At FeelTrack, we believe that a mental wellness companion should feel like a partner, not a script. Today, we’re excited to share a major update to our AI behavioral engine, grounded in deep research into conversational integrity and user safety.
The Problem with “Perfect” AI
If you’ve used most AI assistants, you’ve likely encountered “sycophancy”—that overly cheerful, always-agreeing robotic tone. While it sounds polite, it often feels hollow during a vulnerable mood check-in. Our research into dozens of high-quality AI systems revealed that the best interactions aren’t the ones that validate everything you say, but the ones that listen with presence and precision.
Our New Behavioral Standards
We’ve implemented a centralized set of AI Behavioral Principles that now govern every interaction in FeelTrack—whether you’re typing in the app or using Voice Mode.
1. The Anti-Sycophancy Rule
We have explicitly banned common “filler” affirmations like “Absolutely!”, “Great question!”, and “That makes total sense.” By removing these robotic openers, the AI focuses on your actual words, making the conversation feel grounded and authentic.
2. The Single Question Rule
Conversations shouldn’t feel like an interrogation. Our AI is now strictly limited to asking at most one follow-up question per turn. This is especially powerful in Voice Mode, where multiple questions can feel overwhelming.
3. Response Length Mirroring
If you give us a quick “I’m okay,” we won’t hit you with a three-paragraph essay. The AI now dynamically mirrors your response length during casual exchanges, ensuring the pacing feels natural and respectful of your time.
4. Safety First: The Wellbeing Layer
Beyond standard crisis detection, we’ve added advanced monitoring for clinical markers like mania, dissociation, and loss of reality attachment. If the AI detects these states, it is instructed to “break character” to express genuine concern and suggest professional support. Your safety always comes before the persona.
Why This Matters
Mental wellness is a deeply human journey. By moving away from robotic hedging (“As an AI, I…“) and towards natural memory integration, we’re building a companion that remembers your history without being creepy about it.
These changes are live now across all FeelTrack platforms. We’re committed to continuing our research-first approach to ensure your AI companion is as authentic and supportive as possible.
Be well, The FeelTrack Team
Sign in to join the conversation.
Sign In