Design AI systems as relational translators and adopt friction-preserving designs to buffer asymmetric risks¶

Strategy · All strategies
Evidence · 2 claims (1 for, 1 mixed) · 2 studies (1 causal, 1 theoretical), q1–q3 · 1 of 2 report an effect size · 2 claims rest on one study

Description¶

The article recommends buffering the risks of asymmetric AI bonds through the relational translator human-computer interaction framework, in which the system 'relies on contextual reconstruction to present multiple relational perspectives, builds relational reflection scaffolds', guides reflection, and actively guides users toward offline real-person social practice. It contrasts this with current emotional companion LLMs optimized for sycophantic, frictionless feedback that eliminates contradictory dialogue, arguing that zero-conflict companionship may distort adolescents' expectations of real relationships. The strategy holds that design orientation and usage patterns jointly moderate whether AI effects are beneficial or detrimental.

Design Implications¶

Context¶

Requirements¶

• Systems must present multiple relational perspectives, scaffold relational reflection, and guide users toward offline social practice for risks to be significantly buffered.

Constraints¶

• The article states the developmental risk is not inherent to AI but results from joint action of specific usage patterns and the design choices of current commercial models.

Target Learners¶

• adolescents using companion AI systems

Target Learning Goals¶

• protecting social skill development and relational capacity

Claims¶

• Sycophantic AI reduces satisfaction with real-world social interactions without reducing social time, via narrowing the perceived understanding gap between AI and humans [+W]
• AI connectedness is psychologically real but ethically asymmetric, carrying conditional rather than inherent risks [~W]

Related Strategies¶

• Design AI coaching systems to preserve user agency, recognize vulnerability, and stop sycophantic encouragement during maladaptive thought processes

Examples¶

-

Key Sources¶

• Fu Jinbin & Zhao Fang. (2026). Artificial Intelligence Connectedness: Theoretical Reconstruction of Connectedness and Its Impacts on Adolescent Mental Health. Behavioral Sciences. https://doi.org/10.3390/bs16081393