In last week's post, I flagged that the word "event" is doing heavy lifting across the frameworks I have been drawing on. I claimed that Harel's statechart events and Iqbal's service grammar Events are "fundamentally different framings". That is true, but it undersells what is going on. The difference is not just terminological - it reflects four distinct ontological commitments about what it means for something to happen in a service system.
This post extends the earlier analysis of events, lives, and systems, which established the foundational distinction between moments (events) and processes (trajectories through state space). Here I examine what happens when "event" itself is placed under pressure across four theoretical traditions: Harel's statecharts, Gärdenfors's conceptual spaces, Iqbal's service grammar, and Burgess's promise theory. Each offers a different answer to the question "what is an event?", and the differences matter for how we model, design, and reason about services.
Harel: events as discrete signals
In Harel's (1987) statecharts, an event has a precise meaning: an instantaneous occurrence that triggers a state transition. Harel and Politi (1998) are explicit: "The input to a reactive system consists of a sequence of stimuli - events and changes in the values of data elements - that are generated by the environment". Events arrive from outside the system or are generated internally by other parts of it. They are punctual - they happen at an instant. They have no duration, no internal structure, no semantics beyond the fact of their occurrence.
The statechart framework draws a sharp distinction between events and conditions. Events are instantaneous; conditions are persistent. A condition is something that is true at a given moment - the system can check it. An event is something that happens - a signal that can trigger a transition. "The trigger of a transition may be an expression that combines some events. It may also include a condition, enclosed in square brackets" (Harel and Politi, 1998). The condition guards the transition; the event triggers it.
Harel's events are deliberately minimal: they carry no information about why they happened, who caused them, what they consist of, or what they mean. A referral arriving, a button being pressed, and a timeout expiring are all events in the same sense. The semantics are entirely in the state model - what state the system was in, what state it transitions to, and what actions are generated as a result. This minimalism is what makes statecharts tractable for modelling complex reactive systems.
However, a statechart treats REFERRAL_RECEIVED identically whether the referral was a routine administrative transfer or an urgent escalation driven by clinical deterioration. The trigger is the same; the state logic fires the same way. Whatever matters about the character, the motivation, and the experience of the event is stripped away. For implementation this may be acceptable; for design - where we need to reason about why things happen and what they mean to the people involved - it is a significant loss.
Gärdenfors: events as structured transformations
Gärdenfors's (2017) treatment in The Geometry of Meaning is radically richer. In the conceptual spaces framework, an event is not a signal but a structured transformation with at least two vectors and one object. His two-vector condition states: "An event must contain at least two vectors and one object; these vectors are a result vector representing a change in properties of the object and a force vector that causes the change".
A prototypical event involves an agent generating a force vector (or more generally a force pattern) that affects a patient, causing a change in the patient's properties. Oscar pulling a sledge up a hill is an event: the force vector is Oscar's pulling, the result vector is the sledge's movement uphill, the patient is the sledge, the agent is Oscar, and counterforces (gravity, friction) explain why the result vector does not simply mirror the force vector.
The event has internal structure: an agent who generates the force (sometimes absent - rain falls without an agent), a patient who undergoes the change, a force vector representing the cause, a result vector representing the change in the patient's properties, and optionally instruments, recipients, counterforces, and goals. This connects directly to the thematic roles that Fillmore's (1968) case grammar identifies and that the case grammar post explores in detail.
Crucially, Gärdenfors defines states as limiting cases of events: "In the limiting case when the result vector is the identity vector (with zero length), the event is a state. However, an identity result vector does not imply that the force vector is zero. A state can be maintained by balancing forces and counterforces, for example, when a prop prevents a wall from falling". What Harel calls a state (a configuration the system is in), Gärdenfors reveals as a special case of an event where forces are in equilibrium. The wall is not falling - that is a state - but forces are actively being exerted to keep it that way.
There is a question, however, about how far this model extends. Gärdenfors developed it for the cognitive semantics of physical actions - pushing, pulling, moving objects through space - where force and result have intuitive physical meaning. When we apply it to abstract service events, the "force vector" of submitting a benefits application is institutional (a system accepting data), procedural (validation rules being applied), and social (a claim being registered with an authority). The metaphor is productive - it directs attention to causation and transformation - but it should be held lightly. The same caution applies here as in the case grammar post: the force dynamics vocabulary is a resource for design thinking, not a literal mechanics of institutional action.
Iqbal: events as persistent conditions of unmet need
In Iqbal's (2018) 16x frame, Event (2T) occupies a specific structural position: it is the "why" factor of the "experiences" promise. The definition is: "Things with shortfalls, or why there will be demand for affordances." Where Harel's events are instantaneous signals and Gärdenfors's events are structured transformations, Iqbal's Events are conditions - persistent states of the world that create demand.
As I mapped out in the earlier post on Iqbal's grammar, the four "why" factors in the 16x frame form a revealing set. An Artefact (2A) is a thing with shortcomings that creates demand for performances; an Event (2T) is a thing with shortfalls that creates demand for affordances; a Capability (2C) is a thing with skills that supplies performances; a Resource (2R) is a thing with surpluses that supplies affordances.
Each is a persistent condition. An artefact has a shortcoming that endures until a performance addresses it; an event has a shortfall that endures until an affordance meets it. These are not happenings; they are states of the world that explain why the service exists.
A patient on a waiting list with a deteriorating condition is an Event in Iqbal's sense. The Event does not happen at an instant; it persists as long as the shortfall exists. In the earlier discussion, I said Iqbal's Event is "a state, not an event at all" in Harel's terms. That is correct for the mapping, but its implications deserve more development.
If the foundational demand condition for a service is not an event but a state, then the entire "event-driven" framing of service design may be misconceived. Services may be fundamentally about maintaining, transforming, or failing to transform states - not about responding to events. The verb-centred framing I critiqued in the beyond good services post shares this bias toward action and happening; Iqbal's usage quietly reorients attention toward the persistent conditions that make action necessary.
Burgess: events as assessments
Promise Theory adds a fourth perspective. In Burgess's (2020) framework, events are fundamentally linked to observation and assessment. "Every change involves a fundamental act of observation (an assessment) of promises kept (events occurred)". An event, in Promise Theory, is what an agent observes when assessing whether a promise has been kept. "The events are assessed to have occurred when the observer accepts the promise of the events as kept".
This is a deeply perspectival view. Events do not exist independently of observers. What counts as an event depends on who is assessing, what promises they are attending to, and what their scope of observation includes. An "Event Driven Agent makes a promise conditionally on the sampling of message events" - the agent only responds to events it has promised to attend to.
For service systems, this has a an implication: the same happening in the world may constitute different events (or no event at all) depending on the observer. A referral arriving is an event for the receiving team (it triggers their workflow), but it may not be an event for the patient (who does not know it has happened) or the commissioner (who only sees aggregated data). The event is not a fact about the world; it is a relationship between an occurrence, an observer, and a promise.
The political implications are worth pressing. In public services, whose observations count as events is never a neutral question. If a patient's deterioration is not captured by the system's event vocabulary - if no sensor samples it, no promise conditions on it - then it does not exist from the system's perspective, however real it is to the patient. The choice of what to monitor, what to count, and what to respond to is a design decision with distributional consequences. Burgess's framework makes this visible in a way that the other three do not.
The referral across four frameworks
Consider a GP referring a patient to a specialist. What is the event? Each framework gives a different answer, and the differences are instructive.
For Harel, the event is the signal REFERRAL_RECEIVED that triggers a transition in the specialist service's statechart - from "no active referral" to "referral under review". It is instantaneous, structureless, and triggers state logic. For Gärdenfors, the event is a structured transformation: the agent (GP) generates a force vector (clinical judgment, formal referral action) that affects the patient; the result vector is a change in the patient's status from "managed in primary care" to "referred to specialist"; the event has manner (urgent vs routine), instrument (e-referral system), and counterforces (pressure not to refer, capacity constraints in the receiving service).
For Iqbal, the event is not the referral itself. The Event (2T) is the patient's condition - the ongoing state of clinical need that creates the demand for the referral in the first place. The patient's shortfall is what explains why the service promise of specialist care exists. The referral is a response to the Event, not the Event itself. For Burgess, the event depends on the observer. For the specialist team, it is the assessment that a promise of referral has been kept - they observe the referral arriving and assess whether it meets the criteria they have promised to respond to. For the patient, the "event" might be the GP saying "I am going to refer you" - an assessment of the GP's promise to help. For an auditor, the event is whatever appears in the data system as evidence that a referral was made.
Each perspective captures something the others miss. Harel captures the logic of what happens next: given the signal, what transitions fire, what states change, what downstream events are generated. Gärdenfors captures the cognitive structure of the happening - the forces at work, the changes undergone, the roles played - connecting events to human understanding. Iqbal captures the economic rationale: why does this service exist, what demand does it meet, what promise is being fulfilled. Burgess captures the epistemological structure: who knows what happened, from whose perspective, under what conditions is the happening recognised as an event at all.
Aspect: the temporal structure of service events
One dimension the comparison reveals as underdeveloped is the temporal structure of events. Gärdenfors (2017) provides the richest treatment here, drawing on Vendler's (1957) classic aspectual classification. Vendler distinguished four categories of events (to which many researchers have added a fifth), classified by whether the event is extended or punctual and whether it has a definite endpoint (fixpoint) or not.
An accomplishment is extended and telic - it unfolds over time and ends in a definite state. Building a bridge is Vendler's example. An achievement is punctual and telic - it happens in an instant and reaches a definite endpoint. Realising an error is an achievement. An activity is extended and atelic - it unfolds over time but has no inherent endpoint. Walking is an activity. A semelfactive is punctual and atelic - it happens in an instant but has no definite endpoint. Knocking is a semelfactive.
Gärdenfors (2017) gives these categories a geometric interpretation: aspects are path properties of the result vector. An accomplishment is a path with an endpoint; an activity is a path without one. An achievement is a punctual change to an endpoint; a semelfactive is a punctual change without one. In the limiting case where the result vector has zero length, the event is a state.
These distinctions matter for service design because different service events fall into different aspectual categories and this has design consequences. "Book an appointment" is an accomplishment: it unfolds over time (search, select, confirm) and ends in a definite state (appointment booked). "Wait for a referral" is an activity: it is extended but has no inherent endpoint from the patient's perspective - it simply continues until something else happens. "Receive a notification" is an achievement: it is punctual and reaches a definite state (notified). "Be eligible for a benefit" is a state: no change, but forces (eligibility rules, evidence requirements, institutional processes) are in play.
These distinctions are invisible to Harel's model, which treats all events as punctual triggers regardless of their internal temporal structure. They are also invisible to Iqbal's model, which treats events as persistent conditions. Only the Gärdenfors framework provides the vocabulary to distinguish them, and the distinctions have consequences for design. An accomplishment needs progress indication and the possibility of abandonment. An activity needs duration awareness and escape. An achievement needs confirmation and reversibility. A state needs monitoring and threshold detection. Croft's (2012) three-dimensional decomposition of events - temporal, qualitative, and causal - reinforces this: events can be broken apart along each dimension, and attending to only one (as Harel's model does with the causal-logical dimension) systematically obscures the others.
What the comparison reveals
These four frameworks are not just different definitions of the same word. They operate at different ontological levels: Harel at the logical level (what state changes follow from what triggers), Gärdenfors at the cognitive-semantic level (what the event means as a structured transformation), Iqbal at the economic level (what demand the event represents), and Burgess at the epistemological level (whose observation constitutes the event).
Can these levels be integrated? Partially, and the gaps are instructive. Iqbal's Event (persistent demand condition) maps naturally to a Harel condition: the patient's shortfall is a condition that guards transitions, and the statechart can model it as a guard on the transition to specialist review. Burgess's perspectival events map to Harel's orthogonality feature: different actors can be modelled as orthogonal regions, each with their own state and event sensitivity.
Gärdenfors's framework offers a bridge between Harel and Iqbal that is suggestive but should be tested. If Iqbal's Events are states maintained by force dynamics - a patient's need persisting because the forces that could resolve it (clinical capacity, decision-making, resource allocation) have not been applied - then the Gärdenfors model explains not just what the demand is but why it persists. The forces are institutional and biological rather than physical, however, and whether the vectorial model genuinely illuminates these or merely relabels them is an open question. The bridge is metaphorical, and its load-bearing capacity remains to be demonstrated.
The force dynamics that Gärdenfors describes - the agent-patient structure, the force vectors and counterforces, the manner and instrument of action - have no natural representation in statecharts. A statechart event REFERRAL_RECEIVED carries none of this semantic structure. It does not encode why the referral was made, what forces produced it, how it was accomplished, or what it means to the entities involved. Talmy's (1988) force dynamics, on which Gärdenfors builds, distinguishes four types of causation - volitional, physical, affective, and inducive - each encoding a different causal structure. Service events involve all four, but statecharts reduce them to a single undifferentiated trigger.
Implications for service representations
This analysis reveals something about the service design representations I critiqued earlier. Journey maps and service blueprints implicitly operate closer to Gärdenfors's event model than to Harel's. When a journey map shows "Patient receives referral letter", it describes a structured event with an agent (the service), a patient (the person), an instrument (the letter), a result (the person knows about the referral), and an experiential quality. The problem is not that journey maps model events wrongly but that they model them implicitly. The agent-patient structure, the force dynamics, the counterforces, the manner - all are present in the designer's mind but none is formally specified. The map is a construal in Gärdenfors's (2017) terms: a mental model of an event with a particular focus of attention selected from it, leaving most of the event's structure unstated.
Statecharts go in the opposite direction: they make event logic explicit but strip event semantics entirely. What is missing is a representation that combines the logical precision of statecharts with the semantic richness of Gärdenfors's event model and the perspectival awareness of Burgess's promise theory.
This has three implications for the broader service grammar project. First, the grammar needs to operate at multiple ontological levels simultaneously. States and transitions are Harel-level (logical); promises are Iqbal/Burgess-level (economic/epistemic); the Gärdenfors level (cognitive/semantic) is largely absent from current service design tools.
Second, force dynamics may be what is missing from "experience" modelling. Service design talks extensively about experience but has weak tools for representing it. Gärdenfors's force vector model - where events have agents, patients, instruments, counterforces, manner, and goals - may provide the cognitive science grounding that experience modelling lacks. The anxiety of waiting is not captured by a state label; it is a force-dynamic situation where the patient's counterforces (desire for resolution, fear of bad news) interact with the system's forces (processing constraints, triage logic).
Third, and this connects to an argument I made about design principles, principles may fail partly because they operate at the wrong ontological level. Principles typically operate at the semantic level - "services should be clear, empathetic, efficient" - while the systems they seek to influence operate at the logical level (Harel) and the economic level (Iqbal). A principle about clarity cannot change a state transition; a principle about empathy cannot restructure a promise network. The ontological mismatch may help explain why principles so often fail to grip.
The tension between these four event ontologies is not resolved here. But naming it makes it workable, and the aspectual classification offers a concrete next step: attending to whether a service event is a state, an accomplishment, an achievement, an activity, or a semelfactive is something designers can do now, with existing tools, and it would already enrich the way service processes are specified. The three frames post that follows in this series extends this analysis from events to situations, examining how Fillmore's linguistic frames, Goffman's situational frames, and Dorst's design frames each construe the structured backgrounds against which events are understood.
References
Burgess, M. (2020). A Treatise on Systems, Volume 2: Intentional Systems and Interaction. ChiTek-i.
Croft, W. (2012). Verbs: Aspect and Causal Structure. Oxford University Press.
Fillmore, C.J. (1968). The case for case. In Bach, E. and Harms, R. (eds.) Universals in Linguistic Theory. Holt, Rinehart and Winston.
Gärdenfors, P. (2017). The Geometry of Meaning: Semantics Based on Conceptual Spaces. MIT Press.
Harel, D. (1987). Statecharts: A visual formalism for complex systems. Science of Computer Programming, 8(3), 231-274.
Harel, D. and Politi, M. (1998). Modeling Reactive Systems with Statecharts: The Statemate Approach. McGraw-Hill.
Iqbal, M. (2018). Thinking in Services: Encoding and Expressing Strategy through Design. BIS Publishers.
Talmy, L. (1988). Force dynamics in language and cognition. Cognitive Science, 12(1), 49-100.
Vendler, Z. (1957). Verbs and times. The Philosophical Review, 66(2), 143-160.