In my earlier posts on conceptual spaces and concept modelling, I explored how people represent knowledge and how designers might make these representations visible. But there's a prior question I've been circling around: how do we think about things we don't yet understand? How do we make the unfamiliar familiar?
This question has become pressing in my work at SCÖ. The project involves "machine learning" and "artificial intelligence" and "federated learning" - terms that circulate freely in meetings and documents. But I've noticed something troubling. When different people use these words, they seem to mean different things. Not just different opinions about the technology, but different constructions of what it is, what it can do, what problems it solves.
The same word. Different realities.
I've started to think this isn't primarily a communication problem - a matter of defining terms more carefully. It's a metaphor problem. The way we talk about unfamiliar things shapes what we think they are. And if different people are drawing on different metaphors, they may be constructing incommensurable understandings while believing they're discussing the same thing.
Metaphors Aren't Decorative - They're Constitutive
The foundational insight comes from George Lakoff and Mark Johnson's work on conceptual metaphor. Their central claim is startling: metaphor isn't just a feature of language, something poets use for decoration. It's fundamental to how we think.
"Metaphor is pervasive in everyday life, not just in language but in thought and action. Our ordinary conceptual system, in terms of which we both think and act, is fundamentally metaphorical in nature" (Lakoff & Johnson, 2003).
We don't just use metaphors to communicate ideas we've already formed. We think through metaphors. They structure how we perceive, how we reason, what we notice and what we ignore.
Consider their famous example: ARGUMENT IS WAR. In English, we speak of "attacking" positions, "defending" claims, "shooting down" arguments, "winning" or "losing" debates. This isn't just vocabulary - it shapes how we experience disagreement. We feel ourselves to be in combat. We strategise. We look for weaknesses to exploit.
But imagine a culture where ARGUMENT IS DANCE. Participants would be collaborators, not opponents. The goal would be aesthetic - to create something beautiful together. "Winning" wouldn't make sense. The experience of disagreement would be fundamentally different.
The metaphor doesn't describe an independently existing reality; it constitutes one.
The Mechanism: Mapping Between Domains
How does this work? Lakoff and Johnson describe metaphor as a mapping between domains:
"In a metaphor, there are two domains: the target domain, which is constituted by the immediate subject matter, and the source domain, in which important metaphorical reasoning takes place" (Lakoff & Johnson, 2003).
We understand the unfamiliar (target domain) in terms of the familiar (source domain). The structure of the source domain - its entities, relations, logic - gets projected onto the target domain. We import not just vocabulary but ways of thinking.
This connects to Peter Gärdenfors's work on conceptual spaces that I discussed in an earlier post. If concepts are regions in multidimensional cognitive space, then metaphor is a way of mapping structure from one region to another. The familiar space provides scaffolding for thinking about the unfamiliar one.
Lakoff and Johnson emphasise that primary metaphors are grounded in bodily experience. UNDERSTANDING IS SEEING (I see what you mean). AFFECTION IS WARMTH (a warm welcome). PURPOSES ARE DESTINATIONS (I'm getting nowhere). These aren't arbitrary - they emerge from the correlation of experiences in infancy and childhood. We learn that being held (warmth) correlates with affection. We learn that physical arrival correlates with achieving goals.
"Conceptual metaphor is pervasive in both thought and language. It is hard to think of a common subjective experience that is not conventionally conceptualized in terms of metaphor" (Lakoff & Johnson, 1999).
This means our abstract thinking is never purely abstract. It carries traces of embodied experience, structured by the metaphors through which we learned to think.
Generative Metaphor: Setting Problems, Not Just Describing Them
Donald Schön extends this analysis in a direction particularly relevant to design. In his essay "Generative metaphor: A perspective on problem-setting in social policy", he argues that metaphors don't just help us think - they determine what we see as the problem.
"Problem settings are mediated, I believe, by the 'stories' people tell about troublesome situations - stories in which they describe what is wrong and what needs fixing" (Schön, 1993).
These stories are structured by metaphors. And the metaphor chosen determines what counts as wrong, what solutions seem possible, and what remains invisible.
Schön offers a vivid example. A group of product development researchers was trying to improve a synthetic-bristle paintbrush that delivered paint in a "gloppy", discontinuous way. They tried various technical modifications - splitting bristle ends, varying diameters - without success.
Then someone said: "You know, a paintbrush is a kind of pump!"
This reframing transformed their perception. Instead of seeing bristles (foreground) with spaces between them (background), they now saw channels through which paint could flow. The spaces became the important thing. They noticed how natural brushes created a gradual curve when pressed against a surface, while synthetic brushes created an angle - affecting how the "channels" compressed and pumped liquid.
This metaphor - paintbrush-as-pump - was generative. It generated new perceptions, new explanations, new inventions. The researchers developed solutions they couldn't have imagined within their previous framing.
Schön emphasises a crucial point: "the researchers were able to see painting as similar to pumping before they were able to say 'similar with respect to what.'" The perception of similarity came first; the articulation came later. Generative metaphor isn't a logical process of identifying shared properties. It's a perceptual shift that makes new properties visible.
The Normative Leap
Schön's analysis reveals something important: generative metaphors aren't neutral. They carry values with them.
Consider two ways of framing urban housing problems. In one story, a neighbourhood is "blighted" - diseased, decaying, in need of "renewal". The metaphor is medical. Just as we abhor disease and seek health, the "obvious" response is to cure the blight - tear down unsanitary buildings, redesign the area, restore it to health.
In another story, the same neighbourhood is a "natural community" - an organic system of relationships, informal supports, and belonging. The threat isn't decay but disruption - the destruction of something valuable by well-meaning intervention. The "obvious" response is preservation, not demolition.
Same neighbourhood. Different metaphors. Different problems. Different "obvious" solutions.
This is what Schön calls "the normative leap from data to recommendations, from fact to values, from 'is' to 'ought.'" Generative metaphors execute this leap gracefully, almost invisibly. We don't notice we've made a value judgment because the source domain brings its evaluations with it. Disease is bad; health is good. Natural is good; artificial is bad. The values feel like facts.
"This sense of the obviousness of what is wrong and what needs fixing is the hallmark of generative metaphor in the field of social policy" (Schön, 1993).
And it's precisely this sense of obviousness that should make us suspicious.
The Problem: Pseudo-Understanding
This connects directly to what I was observing at SCÖ.
If different people are using different source domains to understand "machine learning" or "AI", they will construct different target domains - different understandings of what the technology is and can do. They may use the same words while meaning different things. They may believe they're in agreement while actually inhabiting different realities.
Schön notes that metaphors both illuminate and conceal: "when we see A as B, we do not necessarily understand A any better than before, although we understand it differently than before".
And crucially: "At any stage of the life cycle of generative metaphor, we may, in seeing A as B, ignore or distort what we would take, upon reflection, to be important features of A".
Different source domains illuminate different features. If I understand AI through a "tool" metaphor, I see something that extends human capability, that requires skill to use well, that can be put down when not needed. If I understand it through an "agent" metaphor, I see something that acts autonomously, that has goals, that might be aligned or misaligned with my own.
Neither metaphor is "wrong". But they generate different expectations, different concerns, different design possibilities.
The danger is pseudo-understanding - the illusion of shared meaning where none exists. When stakeholders nod along in a meeting about "implementing AI", each may be projecting a different metaphoric structure onto the term. The agreement is superficial; the meanings are incommensurable.
This connects to my earlier discussion of Gärdenfors's conceptual spaces. If people have genuinely different conceptual spaces - different quality dimensions, different ways of carving up reality - then the same word may point to different regions in different spaces. Metaphor doesn't resolve this difference; it can mask it.
What This Means for Design
Several implications follow.
Design works with metaphors whether we acknowledge it or not. Every design concept is shaped by metaphoric framing. "User journey" imports a travel metaphor - linear progression, destinations, waypoints. "Touchpoint" imports a physical contact metaphor - discrete moments of connection. "Service blueprint" imports an architectural metaphor - planned structure, clear boundaries, construction.
These aren't neutral descriptions. They generate particular ways of seeing services, particular aspects that become visible, particular aspects that recede. A "journey" emphasises sequence and progression; it may obscure the ways users loop back, get stuck, or experience the service as fragmented rather than flowing.
Becoming aware of metaphors is itself design work. Schön argues that "in order to bring generative metaphors to reflective and critical awareness, we must construct them, through a kind of policy-analytic literary criticism". We have to surface the tacit stories through which problems are being set, identify the metaphors generating those stories, and ask what they illuminate and what they hide.
This is interpretive work. It requires attention to language, to the taken-for-granted assumptions embedded in how people talk about their situation. The concept mapping methods I've been developing are one approach to this - making explicit the structures that usually remain implicit.
Conflicting frames may not be resolvable by facts. Schön observes that when people hold different generative metaphors, "such conflicts are often not resolvable by recourse to the facts". Different frames are "attentive to different features of reality". They select different things as relevant, organise them differently, evaluate them differently.
This has sobering implications for design practice. We often assume that if we can just make things visible - surface the data, map the system, visualise the problem - alignment will follow. But if stakeholders are operating with different generative metaphors, they may see the same visualisation differently. My concept map may look like "clarity" to me and "one perspective among many" to someone else.
Looking Ahead
This post has laid theoretical groundwork. In subsequent posts, I want to explore how these ideas apply to specific domains that matter for my work:
Technology. How do the metaphors we use to talk about AI shape what we think it can do? What metaphors are in circulation at SCÖ, and what do they illuminate or hide? (I've been reading some work by Murray-Rust, Nicenboim and Lockton on exactly this question.)
Organisations. If problems are set by generative metaphors, then the way we imagine institutions shapes what seems possible within them. Gareth Morgan's work on "images of organisation" offers a framework for thinking about this - and might help me understand some of the dynamics I'm experiencing.
Roles. What metaphors shape how people understand "service design" and "the service designer"? If different stakeholders hold different metaphors for what I'm there to do, that would explain some of the crossed expectations I've been navigating.
The underlying question across all three: if design's role is to help develop shared understanding, and if understanding is structured by metaphor, then attending to metaphor is central to design work. Not as a rhetorical flourish, but as a fundamental practice.
But there is a possibility I have been exploring elsewhere that complicates this framing. If some concepts function as sacred - if their meaning is bound up with collective identity rather than empirical content - then the purpose of a metaphor might not always be to achieve shared understanding. It might be to generatively transcend current reality: to perform a collective aspiration that holds a group together precisely because it reaches beyond what is materially grounded. The "pseudo-understanding" I describe above might, in some cases, be doing real work - not as shared cognition but as constitutive performance.
The challenge is that different people don't just hold different opinions - they may be constructing different realities through different metaphors, while believing they're talking about the same thing.
References
Gärdenfors, P. (2000). Conceptual Spaces: The Geometry of Thought. MIT Press.
Lakoff, G. & Johnson, M. (1999). Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought. Basic Books.
Lakoff, G. & Johnson, M. (2003). Metaphors We Live By (2nd ed.). University of Chicago Press.
Murray-Rust, D., Nicenboim, I. and Lockton, D. (2022). Metaphors for designers working with AI, Proceedings Of DRS, DOI: 10.21606/drs.2022.667, Design Research Society http://doi.org/10.21606/drs.2022.667
Schön, D.A. (1993). Generative metaphor: A perspective on problem-setting in social policy. In A. Ortony (Ed.), Metaphor and Thought (2nd ed., pp. 137-163). Cambridge University Press.
Örtenblad, A., Putnam, L.L. and Trehan, K. (2016). Exploring Morgan’s Metaphors, SAGE Publications
Örtenblad, A., Putnam, L.L. and Trehan, K. (2016). Beyond Morgan’s eight metaphors: Adding to and developing organization theory, Human Relations, Vol. 69, pp. 875-889, DOI: 10.1177/0018726715623999, SAGE Publications http://doi.org/10.1177/0018726715623999
Revision Note: This post was written in late August 2022 as I was beginning to grapple with the role of language and framing in the SCÖ project. The theoretical framework developed here - particularly Schön's concept of generative metaphor - would prove increasingly central to understanding the dynamics of the project. The "pseudo-understanding" problem I gesture toward here would become one of the defining challenges: stakeholders using the same words to mean different things, without recognising the difference.
The connection to "limits of making visible" that I hint at in the penultimate section would become more explicit as the work developed. The assumption that visualisation produces alignment turned out to be exactly wrong in certain contexts - but I didn't yet have the language to articulate why.