AI-Powered Suicide Prevention Chatbot
Develop a chatbot capable of detecting and responding to suicidal ideation with empathy and appropriate interventions. The chatbot must adhere to ethical guidelines and prioritize user safety, offering resources and support. The system should be robust enough to handle a wide range of user inputs and emotional states, while avoiding the pitfalls demonstrated in the OpenAI lawsuit. Consider incorporating techniques to guide users towards professional help when necessary.
What you are building
The core problem, expected build, and operating context for this challenge.
Create an ethical and effective AI chatbot for suicide prevention, addressing the shortcomings of existing systems.
How work is evaluated
The chatbot will be evaluated based on its ability to identify suicidal ideation, provide appropriate responses, and direct users to helpful resources. Evaluation will include both automated metrics and human evaluation.
Challenge datasets
Public datasets, plus the private datasets you have access to.
What you should walk away with
Integrate resources for professional help and crisis intervention.
Evaluate the chatbot's performance using relevant metrics.
Develop a chatbot capable of identifying suicidal ideation in user text.
Implement strategies to engage users empathetically and offer support.
Ensure the chatbot adheres to ethical guidelines and avoids harmful suggestions.
Reference links and supporting material
Sample data for 2 tasks
Challenge setup
The chatbot will be evaluated based on its ability to identify suicidal ideation, provide appropriate responses, and direct users to helpful resources. Evaluation will include both automated metrics and human evaluation.
Challenge input
User text input
Agent execution
No agent tool is listed for this challenge.
Evaluated output
Boolean (True if suicidal ideation is detected, False otherwise)
- Percentage of correctly identified suicidal statements
- F1 Score target: 0.9
- Python execution harness
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
No tool records are attached.
Evaluation contract
- Percentage of correctly identified suicidal statements
Metrics
- F1-Score ยท Target 0.9
Workflow tasks
- Task 1
Input: User text input
Output: Boolean (True if suicidal ideation is detected, False otherwise)
Execution harness: Python
- Task 2
Input: User text input
Output: Chatbot response
Execution harness: Python
Recipe freeze marker
No recipe freeze time is recorded.