FAICO: Framework for AI Communication

Overview

FAICO (Framework for AI Communication) identifies five dimensions of how AI systems communicate with human collaborators in co-creative contexts. Each dimension influences user experience, trust, collaboration feelings, confidence, enjoyment.

The framework emerged from systematic review of 107 HCI papers and validation through focus groups with AI/HCI/design practitioners.

The Five Dimensions

Modalities

The channels through which AI communicates: text, speech, visual, haptic, embodied.

Multimodal communication enhances user experience across demographics. Combining voice, embodied communication, dialogue, and facial expressions gives users more accurate perception of AI, improving collaboration feelings and social presence.

Response Mode

How AI initiates communication: proactive (AI initiates) or reactive (responds when prompted).

Proactive AI can feel intrusive in creative flow but improves trust and encourages critical evaluation of output. Reactive AI feels less annoying but may underserve users who don’t know what to ask.

Timing

When communication occurs: synchronous (real-time) or asynchronous (delayed/notification-based).

Synchronous suits improvisational co-creation and ideation. Asynchronous supports reflection and experienced users who want uninterrupted flow.

Types

What AI communicates: explanation (reasoning behind outputs), suggestion (alternatives and recommendations), feedback (critique of user contributions).

Divergent suggestions help early creative phases; convergent suggestions help refinement. Clear reasoning enhances trust, but disclosing uncertainty diminishes it.

Tone

How communication is expressed: politeness, warmth, friendliness, cultural alignment.

Warmth and friendliness improve satisfaction. Cultural alignment affects perception, users prefer AI aligned with their communication norms.

When to Use This Framework

• Designing AI communication in creative or collaborative systems
• Evaluating existing co-creative AI for communication gaps
• Identifying why a collaboration feels “tool-like” vs “partner-like”
• Structuring user research about human-AI interaction preferences

Limitations

• Validated primarily in Western academic contexts
• Focus groups were small (12 participants) and expert-heavy
• Creative domains studied were writing and design, may not generalize to music, physical making, etc.
• Doesn’t prescribe specific implementations, only dimensions to consider

Use Cases Proposed

• Design cards: Each dimension as a card for ideation
• Configuration tool: Let users customize AI communication preferences
• Evaluation checklist: Assess existing systems against the five dimensions

Related: 01-atom—communication-contribution-distinction, 01-atom—tool-vs-partner-perception, 07-molecule—ui-as-ultimate-guardrail