TABLE 3.

The six major challenges of HCAI, considering implemented design decisions and current technical limitations.

HCAI challenge	Design decision	Technical evidence	Main limitation
Human wellbeing	Charismatic and adaptive character design; cocreation with children	Customized voice model (VITS); behavioral modulation; public interaction in events and workshops	Absence of systematic user perception studies
Responsible design	Preference for locally deployed SLM and controlled prompt architecture	Local execution (Gemma 3); structured memory (JSON); personality modulation	Lack of formal bias auditing and explainability tools
Privacy	Embedded processing and offline operation	On-device inference (Jetson); local data storage	Data governance policies are not yet formally consolidated
Design and evaluation framework	Iterative development process (TRL 5–6) and migration to ROS 2	Modular multimodal pipeline; latency measurements; multithreaded architecture	Absence of controlled experimental benchmarking
Governance	Local autonomy strategy and restriction in high-risk domains	Non-deployment in healthcare; institutional control of the AI stack	No formal external ethical oversight
Human–AI interaction	Embodied multimodal architecture	Synchronized lip-sync; eye tracking; parallel vision–speech execution	Lack of quantitative comparison with the chatbot baseline