AVOL™ Evaluation Contract
Status: Normative
Version: 1.1
Effective Date: 09-January-2026
The HTML content on avolontology.org is authoritative. PDFs are non-authoritative reference artifacts.
1. Purpose and Scope
This Evaluation Contract defines the formal, binding specification governing how an AVOL™ (Adaptive Value Ontology Layer) evaluation is formed, executed, and interpreted.
This document is authoritative for determining:
- What constitutes a valid AVOL evaluation
- What inputs are admissible
- Where AVOL responsibility terminates
- What AVOL explicitly does not provide
- What guarantees AVOL provides as explicitly enumerated in Section 8
- How outputs may be relied upon by machines, agents, regulators, and counterparties
This Evaluation Contract extends but does not modify the Canonical Explainer or the Declared Evaluation Context. All constraints, prohibitions, and definitions in those documents are binding and controlling.
2. Definitions
2.1. Evaluation
An Evaluation is a deterministic computational act performed by AVOL that transforms a declared set of admissible inputs into a single, scalar, normalized value representation using predefined conversion and aggregation rules.
An evaluation exists only when all required inputs are present and valid, and the evaluation is executed under a declared ruleset version. For the avoidance of doubt, the term ‘evaluation’ as used herein denotes a mechanical value translation process and does not imply judgment, assessment, suitability analysis, or decision-making of any kind.
2.2. Evaluation Output
An Evaluation Output is the sole canonical artifact produced by an AVOL evaluation. It consists of:
- A numeric scalar value representing the aggregated, cash-equivalent translation of declared value components
- Associated metadata sufficient to identify:
- Declared Evaluation Context identifier
- Ruleset version
- Aggregation formula identifier (or ruleset-internal formula reference)
- Conversion constants used
- Evaluation timestamp
Associated metadata is descriptive only and does not modify, contextualize, or expand the semantic meaning of the Evaluation Output. No other semantic meaning is attached to the output.
3. When an Evaluation Exists
3.1.
An AVOL evaluation exists if and only if all of the following conditions are satisfied:
3.1.1.
The Declared Evaluation Context is specified and valid
3.1.2.
A versioned AVOL ruleset is specified
3.1.3.
All required inputs (Section 4.1) are provided and valid
3.1.4.
The evaluation is executed deterministically under that ruleset
If any condition above is not met, no evaluation exists, and no output is valid or authoritative.
4. Inputs and Preconditions
4.1. Required Inputs
The following inputs are mandatory for all AVOL evaluations:
4.1.1. Option Definition – A complete description of the transaction option being evaluated, including:
- Base monetary price, which shall always be admissible regardless of the presence or absence of non-price components.
- An evaluation may yield a zero aggregate non-price value where no in-scope, mechanically knowable non-price components exist for the option under the applicable Declared Evaluation Context.
- All quantifiable, non-price value components that are both (i) declared as in-scope by the applicable Declared Evaluation Context and (ii) mechanically knowable at the time of evaluation.
4.1.2. Declared Evaluation Context – A formally declared context specifying:
- Domain and transaction type
- Temporal scope (e.g., effective dates)
- Scope constraints on admissible components
4.1.3. Ruleset Version Identifier – The specific AVOL ruleset version governing:
- Inclusion/exclusion rules
- Aggregation formula
- Conversion constants
4.2. Optional Inputs
4.2.1.
AVOL does not accept optional inputs that alter computation.
4.2.2.
Any input not explicitly declared as required is out of scope and ignored.
4.2.3.
User attributes, preferences, loyalty status, intent signals, or probabilistic data are not admissible inputs at the canonical layer.
4.2.4.
Incentive or margin fields (including commission, take-rate, partner status, placement agreements) are not admissible inputs within the canonical layer.
4.3. Missing or Malformed Inputs
4.3.1.
Invalid evaluations must not produce outputs
4.3.2.
Missing inputs designated as required by the applicable Declared Evaluation Context invalidate the evaluation.
4.3.3.
Malformed inputs invalidate the evaluation.
4.3.4.
AVOL will not infer, repair, substitute, or impute missing or malformed data.
4.3.5.
When an evaluation is invalid, AVOL must return an invalidation status and must not return a numeric substitute value.
5. Relationship to AVOL-Base and Declared Evaluation Context
5.1. AVOL-Base
AVOL-Base is the default instantiation of this Evaluation Contract. Unless explicitly declared otherwise, all evaluations are defined to be AVOL-Base evaluations.
5.2. Default Context Clarification
If no Declared Evaluation Context is supplied, the evaluation SHALL either (a) fail as invalid, or (b) execute under AVOL-Base. AVOL-Base is a declared, versioned default context explicitly defined as the absence of additional external entitlements. It is not selected at runtime and does not reflect discretionary context choice by AVOL.
5.3. Context Binding
An evaluation is bound to its Declared Evaluation Context. Outputs are not portable across contexts and must not be interpreted outside the context under which they were produced.
6. Determinism and Canonical Guarantees
6.1. Determinism
For a given set of identical inputs, declared context, and ruleset version:
6.1.1.
AVOL must always produce the identical output.
6.1.2.
Outputs are invariant across users, calling systems, execution environment, and time, within the same ruleset version.
6.1.3.
No stochastic, adaptive, or probabilistic behavior is permitted.
6.2. Canonicality
AVOL outputs are canonical representations of translated value components only.
6.2.1.
They are not estimates, forecasts, recommendations, rankings, or suitability determinations.
6.2.2.
Numeric magnitude within an AVOL output conveys no directive, preferential, or ordering semantics and does not imply suitability, recommendation, or selection.
6.3. Inspectability
Every evaluation output must be fully decomposable into:
6.3.1.
Declared aggregation logic.
6.3.2.
Explicit conversion constants.
6.3.3.
Declared input components.
6.3.4.
No opaque or non-inspectable mechanisms are permitted within an AVOL evaluation.
7. Versioning and Ruleset Stability
A ruleset is “allowed” if it is published, versioned, and designated as operative for the integration. AVOL SHALL NOT select among multiple eligible rulesets at runtime. Selection of which ruleset to invoke is an integration-level responsibility and does not constitute AVOL evaluation behavior.
7.1. Any change to conversion constants or aggregation logic constitutes a new ruleset version.
7.2. Outputs generated under different ruleset versions are not equivalent.
7.3. Ruleset versions are immutable once declared.
AVOL provides no guarantee of equivalence across ruleset versions.
8. What AVOL Guarantees
AVOL explicitly guarantees the following, and only the following:
8.1. Neutrality – Computation is invariant to:
8.1.1.
Incentives
8.1.2.
Affiliations
8.1.3.
Commercial outcomes
8.1.4.
User identity
8.2. Scope Fidelity
Only declared, admissible inputs are represented. Nothing more.
8.3. Consistency
Identical evaluations yield identical outputs.
8.4. Representation Accuracy
Outputs accurately represent the arithmetic translation of declared inputs under the declared ruleset.
9. What AVOL Does NOT Guarantee
AVOL explicitly does not guarantee:
9.1. That outputs constitute an actual rebate, discount, credit, or entitlement
9.2. That any value will be realized
9.3. That any benefit will be redeemed
9.4. That outputs reflect future states or outcomes
9.5. That outputs are optimal, superior, or preferable
9.6. That outputs are complete beyond declared inputs
9.7. That outputs improve user satisfaction or utility
9.8. Accuracy or completeness of third-party-provided data sources or externally supplied component values
AVOL does not rank, compare, recommend, predict, optimize, or advise.
10. Boundary of Responsibility – AVOL responsibility terminates strictly at value representation. AVOL is not responsible for:
10.1. Optimization
10.2. Recommendation
10.3. Decision-making
10.4. Action selection
10.5. Outcome realization
Any downstream interpretation, ranking, or action is outside the AVOL boundary.
11. Machine and AI Reliance
11.1. Permitted Reliance
Machines and AI agents may rely on AVOL outputs solely as:
- Canonical, normalized value representations.
- Deterministic numeric inputs for downstream systems.
- Inherently comparable across options evaluated under the same Declared Evaluation Context and ruleset version, without additional translation or weighting by AVOL.
- Representable in a machine-readable form (numeric scalar plus metadata) suitable for upstream automated consumption.
- Treated as context-bound values and not portable across Declared Evaluation Contexts or ruleset versions.
11.2. Prohibited Assumptions
Machines must not assume that AVOL outputs:
11.2.1.
Represent recommendations or rankings
11.2.2.
Are optimized for any objective
11.2.3.
Reflect likelihood of benefit realization
11.2.4.
Encode preferences
11.2.5.
Have verified third-party terms, availability, or applicability
11.3. Comparability Clarification
11.3.1.
Evaluation Outputs may be numerically compared across options evaluated under the same Declared Evaluation Context and ruleset version.
11.3.2.
Comparability does not imply preference, superiority, suitability, or recommendation.
11.3.3.
AVOL does not select, rank, order, or prioritize options.
11.3.4.
AVOL makes no assertion as to which option should be chosen.
AVOL outputs are inputs, not decisions.
12. Legal Posture
12.1. Non-Advisory Nature
AVOL outputs do not constitute advice of any kind, including but not limited to:
12.1.1.
Valuation, appraisal, or suitability determinations
12.1.2.
Financial
12.1.3.
Legal
12.1.4.
Commercial
12.1.5.
Consumer
12.2. No Fiduciary Duty
AVOL assumes no fiduciary responsibility to any party.
12.3. No Economic Guarantee
AVOL does not guarantee economic benefit, savings, or advantage, including any implied savings, discount, or “best deal” claim.
12.4. Canonical Reference Only
AVOL serves solely as a canonical value translation and representation layer.
13. Explicit Prohibitions – The following behaviors are explicitly prohibited within the canonical AVOL evaluation layer:
13.1. Ranking or ordering of options
13.2. Recommendation or steering
13.3. Prediction or inference
13.4. Probabilistic or ML-based valuation
13.5. Personalization
13.6. User-specific weighting
13.7. Incentive-aware computation
Violation of any prohibition invalidates canonical compliance.
14. Contract Integrity
This Evaluation Contract is binding on all AVOL evaluations.
Any system claiming to produce AVOL evaluations must conform fully to this specification.
Partial compliance is non-compliance.