What KIN may process
During an active support session, KIN may process microphone audio, speech-to-text results, conversational timing, and recent context so it can decide whether an optional cue may be useful.
What the beta may store
- Pseudonymous participant and session identifiers.
- Structured events such as timing, repetition, direct-answer, clarification, relevance, and reciprocity opportunities.
- Generated KIN cue text, whether audio delivery succeeded, and basic latency or reliability measurements.
- Derived outcome signals, such as whether a two-second pause followed a timing cue.
What is not stored by default
- Continuous raw conversation recordings.
- Complete conversation transcripts.
- Names or diagnoses within the product-performance event stream.
Optional research recording
Any future research-recording mode should require separate, explicit consent. Ordinary use of KIN should not silently become research participation.
Human validation
Model conclusions may be incomplete or wrong. KIN’s development approach emphasizes user feedback and selective human review. The person with Autism remains the primary authority over whether a cue or interpretation was useful.
Third-party services
The prototype may use Apple services for speech recognition, Supabase for secure backend infrastructure, and OpenAI for context-aware cue generation and natural voice. Their processing is governed by their own agreements and privacy practices.
Not medical treatment
KIN is an experimental communication-support prototype. It is not a diagnostic tool, medical treatment, crisis service, or substitute for professional care.
Questions or deletion requests
For questions or deletion requests, contact outofthestoneage@gmail.com. Early testers may use this address to request access, correction, or deletion of their beta data.