.svg){kind=icon}
Learn
The DIKW pyramid is a hierarchical model that arranges data, information, knowledge, and wisdom as four ascending levels, where each level adds context and meaning to the one below it to support better decisions.
Last updated: October 23, 2026
Why it matters to your mission. Decisions do not run on raw data. They run on wisdom, the top of a hierarchy that climbs from scattered facts through information and knowledge before anyone can act. The DIKW pyramid names that climb. In most organizations the pyramid is not the problem; the speed and reliability of the climb is. Raw data piles up faster than anyone can turn it into information, and information rarely becomes shared knowledge fast enough to matter. The mission advantage goes to whoever ascends the pyramid faster, and more trustworthily, than the adversary.
The DIKW pyramid is a hierarchical model that arranges data, information, knowledge, and wisdom as four ascending levels, where each level adds context and meaning to the one below it to support better decisions. It is one of the most widely used mental models in information science, knowledge management, and data strategy, and it is the standard way practitioners describe how raw facts become the basis for judgment.
The model is most commonly traced to Russell L. Ackoff, whose 1989 paper "From Data to Wisdom" (originating as his 1988 presidential address to the International Society for General Systems Research) linked the terms into a single hierarchy. Ackoff actually proposed five tiers, inserting understanding between knowledge and wisdom. The now-familiar four-level pyramid, with data at the broad base and wisdom at the narrow apex, is a later simplification the information-science literature still debates. That debate matters, and this page treats it honestly below, but the model remains useful as a shared vocabulary for one practical question: how do you turn what you have into what you can act on?
Key takeaways
The DIKW pyramid has four levels, each defined by what it adds to the level below. The table gives the standard definition of each level, the question it answers, and a defense example to make the progression concrete.
| Level | Definition | Question it answers | Defense example |
|---|---|---|---|
| Data | Raw, unorganized symbols and signals with no inherent meaning. Facts and figures as recorded, before interpretation. | None yet | A radar return, a line in a logistics file, a single sensor ping, a name in a document |
| Information | Data that has been organized, contextualized, and given meaning and purpose. | Who, what, where, when | "This track is an aircraft, on this heading, at this time, in this sector" |
| Knowledge | Information connected to other information, experience, and rules so it can be applied. Knowing how. | How | "Aircraft on this profile, correlated with these logistics and signals patterns, indicate a specific activity we have seen before" |
| Wisdom | Applied, evaluated knowledge used to judge what to do, including what is right and what matters. Knowing why, and what to do. | Why, and what should we do | "Given the pattern, the rules of engagement, and the mission, here is the recommended course of action and its risk" |
Ackoff himself mapped the levels to question-words in this way: information answers who, what, where, and when; knowledge answers how; understanding answers why; and wisdom is evaluated understanding. The defense example shows why the hierarchy is a useful frame. A single radar return is data. It becomes information once resolved into a meaningful track, knowledge once correlated with other sources into a recognized pattern, and wisdom only when a commander can weigh that pattern against mission, rules, and risk to decide. Every intelligence and command workflow is, in effect, a machine for climbing this pyramid.
The value is in the climb, not the tiers
The tiers are easy to name. The value is in the climb: data to information, and information to knowledge, is where most organizations stall.
Machines are strong at the lower climbs, correlating sources into knowledge at scale; wisdom, judging what to do and why, stays a human responsibility. The model is commonly credited to Russell Ackoff (1989), and scholars contest its strict logic.
The difference between data and information is context. Data is raw and unorganized; it is the symbol or signal before anyone has interpreted it. Information is data that has been structured, related, and given meaning and purpose so it answers a specific question. In the standard framing, information is "data that are endowed with meaning and purpose." The number 37 is data. "Runway 37 is closed" is information.
This first transition is where most data strategies succeed or fail, because the raw material is overwhelmingly unstructured and multi-source: documents, messages, sensor feeds, and records that arrive in incompatible formats with no shared meaning. Turning that into information at machine speed is the whole challenge of structured versus unstructured data for AI. An organization that cannot reliably perform the data-to-information step never gets to use the top half of the pyramid, because knowledge and wisdom have nothing dependable to build on. This is why AI-ready data is a precondition rather than a nice-to-have: an AI system reasoning over un-contextualized data is reasoning over the pyramid's base with none of the structure the higher levels require.
Knowledge is information connected and applied: it is knowing how, the result of relating pieces of information to each other, to experience, and to rules until you can act on them. Wisdom is the layer above: it is evaluated, judgment-laden application of knowledge, knowing not just how but why, and what is the right thing to do. In Ackoff's framing wisdom "is the ability to increase effectiveness" and "requires a judgment function." Knowledge tells you what the pattern is and how it behaves. Wisdom decides what to do about it given values, consequences, and mission.
For AI and defense this distinction defines the division of labor between machine and human. Machines are now very good at the lower transitions, moving data to information and information to knowledge, by correlating sources and surfacing patterns at a scale no analyst can match. Wisdom, the evaluative judgment at the apex, is where human command responsibility sits and should stay. The right design goal for decision advantage is therefore not to automate the top of the pyramid but to deliver trustworthy, explainable knowledge to the human fast enough that their wisdom has time to operate. The system climbs the pyramid; the human owns the summit.
The DIKW model is most commonly credited to Russell L. Ackoff, who linked data, information, knowledge, understanding, and wisdom into a single hierarchy in his 1989 paper "From Data to Wisdom," published in the Journal of Applied Systems Analysis and based on his 1988 presidential address to the International Society for General Systems Research. Ackoff is generally recognized as the first to put all the terms into one formula, though the concepts have older roots.
The lineage is worth stating accurately, because the model is often presented as if it sprang fully formed from one source. It did not. Milan Zeleny described a related progression (know-nothing, know-what, know-how, know-why) in 1987, and Mike Cooley invoked a similar hierarchy in 1980. The four-level pyramid visual, with data at the base and wisdom at the apex, came from later authors rather than from Ackoff or Zeleny. DIKW is best understood as a widely shared convention with a traceable origin in Ackoff's 1989 formulation and several earlier antecedents, not a single discovery with a single date.
The DIKW pyramid is useful as a shared vocabulary, but it is genuinely contested in the scholarly literature, and a responsible treatment should say so rather than present it as settled fact.
Jennifer Rowley's 2007 paper "The wisdom hierarchy: representations of the DIKW hierarchy" reviewed how major textbooks articulate the model and found a telling asymmetry: there is reasonable consensus on how to define each level, but "less consensus in the description of the processes that transform elements lower in the hierarchy into those above." The field agrees on the boxes and disagrees on the arrows, which are the part that actually matters operationally. Martin Frické's 2009 paper "The knowledge pyramid: a critique of the DIKW hierarchy" went further, arguing the hierarchy rests on "dated and unsatisfactory philosophical positions" and is, as a strict logical model, unsound.
The practical conclusion is not to discard the pyramid but to use it for what it is good at. As a map of how raw material becomes decision-ready understanding, and as a shared language across data, intelligence, and command, it is valuable. As a literal claim that there is a clean, automatic escalator from data to wisdom, it is too neat. The critiques reinforce the point this page has made throughout: the transitions, not the tiers, are where the difficulty and the value live. That is why treating data-readiness as an engineering problem matters more than arguing about the diagram.
The DIKW pyramid matters for AI and defense because modern missions fail at the lower transitions, not the upper ones. Fragmented, multi-source data arrives faster than any organization can turn it into information, and information rarely becomes shared knowledge quickly enough to support a timely decision. The pyramid makes the bottleneck visible: most of the raw material sits at the base, the climb stalls, and wisdom, the only level a commander can act on, stays out of reach.
Two properties make this an infrastructure problem rather than an analyst problem. First, the base of the pyramid is now enormous and overwhelmingly unstructured: sensor feeds, documents, logs, and records accumulate in incompatible formats across systems never designed to talk to each other, so climbing from data to information means giving all of it consistent, machine-usable meaning at scale, the core of data-readiness. Second, the knowledge level depends on correlation across sources: a pattern that constitutes knowledge usually exists only in the combination of records that individually say very little, so moving from information to knowledge requires resolving entities and relationships across the whole environment, not one feed at a time.
This is why climbing the pyramid fast and reliably is a reasoning-infrastructure capability, not a single tool. It depends on representing unstructured data so a system can work with its meaning, the function of knowledge domains, and on connecting resolved facts into the relationships that make information into knowledge. The payoff is decision advantage: reaching trustworthy, explainable knowledge fast enough that human wisdom has time to act before the situation changes.
Torch.AI does not sell a "pyramid," and this page is not a claim to own a model that predates the company by decades. What Torch.AI builds is the government-owned reasoning infrastructure that makes the lower two climbs, data-to-information and information-to-knowledge, fast and reliable on fragmented, multi-source data, so that human wisdom at the top has something trustworthy to act on.
The mapping to the pyramid is direct. ORCUS provides the data-availability layer that connects to existing sources and preserves provenance, so the raw material at the base stays traceable rather than pooled into an opaque mass. NEXUS applies semantic representation and knowledge domains so unstructured data gains the context that turns it into information a system can work with: the data-to-information climb. HALO fuses resolved entities and relationships into a graph, surfacing the patterns that only exist across sources: the information-to-knowledge climb. Because this runs as a system of reason on top of existing systems of record, the authoritative sources stay intact, every element remains traceable, and in a government-owned deployment the customer keeps control of the layer where data becomes knowledge. The apex stays where it belongs: wisdom and the decision remain a human command responsibility, now reachable in time. You can see how this is packaged on the Torch.AI software page.
For evaluators, the fit is specific: if your problem is that you have abundant data and too little usable knowledge, that the climb up the pyramid stalls because your sources are fragmented and un-contextualized, that is a data and reasoning problem, and it is the one this infrastructure is built for. Request a technical walkthrough and we will run it against a representative slice of your own multi-source data, with provenance traced end to end.
The four levels, from bottom to top, are data (raw, unorganized facts and signals), information (data organized and given context and meaning), knowledge (information connected and applied so it can be used), and wisdom (evaluated knowledge used to judge what to do and why). Each level is built from the one below it and adds context, connection, or judgment.
Data is raw and unorganized, a symbol or signal before anyone has interpreted it. Information is data that has been structured and given context and meaning so it answers a specific question such as who, what, where, or when. The data-to-information step is where most data strategies succeed or fail.
Knowledge is information connected and applied, knowing how something works and how to act on it. Wisdom is the layer above: evaluated, judgment-laden application of knowledge, knowing why and what is the right thing to do. In AI and defense, machines are strong at producing knowledge by correlating sources, while wisdom remains a human responsibility.
The DIKW model is most commonly credited to Russell L. Ackoff, who linked the terms into one hierarchy in his 1989 paper 'From Data to Wisdom', based on his 1988 presidential address to the International Society for General Systems Research. Earlier antecedents include Milan Zeleny (1987) and Mike Cooley (1980), and the four-level pyramid diagram itself came from later authors.
It remains a useful shared vocabulary for how raw data becomes decision-ready understanding, but it is contested in information science. Jennifer Rowley (2007) showed there is far less agreement on the transitions between levels than on the levels themselves, and Martin Fricke (2009) argued the hierarchy is logically unsound as a strict model. The transitions, not the tiers, are where the real difficulty and value lie.
Because modern missions fail at the lower transitions. Fragmented, multi-source data accumulates faster than it can be turned into information, and information rarely becomes shared knowledge fast enough to support a timely decision. Climbing the pyramid quickly and reliably, from un-contextualized data to trustworthy knowledge, is a reasoning-infrastructure problem and the foundation of decision advantage.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
Describe where your data stalls before it becomes knowledge. We will scope a session around your fragmented, multi-source data. No diagram debate. Your mission data. Real work.
