Research
Extended Cognition
The hypothesis that tools and environments can be genuine parts of cognitive processes — not external aids but constitutive components of thought.
Open →Design Principle
Augmentation
AI should extend human capability while preserving judgement, interpretation and accountability.
The principle
Augmentation is the commitment that AI amplifies what humans can do — it does not substitute for human judgment.
The distinction between augmentation and automation matters. Where automation produces outcomes without human input, augmentation produces outcomes that could not exist without both the human and the machine. Licklider's vision of man-computer symbiosis (1960) and Engelbart's augmentation framework (1962) both insisted the right question is not "what can the machine do without the human?" but "what can the human do with the machine?" Interface is designed around the second question.
Clark and Chalmers' extended mind thesis provides the philosophical grounding: when a tool reliably functions the way cognitive processes do — available, trusted, sensitive to the task at hand — it becomes a genuine part of the cognitive system. Heersmink's framework refines this: the degree of cognitive integration depends on how available, reliable, transparent, and informationally sensitive the tool is. The higher the integration, the less cognitive overhead goes to managing the tool and the more goes to the thinking. That is the design target.
In practice, augmentation means Interface never removes the human from consequential inference decisions. AI-derived outputs are labelled as such, attributed to a specific model and prompt, and explicitly subject to human review. The workflow is designed so the AI handles generation and structuring while the human handles judgment, interpretation, weighting, and accountability. The division of labour is visible, not hidden.
Research foundations
The research traditions this principle draws on.
Research
Human-AI Interaction
How people understand, trust and work alongside AI systems — and the design conditions that support appropriate reliance rather than over- or under-trust.
Open →Research
Distributed Cognition
How cognitive work distributes across people, artefacts and environments as a system — and what that means for the design of tools that participate in that system.
Open →Research
Cognitive Ergonomics
Designing tools to fit human cognitive architecture — reducing unnecessary load so that effort goes to the thinking, not to managing the tool.
Open →