Task, World, Behavior: Three Lenses for Ontology Modeling Clarity
This article introduces a three-perspective framework—Task, World, and Behavior—drawn from Mission Engineering to verify ontology models, using equipment deployment and public opinion research examples to show how to align requirements, reality, and execution while distinguishing plans from actual events.
Introduction: The Gap Between Ontology and Scheduling
An enterprise may have a well-defined equipment ontology covering classification, department ownership, and maintenance records. Yet a scheduler still asks: "Can this job be arranged for tomorrow morning?" The ontology answers equipment type and maintenance status but not whether the job's required capabilities can be met, whether equipment can arrive on time, who arranges transport, or who receives it. This article proposes three perspectives from Mission Engineering—Task, World, and Behavior—to check ontology content for completeness and coherence.
Three Perspectives: What Each Asks
Task Model : What must be achieved? What are the goals, conditions, tasks, and dependencies that define completion?
World Model : What do we know about relevant objects, capabilities, states, relationships, mechanisms, and constraints? What is verified, recorded, or assumed?
Behavior Model : Who does what, how are handoffs performed, and what happens when things go wrong?
These perspectives help organize and cross-check model content; they do not replace the construction processes described in earlier articles, nor must every project produce three separate diagrams upfront.
Task Model: Defining the Goal
A request like "Find me a crane" is insufficient for scheduling. The task model captures location, required capability, deadline, and acceptance confirmation. It structures goals, conditions, tasks, and their dependencies so that changing equipment or transport plans does not require rewriting the business goal.
For example, the goal may be "Have qualified equipment arrive at the designated site before the agreed time and complete acceptance." The model then breaks down verification, resource allocation, and handoff tasks with their dependencies. The system's responsibility may be limited to filtering by type and maintenance status; arrival and acceptance belong to other stages. The model can depict the end-to-end goal while the current delivery scope remains bounded.
This differs from the historical archive query in the previous article, where real-time filtering and arrival scheduling were not required. Only when the business explicitly demands them do we analyze the needed conditions and responsibilities.
World Model: Describing Reality with Evidence
The world model expresses the objects, capabilities, states, relationships, mechanisms, and constraints relevant to the current problem. For instance, "idle" only means not occupied; whether equipment has the needed capability, whether site conditions allow it, and how long transport takes each have separate evidence. Position records carry timestamps—yesterday's location cannot be used for today's arrival judgment.
The model must also show how these conditions affect planning: a changed entry window may shift handoff time; the same person might be needed in two places. Each piece of information is tagged with its source, applicable time, and whether it is verified, recorded, or assumed. The world model covers only what the current work needs to know; it does not require a continuously synchronized digital twin. Even with a clear task, reality may remain unverified; sufficiency depends on the specific question.
Behavior Model: Execution, Handoffs, and Contingencies
Even when reality permits, work does not execute itself. The behavior model specifies who verifies equipment status, who arranges transport, who confirms allocation, who receives. It details inputs, tools, resulting state changes, and records for each step. It also addresses failures: what if status is unknown, resource is occupied, or reception fails?
Example conflict: both status verification and site handover require the same person on-site at overlapping times. Merely writing "responsible person handles" in two process flows does not resolve the conflict. The model must show the clash and force a decision: stagger times, change personnel, or adjust the plan.
Results handoff must be explicit: if status cannot be verified, do we wait, mark "pending verification", and assign someone? Without modeling the off-nominal paths, these questions are left to ad-hoc coordination.
Connecting the Three Perspectives via a Single Task
Assume the business requires equipment on-site before a deadline with acceptance, while the system only handles type and maintenance filtering. Design can then cross-check the three perspectives together before equipment arrives late.
First align the job, equipment, and deadline—don't check last week's plan against this week's requirement. Then examine each perspective's content and cross-checks:
Task Model : Site, deadline, acceptance criteria; system's responsibility boundary. Cross-check: Is the deadline reflected in the transport plan? Are acceptance criteria verified by someone?
World Model : Equipment identity, capability, state, location, entry window; sources, timestamps, unknowns. Cross-check: Do all references point to the same equipment and task? Does evidence support the conditions assumed in the plan?
Behavior Model : Who verifies, arranges transport, receives; required inputs; delay handling. Cross-check: Are there time or resource conflicts? Are results passed to the next step?
Example gaps: deadline missing from transport plan; plan assumes equipment nearby but no location record exists; no receiver assigned on-site. Each gap becomes a concrete to-do: what is missing, who adds it, how to verify. Unverified conditions must not be drawn as satisfied.
If equipment arrives late, trace the record to see what actually happened. However, a clear model does not automatically diagnose cause; a late arrival does not prove the filtering assistant erred, and the three perspectives alone cannot assign blame.
Distinguishing Requirements, Plans, Events, and Records
A common confusion is treating preparation as completion. Four distinct concepts must be separated:
Requirement : "Equipment must arrive before agreed time and be accepted" — a completion criterion, not a result.
Plan : "Transport personnel and reception slots arranged" — an arrangement exists, but feasibility still needs checking.
Execution Event : "Equipment actually arrived, reception staff completed verification" — did it happen, when, with what evidence?
Record : "System saved an arrival or acceptance record" — a record exists, but its object, time, source, and match to the event must be verified.
If the same "completion" field means "plan scheduled" in one place and "site acceptance done" in another, queries will answer incorrectly. Shared meanings must be explicit in the ontology and used consistently by data and applications.
How Analysis Results Enter the Ontology
Not every analysis table row becomes an ontology class. Identify which meanings need shared use across the organization and which are specific to this job.
For example, "acceptance complete" appears in multiple contexts; clarify whether it means plan scheduled, site verification finished, or system record saved. Define these distinctions and related object relationships, then test with concrete tasks and records.
Job goals & acceptance criteria → Shared: Job task, completion criteria, applicable objects. Instance links: this task's location, deadline, criteria version, confirmation basis.
Equipment capabilities, states, constraints → Shared: Equipment, capability conditions, state records, sites, relationships. Instance links: which equipment, state at what time, source, pending assumptions.
Transport and site handover → Shared: Transport activity, reception activity, roles, inputs/outputs, state changes. Instance links: concrete plan, personnel, actual events, records.
This table suggests candidates, not a class creation checklist. Reuse existing definitions; don't add unused ones. The ontology defines shared meanings (e.g., "task can have deadline", "reception activity involves certain roles"). Specifics become instances or link to external data. Routing calculations, scheduling, and process execution remain in programs.
Ontology, OPM, and Execution Systems: Division of Labor
Previous article "Is the job done when coffee is finished?" and the Mission Engineering Guide (MEG 2.0) emphasize end-to-end task effects and iterative analysis. The Mission Architecture Style Guide (MASG) notes a bidirectional relationship between architecture and analysis. The three perspectives here adapt those ideas.
OPM (Object-Process Methodology) can express equipment, personnel, transport processes, and state changes; the ontology clarifies shared concepts, relations, and constraints. They cross-check each other; no fixed conversion order is required, and diagrams do not auto-execute.
The earlier distinction between business ontology (meaning confirmation) and technical ontology (platform expression) is separate. Task, World, and Behavior perspectives may need both business and technical collaboration; do not equate World Model with business ontology and Behavior Model with technical ontology.
Ontology does not cover the whole solution: arrival time calculation needs data and algorithms, dispatch needs an execution system, confirmation needs field feedback. Clear concepts are the foundation for collaboration.
Two Construction Approaches: How Far to Apply Each Perspective
Goal-constrained : Work backward from completion conditions to required reality conditions and necessary activities. Equipment arrival suits this.
Open exploration : No need to fix the final mission upfront. Using the public opinion research example from the previous article:
Task : Tentatively "distinguish direct evaluation from retelling"; define what this round verifies, without judging true psychological stance.
World : Organize content, accounts, expressions, attribution, citation relations, sources; unknown principals remain unknown.
Behavior : If multi-person collaboration, assign who adds context, who verifies original text, who reviews candidate definitions, who resolves disagreements.
Note: the research team's work and online accounts' posting behavior are different layers; both are "behavior" but must be distinguished in diagrams and records.
If solo, the behavior part stays brief; expand when multi-person, resource conflicts, and handoff responsibilities arise. Use existing construction cards or exploration records, not three new archives.
Practical Step and Next Article
Pick a current task and ask: what must I judge now? Is the material enough? Who does the next step? Missing definitions? Discuss. Missing material? Gather. No owner? Assign. Model adequacy and actual completion are separate checks.
The next article will cover handling discovered errors: the same equipment list was wrong—how to decide whether to fix the ontology, the data, or the program.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
