h t t p s : / / d o i . o r g / 1 0 . 1 1 4 5 / 1 8 1 0 2 9 5 . 1 8 1 0 3 2 9

Ingest: Requirements Reflection: Requirements as Runtime Entities

Type: kb/sources/types/ingest-report.md

Classification

This is an ICSE 2010 scientific vision and design paper: it defines requirements reflection, sketches a two-stratum runtime architecture, and states research challenges rather than reporting a realized system or controlled evaluation. Author: researchers at Lancaster University and University College London working in requirements engineering, reflective middleware, goal modeling, and self-adaptive systems; the paper locates its proposal within those research literatures and the authors' related modeling work.

Summary

The paper argues that requirements monitoring leaves a running system with low-level derived artifacts but not the high-level goals, assumptions, alternatives, and relationships that motivated them. It proposes reifying those requirements as first-class runtime entities in a goal-oriented meta-model, exposing operations for inspecting and changing them, and semantically synchronizing that requirements stratum with a reflective architecture stratum so changes on either side can affect or be checked against the other. It also proposes runtime reasoning about uncertainty and interactive multi-objective trade-offs. Read it as a concrete structural agenda for putting a software system's purposes on a causal path to adaptation, not as evidence that the architecture works or that the system learns its own requirements model.

Quotes

No source quotes have been retained yet.

Connections Found

The paper is the requirements-specific structural-lineage anchor for the lineage account for reflective theory refinement: it places an explicit model of a system's purposes and goal relations on a proposed causal path to changes in the running architecture, but supplies neither empirical theory refinement nor delayed read-back. It is also a conceptual precedent for Reflection buys addressability and Reflective coverage is graded across representational forms, because its requirements and architecture strata expose different entities, operations, and a required transfer mapping. Compared with Computational Reflection, it specializes causally connected self-representation to requirements and makes cross-stratum synchronization the central unsolved problem.

Extractable Value

  1. A requirements-level self-model can be causally relevant without being a learned theory -- The paper proposes runtime objects for goals, refinements, alternatives, assumptions, conflicts, and their links to architecture. This supplies the self-target and adaptation-path half of reflective theory refinement while keeping the empirical theory-revision half distinct. [quick-win]
  2. The represented vocabulary determines the available questions and interventions -- A KAOS-like meta-model would let a system navigate goal relations, identify responsible agents, inspect assumptions, and invoke operations such as adding, deleting, replacing, or assigning requirements and goals. This gives a concrete requirements-specific instance of runtime representation creating a control surface. [quick-win]
  3. Causal connection requires an explicit transfer mechanism between representations -- The proposal separates requirements and architecture into strata, each with base and meta-levels, then requires semantic synchronization so a requirements change can generate architectural changes and an architectural change can be checked against requirements. Merely retaining goals does not establish this connection. [deep-dive]
  4. Most of the effective update space remains designer-supplied -- The paper leaves the goal language and meta-model, exposed primitives, requirements-to-architecture mapping, monitoring boundary, uncertainty formalism, decision procedure, and human interaction policy to designers. Runtime reprioritization or replacement therefore occurs inside a fixed decomposition rather than revising the machinery that defines valid representations and adaptations. [deep-dive]
  5. The evidence supports historical positioning, not effectiveness -- The paper offers a motivating vacuum-cleaner scenario, conceptual comparison with prior work, and a research agenda, but no implementation, benchmark, ablation, or outcome data. It can ground a lineage claim and sharpen a baseline; it cannot show that requirements reflection improves adaptation, explanation, safety, or learning. [just-a-reference]

Limitations (our opinion)

The central mechanism is underspecified and untested. The paper does not define or evaluate the semantic synchronization that would translate a changed goal into safe architectural changes, nor does it show that reverse monitoring can diagnose which requirement should change. Its vacuum-cleaner case is illustrative, not evidence, and the proposed KAOS-derived representation may omit requirements or environmental phenomena that do not fit its designer-chosen ontology. Human participation is proposed for resolving multi-objective conflicts, but the allocation of authority, elicitation of preferences, and behavior under disagreement are not evaluated. Most importantly for reflective theory refinement, the paper does not learn or empirically revise the requirements meta-model, synchronization rules, monitors, adaptation operators, or evaluation policy; it proposes reasoning and modification within that supplied structure.

Update the lineage account for reflective theory refinement to replace its bare requirements-reflection DOI with this ingest and state explicitly that the paper supplies a proposed requirements-level self-model and synchronization obligation, while leaving empirical read-back, theory revision, and the adaptation decomposition designer-supplied.