The problem archetypes solve

Ask an analyst to imagine the future of a technology or a region and the mind almost always reaches for one story, usually a continuation of the present. That single story feels like analysis but it is really a projection of the analyst's current beliefs. Scenario archetypes exist to break this habit. Instead of asking what will happen, they ask the analyst to populate a fixed set of qualitatively different futures, so that outcomes the mind prefers to skip get written down and taken seriously.

Dator's four generic futures

Jim Dator, who founded the Hawaii Research Center for Futures Studies, argued from decades of surveying futures work that almost every image of the future collapses into one of four kinds. Growth is the official future of most institutions: more of the present, bigger, richer, more connected. Collapse is the failure of the system through economic, ecological or social breakdown. Discipline is a future organised around a constraint, where society reins itself in to live within limits, whether ecological, moral or resource-based. Transformation is discontinuity, usually technological or spiritual, in which the rules of the game themselves change and the old measures stop applying. Dator's claim is not that these are equally likely but that a serious foresight exercise should hold all four in view at once.

Why a fixed set reduces bias

The power of the archetypes is that they are chosen for you. A blank-page brainstorm tends to cluster around a single mood, optimistic in a boom, fearful in a crisis. By contrast, a fixed set of four forces the analyst to write a credible growth story even when they feel gloomy, and a credible collapse story even when they feel confident. This is the same logic that structured analytic techniques bring to intelligence work: the method compensates for the known tilt of the individual mind. The archetypes are a bias-reduction device before they are anything else.

Archetypes in DSI's own work

DSI's briefing The Architecture of the Watched World closed not with a single prediction but with five explicit forecasts about the trajectory of the surveillance state. Read against Dator, that set is doing archetype work: some forecasts describe growth of the watched world, one describes a disciplined settlement in which societies accept limits on surveillance, and others gesture at transformation as new technologies rewrite the terrain. Presenting a spread rather than a point is the archetype instinct in action, even when the four names are not used explicitly.

Where the generic set runs out

Archetypes are a strong starting frame, not a finished product. Because they are generic, they say nothing about the specific driving forces of your particular question, and four boxes can flatten a rich problem into caricature. The honest use is as scaffolding: begin with the four to guarantee coverage, then move to a method that engages the actual forces at work. The next lesson does exactly that, replacing generic futures with quadrants built from the specific uncertainties that matter most.

Two axes, four futures

The deductive two-by-two is the workhorse of practical scenario planning. The analyst identifies the driving forces bearing on a question, selects the two that matter most and are least predictable, and draws them as crossed axes. The four quadrants that result are four internally consistent futures, each defined by a different combination of how the two forces resolve. Because the structure is deduced from the axes rather than dreamt up, the four scenarios are guaranteed to differ from one another in a disciplined way.

The Global Business Network method

This approach is associated above all with the Global Business Network, the consultancy founded in 1987 by Peter Schwartz, Stewart Brand, Napier Collyns and others, drawing on the scenario tradition Schwartz and Pierre Wack had developed at Royal Dutch Shell in the 1970s. Schwartz set it out in The Art of the Long View. The GBN process runs from a focal decision, to the driving forces, to ranking those forces by importance and uncertainty, to choosing the two critical uncertainties as axes, and finally to fleshing out and naming each quadrant. Its lasting appeal is that it produces a small, memorable set of futures a busy decision-maker can actually hold in mind.

Why importance and uncertainty both matter

The whole method lives or dies on the choice of axes, and the two tests are non-negotiable. A force that is important but near-certain, an ageing population in much of Europe, for instance, belongs in every scenario as a fixed assumption, not on an axis, because it does not distinguish the futures from one another. A force that is uncertain but unimportant generates four scenarios that no decision-maker cares about. Only a force that is both, high impact and genuinely open, does useful work as an axis. Getting this wrong is the single most common way the method fails.

Choosing axes well

Good axes are also independent of each other, so that all four quadrants are plausible rather than two being self-contradictory. Suppose a European energy strategist asks about the 2035 grid. Candidate axes might be the pace of electricity demand and the degree of European supply-chain autonomy in critical minerals, forces from the same weaponised-interdependence terrain DSI writes about. Both are important, both are uncertain, and they are largely independent, so the four quadrants, from an autonomous high-demand world to a dependent low-demand one, are each coherent and worth planning against.

Failure modes to guard against

Three failures recur. Axes that secretly measure the same thing collapse four scenarios into two. Axes chosen for drama rather than relevance produce vivid but useless quadrants. And analysts often smuggle their preferred future into one quadrant and treat the other three as strawmen. The corrective is DSI's report standard of probability-banded scenarios, modelled on J.P. Morgan's Decision-Making under Deep Uncertainty: each of the four futures is presented with an honest confidence range rather than one being quietly favoured. The discipline of banding keeps the set balanced.

Building up instead of down

The deductive method starts from a frame, the two axes, and fills in the futures they imply. The inductive method reverses the direction. It starts from the ground, with a large collection of signals, weak indications of change picked up from research, data and expert conversation, and lets the scenarios emerge from how those signals hang together. Nothing is imposed first. The analyst's task is to notice which signals reinforce one another and cluster into a coherent picture, then to name and develop each cluster as a distinct future.

Clustering signals into stories

The craft of the inductive method is disciplined clustering. Faced with dozens of signals about, say, the future of a supply chain, the analyst groups those that point the same way, a reshoring incentive here, a tariff there, a new refining capacity elsewhere, into a story of regionalisation, while a different cluster forms a story of deepening interdependence. The strength of the approach is that it can surface futures no pre-chosen axis would have generated, because it is led by what the evidence is actually showing rather than by the analyst's initial hunch about which two forces matter. Its weakness is that clustering is a matter of judgement and can drift toward the stories the analyst already favours, so the signal set must be broad and honestly assembled.

Branching-point scenarios

A distinct member of the family is the branching scenario, which is appropriate when a single decision or event dominates the future. Here the analyst identifies a fork, an election, a court ruling, a decision to invade or not, and builds the tree of futures that flows from each branch. The scenarios share a common trunk up to the branching point and then diverge sharply. This structure is honest about where the real uncertainty lives: it concentrates attention on the moment that matters rather than spreading it evenly. It suits geopolitical questions where one actor's choice reshapes everything downstream.

Incremental scenarios and choosing a direction

A lighter relative is the incremental scenario, in which the analyst takes a baseline projection and varies a single factor at a time, higher demand, slower adoption, a delayed regulation, to see how sensitive the outcome is to each. It is less about rich alternative worlds and more about stress-testing one expectation. Choosing among the building directions is itself an analytic decision. Use the deductive two-by-two when two forces clearly dominate; use inductive clustering when the field is open and signal-rich; use branching when one pivotal choice looms; use incremental variation when you mainly need to test the robustness of a single view. The methods are complementary tools, not rivals.

A different question

The archetype, two-by-two and inductive methods are all exploratory: they map the space of what could happen and stay neutral about which future is desirable. Normative foresight reverses the stance. It begins by asking what future we would prefer, then treats that preferred future as the fixed point and reasons backwards. The shift is from describing possibility to designing a path, and it changes the analyst's role from observer to a partner in steering. This is the visioning tradition, and its central technique is backcasting.

What backcasting is

BackcastingWorking backwards from a preferred future end state to the steps, decisions and milestones that would connect the present to it. Developed by John Robinson as a deliberate inversion of forecasting. was named and developed by the Canadian scholar John Robinson in the early 1980s, originally in energy-policy work, as a deliberate inversion of forecasting. A forecast asks where present trends will carry us. Backcasting asks: if a particular desirable end state were achieved by some target year, what sequence of events, decisions and milestones would have had to occur to get there? The analyst fixes the destination first, then works backwards along the timeline, identifying the steps immediately before the goal, then the steps before those, until the chain reaches the present. The output is not a prediction but a route.

When steering beats predicting

Backcasting earns its place when the future in question is not something to be passively awaited but something an actor intends to shape, and when present trends point somewhere the actor refuses to accept. A city aiming to be carbon-neutral by a fixed year gains little from a forecast that simply extrapolates current emissions; it needs the chain of decisions that would bend the curve to zero. The same logic serves a firm setting a long-horizon strategic goal, or a government defining a security posture it means to reach. Where forecasting is fatalistic, describing the future as something that happens to you, backcasting is agentive, describing it as something you build.

The risks of starting from the answer

The very feature that makes backcasting powerful, fixing the destination first, is also its danger. Once a preferred future is chosen, the mind is prone to motivated reasoning: the analyst constructs a comfortable path and quietly discounts the obstacles that make it implausible. A backcast can become a wish dressed as a plan. Two disciplines guard against this. First, the preferred future must be genuinely feasible, tested against the same driving forces the exploratory methods surface, not merely attractive. Second, the backcast should be paired with an exploratory set, so the route to the desired future is stress-tested against the collapse and discipline archetypes and against the two-by-two quadrants where the plan would fail. Held honestly, normative foresight does not replace exploration; it completes it by adding the question of where we actually want to go.

Events are not independent

Most of the methods so far quietly assume that driving forces act on their own. Reality is entangled: a supply shock raises the odds of a policy response, which changes the odds of a technological shift, which alters the shock. Cross-impact analysisA method, devised by Gordon and Helmer, that estimates how the occurrence of each future event would raise or lower the probability of the others, capturing their interdependence in a matrix. was developed in the 1960s by Theodore Gordon and Olaf Helmer, who had earlier built the Delphi method at RAND, precisely to capture this entanglement. The analyst lists a set of possible future events, estimates the standalone probability of each, and then fills in a matrix of how the occurrence of each event would raise or lower the probability of every other. The scenarios that survive are the combinations of events that remain internally consistent once these interactions are accounted for.

What the cross-impact matrix reveals

The value of the exercise is less the precise numbers, which are estimates, than the structure it forces into view. Building the matrix compels the analyst to ask, for every pair of events, does this one make that one more or less likely, and by how much. The answers often overturn intuition: an event judged likely in isolation may become improbable once the events that would suppress it are considered, and a marginal event may become near-certain once its enablers are in place. The method turns a list of independent guesses into a web of conditional relationships, which is a far more honest picture of how a future actually assembles itself.

Morphology: futures as combinations

Morphological analysisFritz Zwicky's method of decomposing a question into parameters, enumerating each parameter's possible states, and combining them under a consistency filter that discards incompatible pairs, leaving coherent scenarios., devised by the astrophysicist Fritz Zwicky at Caltech in the mid-twentieth century, approaches the problem from the component side. The analyst breaks a question into its key parameters, then lists the possible states each parameter could take, demand high, medium or low; supply autonomous or dependent; regulation strict or loose. Every combination of one state per parameter is a candidate scenario. Because the number of combinations explodes quickly, the essential move is the consistency filter: the analyst marks which pairs of states are mutually incompatible and discards every combination containing an impossible pair. What remains is a set of internally coherent scenarios assembled systematically rather than intuitively.

Intuitive versus formal

Set against the two-by-two, these methods trade memorability for rigour. The two-by-two gives four vivid futures a decision-maker can hold in mind but reduces the whole question to two forces. Cross-impact and morphology can handle many events and many components, and they expose interactions the two-by-two hides, at the cost of complexity and a heavier apparatus. Neither is simply better; the choice depends on whether the priority is communication or completeness. A mature foresight practice uses the two-by-two to communicate and the formal methods to check that the intuitive set has not missed a coherent combination.

The bridge to the French school

These formal, combinatorial methods lead directly into the French tradition of la prospective, developed by Gaston Berger and later systematised by Michel Godet. Godet's toolkit, with instruments such as MICMAC for analysing the influence among variables and Morphol for structured morphological work, is the formal end of the discipline this module has traversed. The path runs from generic archetypes, through the intuitive two-by-two, to inductive and normative building, and finally to the structured combinatorial methods that treat scenario-building as a rigorous, auditable procedure. Knowing where a given exercise sits on that spectrum, from quick and communicable to slow and exhaustive, is itself a mark of the mature analyst.