An intention is a directed mental or declared state in which an actor means to pursue, perform, avoid or bring about something in the future.
Intentions sit between possibility and action. A person can want something without intending to pursue it, intend to act without having acted yet, and begin acting while later abandoning the original intention. Organisations, teams and systems can also be described as having intentions when authorised plans, commitments or control policies direct future behaviour, but those institutional intentions should remain traceable to actual decision structures rather than treated as mysterious collective minds.
Good classification therefore separates intention from desire, preference, goal, commitment, plan and action. It also distinguishes what an actor explicitly states from what an observer infers from behaviour.
Quick answer: how should intentions be categorised?
- Target: what outcome, action or avoided state is intended?
- Source: explicitly declared, documented, authorised or inferred?
- Specificity: vague direction, defined goal, concrete action or implementation intention?
- Commitment strength: tentative, preferred, decided, committed or binding?
- Time horizon: immediate, near-term, long-term or open-ended?
- Conditions: unconditional or dependent on triggers, resources, approval or state?
- Feasibility: possible, constrained, blocked or impossible under current conditions?
- Conflict: compatible with or competing against other intentions?
- Evidence: statement, plan, behaviour, resource allocation, repeated action or institutional record?
- Execution: not started, active, paused, completed, abandoned or superseded?
- Revision: what would cause the intention to change?
This article complements How to Categorise Goals, How to Categorise Actions, How to Categorise Commitments and How to Categorise Preferences. A goal describes a desired end state; an intention is the actor’s directed stance toward pursuing or avoiding one.
Intention is not desire
A person may desire an outcome without intending to pursue it. Someone can want to travel, become fitter or learn another language while having made no decision to act. Desire concerns attraction; intention concerns directed future action or state.
Intention is not preference
A preference ranks alternatives. An intention selects or adopts a course. A student may prefer studying in the morning but intend to study tonight because of a deadline.
Intention is not goal
A goal can exist as an external target: score 75, finish the report, reduce latency. An intention records that an actor means to pursue that goal. Systems should not infer intention merely because a goal is available.
Intention is not commitment
A commitment creates stronger expectations toward oneself or others. An intention may remain revisable and private. Once an actor promises, signs, allocates resources or creates reliance, the intention may become a commitment with additional obligations.
Intention is not plan
A plan specifies route, sequence, resources or timing. An intention can exist before a plan. “I intend to improve my mathematics” is an intention; a weekly practice schedule and diagnostic sequence form part of the plan.
Intention is not action
Actions are observable events. Intentions concern what an actor means to do. A stated intention can fail to become action; an action can occur accidentally or under compulsion without matching the actor’s intention.
Declared intentions are directly stated
“I will revise tonight”, “we intend to launch in October” and “the committee intends to review the policy” provide explicit evidence. Direct statement reduces uncertainty about the declared state but not about sincerity, feasibility or eventual execution.
Documented intentions leave durable records
Approved plans, meeting minutes, roadmaps and formal resolutions can preserve institutional intention. The document should identify who had authority to establish that intention and under which version.
Inferred intentions are interpretations
Observers infer intention from behaviour: repeated visits may look purposeful, resource allocation may signal strategic intent, and a student choosing a method may appear to intend a certain solution path. These inferences can be useful but should remain marked as interpretations rather than direct facts about internal states.
Behavioural evidence can underdetermine intention
The same action can result from habit, accident, coercion, experiment, imitation or multiple possible goals. Intention inference becomes stronger when several independent behaviours align and alternative explanations weaken.
Immediate intentions target the next action
“I intend to answer this question now” is close to execution. Immediate intentions have shorter gaps in which priorities, conditions or preferences can change.
Proximal intentions target the near future
A learner may intend to finish revision this weekend. A team may intend to deploy a feature next sprint. These intentions can support planning because timing is bounded but still leave room for intervening constraints.
Distal intentions target longer horizons
Career, institutional and strategic intentions may span years. Their content is often less specific and more vulnerable to environmental change, making versioning and intermediate commitments important.
Open-ended intentions lack fixed timing
“Someday I intend to write a book” expresses direction without a scheduled execution state. Open-ended intentions should not be treated as equivalent to scheduled commitments.
Abstract intentions describe direction
“Become healthier”, “improve quality” or “strengthen research” expresses a directional intention but leaves action selection unresolved.
Specific intentions identify an action or outcome
“Complete three practice papers before Friday” narrows the target, time and success condition. Specificity makes execution and later verification easier.
Implementation intentions link conditions to actions
“If I reach home by 7 pm, I will revise algebra for 30 minutes” connects a trigger to a response. This structure reduces the need to make the same decision repeatedly at execution time.
Conditional intentions depend on future states
“We intend to proceed if approval is granted” is weaker than an unconditional intention. The activation condition should be explicit so readers do not mistake contingency for commitment.
Contingent intentions preserve optionality
An actor may intentionally keep several paths open until uncertainty resolves. This is not indecision if preserving options is itself a deliberate strategy.
Positive intentions seek action or state
Learn, build, visit, complete, expand and communicate are positive directions toward bringing something about.
Negative intentions seek restraint or avoidance
“I intend not to disclose the information” or “the system should avoid exceeding this limit” aims at preventing an action or state. Negative intention should not be represented as absence of intention.
Tentative intentions remain weakly committed
“I am thinking of applying” or “we may expand” can signal direction without a settled decision. Classification should preserve tentative status instead of upgrading it into a plan.
Settled intentions follow decision
Once alternatives have been considered and one path chosen, the intention becomes more stable. It may still change if assumptions fail or constraints shift.
Committed intentions create stronger expectations
Allocating money, booking resources, publishing a roadmap or making a promise raises the cost of reversal. At this point intention begins to overlap with commitment and obligation, and those separate objects should be linked.
Intentions can conflict internally
A person can intend to save money and travel, study and rest, grow a business and preserve free time. Conflicting intentions compete for resources and can explain inconsistent action without requiring irrationality.
Intentions can be hierarchically related
A higher-level intention to improve mathematics may generate lower-level intentions to practise algebra, review errors and seek feedback. Hierarchy connects direction to executable action.
Means intentions serve end intentions
An actor may intend to wake earlier only because it enables study before school. Distinguishing means from ends helps explain why one intention disappears when a better route becomes available.
Intrinsic intentions concern outcomes valued in themselves
Someone may intend to read because reading is enjoyable, not because it serves another goal. Intrinsic and instrumental intentions can coexist.
Feasibility constrains intention quality
An intention can be sincere yet infeasible because of time, resources, capability, law or dependencies. Feasibility should be classified separately from sincerity or commitment strength.
Blocked intentions remain intentions
A learner may genuinely intend to practise but lack materials or time. A team may intend to deploy but be blocked by an approval. Failure to act does not by itself prove absence of intention.
Intentions compete with constraints
Hard constraints can prevent execution entirely; soft constraints can reduce probability or increase cost. Linking intentions to How to Categorise Constraints makes blocked pathways visible.
Intentions compete with priorities
A person can hold many genuine intentions but only act on the highest-priority ones under limited capacity. Intention therefore should not be inferred from priority alone, and priority should not be inferred from intention alone.
Resources reveal seriousness but do not prove intent
Budget, time and staffing allocation can support an inference that an intention is serious. Yet resources may be allocated for option preservation, compliance or experimentation rather than settled commitment.
Repeated behaviour strengthens inferred intention
Consistent action over time can make one interpretation of intent more plausible, particularly when alternative explanations weaken. Still, observed behaviour should not erase the distinction between evidence and internal state.
Intention–action gaps are informative
When stated intentions repeatedly fail to become action, investigate friction: weak commitment, poor planning, conflicting priorities, missing cues, low capability, environmental barriers, changing preferences or unrealistic scope.
The first weak link may sit between intention and execution
A learner may fully intend to study but lack a start cue, diagnostic target or manageable task. Better intervention repairs the first missing operational link instead of judging the intention as insincere.
Intentions can be abandoned
An intention may cease because the goal loses value, evidence changes, costs rise, a better alternative appears or the actor deliberately reprioritises. Abandonment should have a reason when the history matters.
Intentions can be superseded
A newer intention can replace an older one without implying that the earlier state was false. Preserve version and effective time rather than overwriting history.
Intentions can be deferred
Deferral changes timing while preserving direction. A deferred intention should remain distinct from abandonment and from active execution.
Intentions can be fulfilled
Fulfilment occurs when the intended action or state is achieved. But completion of one action does not necessarily fulfil a higher-level intention if the broader target remains unmet.
Outcome does not prove intention retrospectively
If an outcome benefits someone, we cannot infer they intended it. Accidents, external forces and unintended consequences can produce the same result. Intention attribution should rely on evidence available before or during action, not outcome alone.
Bad outcomes do not prove bad intentions
Good intentions can produce harmful consequences through poor information, weak methods or unforeseen effects. Evaluation should separate intention, action quality and outcome.
Good outcomes do not excuse harmful intent
An action undertaken for a harmful purpose can accidentally produce a beneficial outcome. Classification prevents moral and causal assessment from collapsing into one result.
Institutional intentions require authorised evidence
“The company intends” should point to board decisions, executive strategy, approved plans, budgets, public commitments or other authorised signals. Statements by unauthorised individuals should not automatically be treated as institutional intent.
Collective intentions require coordination
A team can act as though it shares an intention when members understand a common goal, expect coordinated action and accept role contributions. Mere similarity of individual desires is weaker than coordinated collective intent.
Policy intentions should be separated from implemented effects
A policy can intend to improve access, safety or fairness while producing mixed or opposite outcomes. Policy evaluation should preserve stated intent, mechanism, implementation and observed consequence as separate layers.
Legal intention can have domain-specific tests
Law may infer intention using evidence and standards that differ by jurisdiction and issue. A general classification system should therefore record the governing legal context rather than impose one universal test.
Educational intention can be fragile
Students often genuinely intend to improve but form intentions at the wrong level: “study harder” rather than “after dinner, redo the three questions I got wrong and explain the correction”. Converting broad intention into implementation structure can improve follow-through.
Intentions can be diagnostic signals
A learner’s stated intention reveals motivation and perceived goal. Comparing that intention with actual behaviour can help identify whether the weak link is motivation, planning, capability, environment or feedback.
Preferences can change intention without changing values
A person can value learning consistently while preferring a different study method after new experience. Intention revision may therefore reflect better route selection rather than weakened commitment to the underlying value.
Values constrain acceptable intentions
Ethical, legal and personal values can make some possible intentions unacceptable even if they would achieve a desired outcome. Classification should preserve hard boundaries around what may be pursued.
Conflicting intentions require prioritisation or integration
Some conflicts can be resolved by sequence, compromise or redesign. Others require choosing one intention over another. The resolution process should not erase the existence of the displaced intention if the history matters.
Intentions can become stale
A recorded intention may no longer reflect the actor’s current state. Old plans, abandoned drafts and outdated strategy documents should carry effective dates and lifecycle status.
Reconfirmation matters for consequential action
Before acting on an old intention involving money, publication, privacy or irreversible change, confirm that the intention remains active and authorised rather than assuming persistence.
Intention evidence has different strengths
A casual statement, signed agreement, repeated action, approved budget and irreversible preparatory step provide different evidence. No single signal should be treated as universally decisive.
Sincerity and feasibility are different
An actor can sincerely intend an impossible outcome. Conversely, a feasible action may be stated strategically without genuine commitment. Classify intention strength, capability and feasibility separately.
Intention confidence should be local
We may be certain about the declared intention but uncertain about persistence, sincerity or exact implementation route. One global confidence score can hide these distinctions.
AI should be cautious about intention inference
Language models can infer likely intent from text or behaviour, but those inferences should not be treated as hidden facts about a person. Where consequences matter, systems should prefer explicit user instructions, declared goals and observable evidence over speculative mental-state attribution.
AI agents need explicit operational intention
An agent should act from an authorised instruction or goal, not from an invented guess about what the user “must have intended”. Ambiguity should trigger bounded interpretation, reversible action or confirmation depending on consequence.
A practical intention record
- intention ID and description;
- actor or authorised institution;
- target action or outcome;
- source: declared, documented or inferred;
- specificity level;
- time horizon;
- conditional triggers;
- commitment strength;
- linked goals and preferences;
- required capabilities and resources;
- constraints and blockers;
- competing intentions;
- supporting evidence;
- confidence dimensions;
- execution state;
- revision trigger;
- effective date and version;
- status: tentative, active, deferred, fulfilled, abandoned or superseded.
Worked example: study intention
A student says, “I intend to study more.” The intention is declared but abstract. It has no start cue, duration, subject, method or verification. A more operational version might be: “After dinner on Monday, Wednesday and Friday, I will spend 30 minutes redoing the mathematics questions I previously got wrong and check each answer against the worked solution.”
The underlying goal may be unchanged, but the intention becomes more specific, time-bounded and executable. If follow-through still fails, diagnosis can move to environmental friction, capability, competing priorities or feedback rather than remaining at the level of motivation.
Worked example: organisational intention
A company announces that it intends to reduce emissions. The statement is evidence of declared intent, not evidence that reduction has occurred. Stronger institutional intention may be reflected in approved targets, capital allocation, executive accountability, procurement rules, measurement systems and published milestones.
Outcome evaluation should then compare stated intent with implementation and measured results without collapsing the three into one judgement.
Worked example: inferred intention from behaviour
If a person repeatedly opens a document, edits the same section and saves drafts, one plausible interpretation is that they intend to revise the document. But they may also be reviewing, learning, testing software or demonstrating a workflow. The correct classification is “inferred intention with supporting behaviour”, not “known intention”.
Questions to ask before accepting an intention classification
- Who is the actor?
- What exactly do they intend to do or bring about?
- Is the intention declared or inferred?
- How specific is it?
- How strong is the commitment?
- When is execution expected?
- Which conditions must hold?
- What capabilities, permissions and resources are required?
- What competing intentions exist?
- What evidence supports persistence of the intention?
- Has behaviour begun?
- What would cause revision, deferral or abandonment?
- What outcome would count as fulfilment?
- How costly would a mistaken intention inference be?
Common classification mistakes
- Confusing desire with intention.
- Confusing intention with action or outcome.
- Treating a vague aspiration as a concrete plan.
- Inferring intention from one ambiguous behaviour.
- Assuming outcome proves prior intention.
- Ignoring conflicting intentions and limited capacity.
- Equating sincerity with feasibility.
- Overwriting abandoned intentions instead of preserving history.
- Treating institutional statements as authorised intent without checking source.
- Allowing old intentions to remain active indefinitely.
- Using AI to infer mental states more confidently than evidence permits.
The deeper idea
Intentions are directional states between thought and action. They matter because plans, commitments and behaviour often originate there, but they are not identical to any of those later structures. A useful system preserves whether the intention is merely contemplated, settled, conditional, resourced, active, blocked, fulfilled or abandoned.
That separation also supports fairer diagnosis. When action does not follow intention, the right question is not automatically “did the person really mean it?” The missing link may be planning, capability, resources, cueing, competing priorities, authority or changed circumstances.
To categorise an intention well is to preserve what the actor means to pursue, how explicitly and strongly that intention is held, when and under which conditions it should become action, what blocks it, and what evidence would show that the intention changed.
Final answer
Categorise intentions by target, source, specificity, commitment strength, time horizon, conditions, feasibility, conflict, evidence, execution state and revision. Keep intentions separate from desires, preferences, goals, commitments, plans and actions, and treat inferred intention cautiously whenever consequences depend on being right about another actor’s internal state.