Three years have passed since the concentrated sequence of posts that built this series' formal apparatus - conceptual spaces, state spaces, graphs, promises, statecharts, service grammar - and applied it to institutional critique in "Beyond Technomagic". Since then, the concept has been tested against practice (the statechart work on a data access service), extended through political analysis (the politics of formalism, the formalisation/reification tension), disambiguated (four senses), and grounded as a mode of design inquiry (co-evolution and practice). This post takes stock.
What the 2023 apparatus got right
The central claim - that state-space construction is design work, not technical specification - has been reinforced by every subsequent application. The statechart work confirmed it practically: making states explicit revealed dead-end states, unassigned transitions, and parallel processes that journey maps had concealed, and the questions the formalism forced ("what triggers this transition?", "is there a path out of this state?") were recognisably design questions, not engineering questions. The institutional critique confirmed it politically: technomagic is, structurally, an attempt to plan without the design work that state-space construction represents. The evaluative capacity work will confirm it philosophically: the judgement required to construct a useful state space is closer to phronesis than to technical skill.
The multi-register coherence - the concept's ability to operate across formal, representational, institutional, and governmental levels - has also held up, though it required the disambiguation to make this coherence explicit rather than relying on the reader to track the shifts. The four senses are genuinely different, but they are connected by a shared structure: all four involve constructed frameworks within which purposeful action becomes possible, and in all four the construction is a design activity.
The critique of service design's existing representational toolkit - that journey maps, blueprints, and personas leave states implicit - remains valid. Nothing I have encountered in the subsequent literature or in practice has challenged the observation that action-oriented representations dominate service design and that state-oriented representations would surface dynamics that the dominant tools conceal. Penin and Prendiville's (2025) recent Bloomsbury Handbook entry on service design representations confirms the field's continued emphasis on touchpoints, journeys, and experiences; the state dimension remains largely unaddressed.
What subsequent work complicated
Several aspects of the original apparatus now look less straightforward than they did in 2023.
"Constructed, not given" was the founding claim, but as "Formalisation and Reification" argued, it is better articulated as partly discovered, partly constructed, partly negotiated. The domain has a constraint structure (Burns and Hajdukiewicz, 2017) that shapes what state spaces are possible; the designer selects from the constrained field rather than constructing freely; and different stakeholders will construct differently from the same domain, making the resolution a political process as much as a cognitive one. The original formulation captured the constructedness without capturing the constraints.
The relationship between design and planning is not the temporal sequence that "logically prior" can inadvertently suggest. As the co-evolution post developed, drawing on Dorst (2015) and Gedenryd (1998), the state space and the plan develop together: provisional planning reveals what the state space needs; state-space revision reshapes what plans are possible. The logical priority of state-space construction - the fact that it determines what planning can attempt - is real, but it manifests as a recurring activity within the design process rather than as a preceding phase. This matters for practice because it means state-space thinking cannot be confined to a "discovery" phase; it must persist throughout.
Formalisation is both the solution (making implicit states explicit reveals what journey maps conceal) and the risk (institutional governance converts the designer's provisional model into a reified framework). The tension between formalisation-as-inquiry and formalisation-as-closure is genuine and cannot be resolved by choosing one over the other; the designer must hold both simultaneously, using formalisms to illuminate the domain while resisting their institutional transformation into fixed categories. This is what evaluative capacity - the "compass" - makes possible, and it is why the practice account describes a mode of inquiry rather than a method.
The politics of formalism choice - the recognition that the choice of algorithmic approach (genetic algorithm, classical planner, Bayesian network, reinforcement learner, language model) shapes the state space as much as the choice of variables - was absent from the original apparatus. The 2023 posts treated the Pathway Generator's computational requirements as a given; the later analysis showed that the genetic algorithm's requirements were themselves a design decision with political consequences. "Constructed, not given" now applies to the formalism as well as to the state space.
The mature position
Reading across the full arc, the argument as it currently stands can be reconstructed in seven propositions:
State spaces are constructed - partly discovered from the domain's constraint structure, partly designed through the selection of dimensions and granularities, and partly negotiated among stakeholders with different conceptual spaces and different interests.
Planning presupposes a state space; design constructs one. This makes design the more fundamental activity, though not the temporally prior one; the two co-evolve, with state-space construction recurring whenever the current framework proves inadequate.
Service design's existing representational toolkit leaves states implicit. Journey maps, blueprints, and personas contain implicit state thinking but do not make it formal; this creates gaps, ambiguities, and translation problems that state-oriented representations can surface.
Statecharts and related formalisms offer a way to make states explicit. They reveal dead-end states, unassigned transitions, and parallel processes that existing tools conceal; their value lies not in the completeness of the resulting model but in the questions the formalism forces.
The choice of formalism is political, not merely technical. Different computational approaches encode different ontological commitments about what the domain is and what knowledge counts; the choice shapes the state space as much as the choice of variables.
Institutional governance tends toward formalisation-as-closure - the conversion of the designer's provisional models into reified frameworks. The framework that governs institutional action is not the designer's situated model but a politically determined simplification. Resisting this closure - keeping formalisation in the inquiry register - requires the evaluative capacity that makes design work something more than specification.
The concept's analytical power comes from the connection between its four senses: formal requirements (sense 1) make legible what is absent in institutional planning (sense 3), with representational design practice (sense 2) providing the bridge. Simon's (1996) canonical definition of design - devising courses of action aimed at changing existing situations into preferred ones - is, implicitly, a state-space description; the series' contribution is to make this explicit and to show what it demands.
What remains open
The practice account is necessarily incomplete. "State-Space Construction as Design Inquiry" describes a mode of inquiry rather than a method, and this is deliberate - state-space construction in complex domains cannot be fully proceduralised - but it means the practical guidance available to a designer encountering this work for the first time is limited to "think in states, use these tools, iterate". Whether a more developed repertoire can be articulated without falling into the GPS trap (turning design into a process step) remains to be tested.
The relationship between the full 2023 apparatus and the more pragmatic statechart work has not been resolved. The 2023 sequence built five layers - conceptual spaces, state spaces, graphs, promises, grammars - each building on the previous. The practical work has drawn primarily on statecharts (one layer) and conceptual spaces (another), while the promise theory and grammar layers have remained largely theoretical. Whether these additional layers are essential to the argument or are theoretical scaffolding that served the construction but is not needed for the finished building is an open question. My suspicion is that promise theory remains important - Burgess's (2020) insight that "computation encodes policy" connects the formal and political dimensions directly - but the grammar work may need to be reframed as a research programme rather than a practical tool.
The connection to the thesis on institutional design practice - how the state-space argument relates to the broader questions about design's role in programme management, about the reification gap, about what the designer brings to an institutional setting - has been sketched but not fully developed. The state-space concept is one thread in a larger argument about design's strategic importance in complex institutional environments; it provides the formal grounding for claims about why design is necessary (because someone must construct the framework) and what happens when it is absent (technomagic, procedural planning, reified categories). But how this thread weaves together with the political, psychological, and organisational threads of the thesis remains work in progress.
The contribution
The series' distinctive contribution is the bridge it builds between formal and situated traditions in design. The formal tradition - AI planning, state-space theory, computational ontology - provides precision: it makes specific demands on what a domain model must contain and can verify whether those demands are met. The situated tradition - reflective practice, ecological interface design, participatory design - provides wisdom: it insists that formal models are always partial, that the domain exceeds any representation, and that the people who live and work in the domain have knowledge that formal models cannot capture. The state-space concept, as developed across this series, sits at the intersection: it uses formal requirements to make visible what situated understanding knows is absent, while using situated understanding to keep formal models honest - provisional, revisable, in the service of inquiry rather than closure.
Whether this bridge holds - whether it can bear the weight of practical application in complex institutional environments, or whether it remains a theoretical construction - is the question the next phase of this research, and the thesis, will need to answer.
References
Burgess, M. (2020). A Treatise on Systems Volume 2. ChiTek-i.
Burns, C. and Hajdukiewicz, J. (2017). Ecological Interface Design. CRC Press.
Dorst, K. (2015). Frame Innovation: Create New Thinking by Design. MIT Press.
Gedenryd, H. (1998). How Designers Work: Making Sense of Authentic Cognitive Activities. PhD thesis, Lund University.
Penin, L. and Prendiville, A. (2025). Service Design Representations. In Sangiorgi, D. and Prendiville, A. (eds), The Bloomsbury Handbook of Service Design Research. Bloomsbury.
Simon, H. (1996). The Sciences of the Artificial (3rd ed.). MIT Press.