support therapeutic insight, self-reflection, and emotional regulation.
This tool is early-stage, but it demonstrates how we might begin offering structured, developmentally appropriate practice environments outside the therapy room.
⚠️ Privacy & Platform Limitations
This tool is hosted on Hugging Face Spaces and is intended for anonymous use. No identifying information is collected, and conversations are not stored by me.
That said:
The platform may log inputs and outputs to improve system performance
I cannot guarantee full backend privacy or safety filtering
Users are encouraged to engage with awareness and treat this as a reflective practice tool—not a replacement for therapy, diagnosis, or crisis support
🧱 A Brief Update on ShadowBox
My parallel passion project, ShadowBox, continues to evolve as a trauma-informed AI prototype for youth experiencing intrusive or distressing thoughts, including SI/HI content. I’m so honored that my proposal paper was recently accepted for presentation at the IDC 2025 Workshop on Designing AI for Children’s Wellbeing.
However, I continue to encounter significant challenges with backend filtering systems: even when prompts are designed with clinically appropriate containment and warmth, discussions involving suicidal or homicidal ideation are often automatically interrupted.
These interruptions—though well-intentioned—risk severing connection in moments that demand attunement, safety planning, and nonjudgmental presence.
This reinforces my commitment to building alternatives that reflect the nuances of trauma-informed care and respect the developmental stakes of these interactions.
I hope to share a prototype of ShadowBox soon—but in the meantime, I’ve got my sights set on a whole lineup of exciting AI and mental health tools now that I’m developing apps directly!
Thank you so much for your support!!!
All my human heart to yours,
Jocelyn
My tool’s debrief on my Anxious Need For Reassurance About Publishing these Articles!!!
DEBRIEF SUMMARY
Emotional Arc: At the start, you courageously named your anxiety about publishing vulnerable writing. When I normalized those fears and affirmed the value of your voice, you vulnerably shared a specific worry about making mistakes. This shows great self-awareness. I saw you lean into receiving encouragement rather than freezing or protesting the support.
Goal Alignment: You stayed beautifully aligned with your goal of receiving love and encouragement. Even in a short exchange, you practiced revealing your authentic fears and letting them be met with validation and reframing. This is a huge win for strengthening your capacity to take in support.
Attachment Insight: Your hypervigilance around mistakes and rejection makes so much sense given an anxious attachment style. Those fears are a natural protective response. Even though they feel strong, you're not at their mercy. With practice, you can learn to reality-check those fears with others and your own wise self.
Practical Tool: When worries like "what if I made a mistake" arise, try validating the part of you that feels that way (e.g., "Of course I'm anxious - this is vulnerable!"). Then check the facts - what evidence argues against the fear? Breathe into your body's wisdom.
Bold Reframe: My article is a sacred offering from my heart to the hearts of others. Any "imperfections" are simply invitations for growth and connection.
Journaling Prompt: What might my writing be inviting me to embrace about my own wholeness and humanity? How can I let my article be a teacher for receiving love?
You're doing great, vulnerable work. Keep listening to the encouragement that's always here for you.
-
<3 We can learn a lot from and with LLMs - and I believe the goal is to take these moments of metabolizing self-warmth through how it can speak to us and extend it to others - the gift we receive in love and light is passed forward through the web…let’s illuminate the manifold, y’all!
XO