Affordances
An affordance is a relationship between an object's properties and the capabilities of the agent encountering it. A flat, rigid, knee-high surface affords sitting. A graspable, throwable object affords throwing. The affordance exists as a relationship whether or not anyone perceives it, but it becomes actionable only when perceived. Norman adapted Gibson's ecological concept for design by adding a crucial distinction: affordances can exist without their perceptual cues, and perceptual cues can exist without underlying affordances. The well-designed artifact aligns what it permits with what it advertises. The natural language interface, as the Norman volume argues, violates this alignment at a structural level — offering unbounded capabilities behind a blank text field that signals none of them.
In the AI Story
Norman imported affordances from James Gibson's ecological psychology, where the term described the meaningful possibilities of action that an environment offers to an organism. In Norman's hands, the concept became a design instrument: the designer's task was to ensure that the affordances of an artifact — what it actually enables — were perceivable as such. A door handle affords pulling, and its shape says so. A button affords pressing, and its raised appearance announces the possibility.
The AI era has broken this alignment in a way no previous interface managed. The blank prompt is an interface with effectively infinite affordances — the system can produce any outcome describable in natural language — and zero signifiers. Nothing about the empty text field tells the person what she can ask for, what kinds of requests yield good results, or where the system's capabilities end.
Compounding the challenge, AI affordances are person-dependent in a way no previous affordance has been. A button affords pressing to novice and expert alike. An AI system's affordances depend on the user's articulational capacity: the expert programmer perceives a space of sophisticated refactoring and architectural generation; the novice perceives basic script-writing. The underlying system is identical. The perceived affordance space is radically different, and the difference correlates with education, domain expertise, and linguistic fluency — making the technology that was supposed to democratize capability a new axis of inequality.
The design response, as Chapter 2 of the Norman volume argues, must be progressive affordance disclosure: systems that reveal their capabilities dynamically through conversation rather than requiring the user to arrive already knowing what to ask for. This is a significant departure from traditional affordance design, where the affordances were fixed properties of the artifact rather than emergent features of a negotiated interaction.
Origin
Gibson introduced affordances in his 1979 The Ecological Approach to Visual Perception, describing them as objective relations between environments and organisms. Norman adapted the concept in The Design of Everyday Things (1988) and subsequently refined his account — distinguishing affordances from signifiers in his 1999 essay "Affordance, Conventions, and Design" after observing that the design community had conflated the two in ways that weakened the concept's analytical power.
The AI-era reformulation, traced through the Norman volume's chapters on discoverability and the prompt as design object, recognizes that affordances in conversational interfaces must be treated as emergent and person-relative rather than static and universal — a concession Gibson would have found congenial and Norman found necessary.
Key Ideas
Relation, not property. An affordance is not a feature of an object but a relationship between object and agent. The same handle affords different actions to a child and an adult, to a practiced user and a novice.
Affordance without signifier, signifier without affordance. Glass doors afford walking through but lack signifiers (people walk into them). Painted doorknobs signify grasping but afford nothing (they connect to nothing). Good design aligns the two.
Person-dependent affordances. AI systems' capabilities vary by the user's articulational capacity. This is structurally new and creates inequality where previous interfaces produced uniformity.
Progressive disclosure, conversational edition. Traditional progressive disclosure revealed hidden features through layered interfaces. AI systems must reveal capabilities through dialogue — asking clarifying questions, proposing alternatives, expanding the user's sense of what is possible.
Debates & Critiques
Whether affordances are properties of objects, relations between objects and organisms, or purely perceptual constructs remains contested in ecological psychology. The design community has largely settled on Norman's relational formulation, but the AI era has reopened the question: if affordances depend on the user's articulational capacity, are they properties of the system, the user, or the interaction? The answer matters for how responsibility for discoverability is assigned.
Appears in the Orange Pill Cycle
Further reading
- James J. Gibson, The Ecological Approach to Visual Perception (Houghton Mifflin, 1979).
- Donald A. Norman, "Affordance, Conventions, and Design," Interactions 6, no. 3 (1999).
- Victor Kaptelinin, Acting with Technology: Activity Theory and Interaction Design (MIT Press, 2006).