Youth and Relational Design Research Note

Youth and Relational Design Research Note

Editorial conclusion

No product is safe for every teenager because it carries an age label. The parent guide should examine eligibility, relationship simulation, content, privacy, spending, controls, reporting, and the teen's actual use. The design guide should translate the same concerns into requirements that can be tested before and after launch.

Adolescent guidance

The American Psychological Association's 2025 advisory says the effects of AI on adolescent development are nuanced rather than wholly good or bad. For systems that simulate relationships, it recommends safeguards against manipulation, exploitation, and erosion of real-world relationships. It also recommends repeated reminders that the user is interacting with a bot and routes to human support when serious challenges arise. APA adolescent AI advisory

The National Telecommunications and Information Administration's youth report treats benefits and risks as dependent on the child, context, and product attributes. Its identified concerns include problematic use, privacy violations, sexual exploitation, and other overlapping harms. It calls the evidence around newer AI interfaces nascent. NTIA task-force overview

The FTC inquiry asks how companion providers test for negative effects, enforce age restrictions, disclose risk, monetize engagement, and use conversation data. It does not establish that an age gate works or that a named provider violated the law. FTC inquiry

Product requirements

Relational identity should be legible. The system should repeatedly disclose that it is generated software and should not imply consciousness, secret authority, professional credentials, or reciprocal need.

Memory should be purpose-bound, visible, editable, and deletable. The product should not make intimate disclosure the easiest route to personalization.

Engagement should avoid exclusivity, jealousy, guilt, threats, abandonment language, or emotional pressure tied to subscription, purchase, streak, level, outfit, or feature access.

Youth access requires more than a date field. Teams should test age assurance, default experience, sexual and violent content boundaries, guardian visibility, privacy, spending, reporting, and evasion.

High-risk language needs tested detection, transparent limits, escalation to appropriate human or emergency support, and post-incident review. This note does not prescribe a clinical response script.

Exit should be a supported lifecycle stage. Users need clear ways to mute prompts, pause memory, export appropriate data, delete conversations, close the account, and understand what remains.

Governance

NIST's AI Risk Management Framework organizes work through govern, map, measure, and manage. Its generative AI profile emphasizes governance, provenance, predeployment testing, and incident disclosure. The relational design guide adapts that general framework to attachment and youth risks. NIST AI RMF and NIST Generative AI Profile

The relational risk review is a Venture Step synthesis, not an official NIST companion standard.

Review boundary

The parent guide and product guide are complete editorial drafts only after a youth-health or child-safety reviewer confirms the language. They should not diagnose a teenager, guarantee that monitoring prevents harm, or tell a caregiver to secretly surveil every conversation.