DEVOURED
The Layers of AI Experience
Designer Emily Campbell proposes a six-layer framework to help designers influence AI products beyond just the user interface.
What: The model divides AI experience into Interface, Context, Harness (coordination systems), Model (training/capabilities), Governance (policies/rules), and Emergence (unpredictable behavior) to help designers build more reliable AI systems.
Why it matters: As AI moves from deterministic rules to probabilistic outcomes, designers must move from 'interface designers' to 'systems designers' who can influence the data and constraints feeding the model.
Deep dive
- User Interface: The surface level for oversight and input; roles shift from direct instruction to orchestration.
- Context: The 'map' of user needs, history, and preferences; requires deliberate engineering to avoid 'context rot'.
- Harness: The operational infrastructure (tools/connectors) that allows the AI to perform tasks; manages autonomy levels.
- Model: Selection based on capabilities (reasoning vs. speed); designers must understand model strengths/weaknesses.
- Governance: The layer of compliance and standards; designers need to advocate for policy within the product experience.
- Emergence: Managing the inherent unpredictability of LLMs; requires observability, interpretability, and provenance tools.
Decoder
- Probabilistic design: An approach to product building where outputs are not guaranteed to be consistent, necessitating systems to monitor and manage variability.
- Context rot: The degradation of an AI's performance due to overstuffed memory or conflicting data, leading to inconsistent or incorrect responses.
- Provenance: The ability to trace a specific AI-generated output back to the exact inputs and context that produced it.
Original article
Full article content is not available for inline reading.
Explore all 71 articles in today's digest
Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi