Introducing this Spark
Graduate counseling students rarely get to rehearse suicide risk and behavioral threat assessment before doing it with real youth, because live cases cannot be summoned on demand and practicum exposure is uneven. This AI Spark gives every student unlimited, developmentally realistic practice interviewing a simulated adolescent client and determining risk or threat level, with immediate structured feedback against established clinical frameworks.
Learning Outcome
After this activity, students can conduct a developmentally appropriate assessment interview, elicit information across the core risk domains (ideation, intent, plan, access to means, prior attempts, precipitants, protective factors) or threat domains, determine a risk level using the Joiner et al. (1999) decision tree or classify a threat using the Florida Standardized Threat Assessment Tool (FSTAT), and build an appropriate safety or protective plan. The activity supports the course objectives that students learn to conduct a child clinical interview and learn developmental and social-contextual considerations in working with children and adolescents.
Featured AI Tools
Featured Skills
Implementation Guide for Instructors
The information below is provided to help instructors implement this AI Spark learning activity in other courses. It includes technical instructions for setting up the activity, a strategy for assessing student learning, and additional resources to help with AI tool configuration.
In this Gemini-based activity, graduate counseling students interview a simulated child or adolescent client to practice suicide risk and behavioral threat assessment. The simulator operates in two modes: a client mode that role-plays a fictional youth from a hidden, calibrated case profile and reveals sensitive information only when the student asks well, and a supervisor mode that scores the student against the Joiner decision tree, TASR, and the Florida Standardized Threat Assessment Tool, evaluates the safety plan against the Brown and Stanley template, and produces a structured instructor summary. By withholding information until elicited, the simulator turns AI from a shortcut into a rigorous interviewing and reasoning workout. Though built for counselor training, the elicit-then-assess design adapts to any field that teaches diagnostic or investigative interviewing.
- Go to gemini.google.com and sign in with your FSU Google account.
- Click Gems in the left sidebar, then New Gem in the top right.
- Name the Gem "Risk and Threat Assessment Simulator".
- In the Instructions field, paste the provided system prompt exactly as written. You’ll find the prompt in the Resources tab. It configures the AI to operate in two modes: a client mode that role-plays a fictional adolescent from a hidden, calibrated case profile and reveals information only when the student asks well, and a supervisor mode (triggered when the student types END SESSION) that scores the assessment against the course frameworks and produces a structured summary.
- Optionally, upload the reference instruments as knowledge files for the Gem to draw on. These documents can be found in the Resources tab.
- Click Save.
- Share the Gem link with students. Because Gem link-sharing requires both instructor and students to be on FSU Google for Education accounts, you may instead post the system prompt text directly in the Canvas assignment and have students go to gemini.google.com, create a New Gem, paste the prompt, and save it as their own personal Gem.
Canvas Setup: Post the Gem link (or the system prompt text) and the student micro-training in the assignment. Create a submission point where students upload the PDF of their completed session (the full transcript plus the structured summary, saved with the browser's Save as PDF) and complete the reflection.
Each completed session is captured as a transcript and a structured summary that documents whether the student elicited each core domain and whether their determination matched the case's calibrated framework key, the completeness of their safety plan, and the quality of their interviewing. Because each case is calibrated to a known risk level, determination accuracy is judged directly. A short reflection captures the AI-literacy outcome. This connects to the outcome by measuring the exact interviewing and determination behaviors the activity is meant to build.
Link to the Shared Gem: Risk and Threat Assessment Simulator
Link to System Prompt: Risk and Threat Assessment Simulator Prompt
Instructor Deployment Notes:
- Write constraints explicitly and at high priority. The model will drift into being the clinician or ignore the end-session trigger unless the role lock and trigger are stated forcefully near the top of the prompt.
- Calibrate and pilot cases. Each hidden case specifies the correct level and which details are gated; test a few runs to confirm the levels land consistently.
- Have students start a fresh session per case to reduce role drift in long chats.
Key Behavioral Constraints: The simulator acts strictly as the client and is forbidden from behaving like the clinician. It reveals sensitive risk information (active ideation, intent, plan, access to means, prior attempts) only when the student asks directly and specifically, never volunteering it and never screening the student. It switches to evaluative supervisor mode only when the student types.
Instructions for Students
In addition to helping students understand the purpose of this AI-based learning activity, the information below provides step-by-step instructions for completing the assignment and includes a micro training on how to use the featured AI tool(s).
This activity uses a custom AI simulator built on Google's Gemini. It role-plays a fictional adolescent client you interview by typing back and forth, staying in character within the rules the instructor set, and, at the end, stepping out of the role to give you supervisor-style feedback on your assessment. The goal is not to get an answer from the AI but to practice a real clinical interview. You will interview the simulated client to assess suicide risk or threat. The client reveals the important information only when you ask clearly and specifically, so the activity builds your questioning, your clinical reasoning, and your ability to critically appraise AI-generated feedback rather than accept it at face value.
Step 1. Open the activity. Use the link posted in Canvas and enter your name (access it with your FSU Google account).
Step 2. Choose a case. Select the case type (suicide risk or threat) and the assigned case or difficulty.
Step 3. Conduct the interview. Interview the client as you would in a real session. It will not volunteer the important information, so build rapport and ask directly, including the hard questions about prior attempts and access to means.
Step 4. End and get feedback. When ready, type END SESSION, state your risk or threat level and your safety or protective plan, and the tool will reveal the case and give framework-based feedback.
Step 5. Submit. Save the whole session as a PDF (open your browser print menu, choose Save as PDF) and upload it to the Canvas assignment. Complete the short reflection if assigned.
IMPORTANT: Do not ask the tool to tell you the risk level or write your plan, and never enter real client or personal information. The clinical judgment must be yours.
What to Submit:
- Session record: A PDF of the full session (transcript plus the structured assessment summary), saved by using the browser print menu and choosing Save as PDF, then uploaded to the Canvas assignment.
- Reflection: A brief write-up on how your questioning shaped what the client revealed and how you appraised the feedback.
What is Google Gemini? Google Gemini is an AI-powered productivity tool designed to enhance your work and learning experience across Google Workspace. Read more at Google Gemini.
- How to login to Google Gemini:
- Navigate to Google Gemini
- Enter your FSU email address (fsuid@fsu.edu)
- Enter your FSU password
- Confirm on Duo, if needed
- Select the I understand button to acknowledge the Terms of Service (first time only)
What is a Gem? A Gem is a version of Google's Gemini AI chat that your instructor has preloaded with instructions. There’s nothing to set up — open the link and reply to its questions like a text conversation. Read more about Google Gems.
- Sign in. Open the Gem link and sign in with your Google account. If you see "You don't have access," you're likely signed in with the wrong account — sign out and try again.
- Send a message. Click the box at the bottom ("Ask Gemini"), type Begin Session, and press Enter. Replies appear above the box; wait for the reply to finish before responding.
- Reply in the same chat. The Gem remembers the whole conversation. Don't click "New chat", which will start the session over.
- Lost the page? Go to gemini.google.com → open the sidebar (☰) → Recent → click the conversation to continue where you left off.
- Copy the output. Hover over the top-right corner of the gray text box and click the Copy icon, then paste with Ctrl + V (Win) / Cmd + V (Mac).
- Good to know. Type text, rather than use voice (Gems don't work with Gemini Live). Don't enter passwords or sensitive personal information. Keep in mind that the Gem may ask you to expand a short answer.
- Access via FSU. Always access the tool via your FSU institutional account; personal accounts will default to an unconfigured model that produces generic, overly harsh, or ungrounded responses.