Source-linked AI summary
Revisiting the Core Ontology and Problem in Requirements Engineering
Ivan Jureta, John Mylopoulos, Stephane Faulkner
TL;DR
The paper argues that Zave and Jackson’s core ontology is incomplete because stakeholders communicate attitudes and other concerns beyond propositional attitudes. It proposes a foundationally grounded ontology based on stakeholder communication and derives a revised requirements problem that addresses limitations of the earlier formulation.
Problem
Zave and Jackson’s ontology omits attitudes, even though stakeholders communicate them alongside beliefs, desires, and intentions.
Method
The paper classifies stakeholder communications using speech-act theory and defines RE concepts and relationships in a restricted vocabulary grounded in the DOLCE foundational ontology.
Results
The revised ontology covers stakeholder communications through concepts including goals, softgoals, quality constraints, plans, and domain assumptions, and yields a new requirements problem formulation.
Takeaways & Limitations
The ontology establishes revised minimum-information standards for RE languages and criteria for determining whether RE has been successfully completed.
Takeaways & Limitations
The revised formulation accepts some claims through available stakeholder evidence rather than formal proof, within the scope of a given development project.
Abstract
from arXiv · showhide
In their seminal paper in the ACM Transactions on Software Engineering and Methodology, Zave and Jackson established a core ontology for Requirements Engineering (RE) and used it to formulate the "requirements problem", thereby defining what it means to successfully complete RE. Given that stakeholders of the system-to-be communicate the information needed to perform RE, we show that Zave and Jackson's ontology is incomplete. It does not cover all types of basic concerns that the stakeholders communicate. These include beliefs, desires, intentions, and attitudes. In response, we propose a core ontology that covers these concerns and is grounded in sound conceptual foundations resting on a foundational ontology. The new core ontology for RE leads to a new formulation of the requirements problem that extends Zave and Jackson's formulation. We thereby establish new standards for what minimum information should be represented in RE languages and new criteria for determining whether RE has been successfully completed.
1 Introduction
Zave and Jackson’s ontology established minimum information standards and a requirements problem for judging RE completion, but its interpretation and coverage remain contested. The paper proposes a stakeholder-grounded ontology and revised requirements problem to address these limitations.
- Zave and Jackson’s core ontology specifies minimum information that RE languages should represent and supports a precise requirements problem.
- The ontology’s characterization is deficient because it cannot represent partial requirement fulfilment or distinguish better specifications under fixed assumptions and requirements.
- Non-functional requirements and preferences are omitted from the framework without substitute concepts.
- The paper proposes a core ontology grounded in stakeholder–engineer communication and uses it to formulate a new requirements problem responding to these limitations.
2 Baseline
The paper selects RE concepts by analyzing stakeholder communication as speech acts and distinguishing communicated content from its illocutionary force. It grounds the resulting ontology in DOLCE so that beliefs, desires, intentions, and attitudes can be represented distinctly.
- The ontology starts from stakeholders communicating information to engineers, who classify the content as requirements or domain assumptions before writing specifications.
- Speech-act theory distinguishes communicated content from illocutionary force, such as expressing a desire for a plane ticket to be booked.
- A minimal RE ontology must capture and distinguish beliefs, desires, intentions, and attitudes communicated by stakeholders.
- The concepts are defined by mapping them to DOLCE, a foundational ontology chosen to make philosophical assumptions explicit and support rigorous comparison.
3 Core Ontology for RE
The proposed RE ontology classifies stakeholder communications by speech-act modality and represents goals, quality constraints, softgoals, plans, domain assumptions, and attitudes. It extends prior formulations by making nonfunctional requirements, justified approximations, intentions, and stakeholder evaluations explicit.
- Classifying communicated content: The ontology separates communicated content from modality, using belief, desire, intention, and attitude to classify stakeholder speech acts.The engineer first identifies speech-act type, then associates its content with the relevant modality.
- Goal, quality constraint, and softgoal: Goals represent desired content that neither describes qualities nor constrains quality values, while quality constraints specify desired values for qualities.This distinction grounds the functional versus nonfunctional requirement taxonomy in the ontology’s treatment of qualities.
- Goal, quality constraint, and softgoal: Quality constraints require well-defined, stakeholder-shared quality spaces, because subjective or poorly communicable evaluations make verification elusive.The paper therefore requires shared or acceptable quality spaces when defining quality constraints.
- Goal, quality constraint, and softgoal: Softgoal satisfaction requires quality constraints that justifiably approximate softgoals, replacing unsupported contribution claims with explicit correlation and justification.The ontology treats justified approximation as a more rigorous interpretation of contribution links rather than rejecting those links outright.
- Plan: Plans encode communicated intentions and should lead to states satisfying goals and quality constraints without violating domain assumptions.The requirements problem is recast around finding plans P such that domain assumptions and plans support the desired goals and quality constraints.
- Plan: The revised requirements problem uses non-monotonic reasoning and accepts sufficient evidence within the project scope rather than requiring deductive proof.Runtime execution must remain consistent with the plan constraints, while third-party services may be represented as input-output restrictions rather than fully specified behavior.
4 Requirements Problem Revisited
The revised requirements problem seeks a specification that satisfies compulsory concerns, addresses optional concerns as far as feasible, and respects stakeholder preferences across alternatives. It also replaces a monotonic satisfaction view with a formulation that accommodates inconsistent assumptions, softgoal approximations, and attitudes.
- A non-monotonic consequence relation is proposed because adding information can invalidate tentative evaluations, while attitudes remain relevant to lifecycle-wide stakeholder evaluation.
- The revised problem starts from communicated compulsory and optional assumptions, goals, quality constraints, softgoals, plans, and attitudes.
- Alternative consistent subsets of domain assumptions are considered because communicated assumptions may conflict.
- Preferences require selecting a solution for which no preferred alternative combination of assumptions, goals, or quality concerns exists.
- The engineer must find a specification whose plans make goals hold, avoid violating domain assumptions, and satisfy quality constraints, including justified softgoal approximations.
- The formulation seeks specifications that come as close as feasible to compulsory requirements, satisfy as many optional requirements as feasible, and rate favorably in stakeholder attitudes.
5 Conclusions
The paper revisits RE’s basic ontology by grounding stakeholder communications in explicit speech acts and mental states. It proposes concepts and relationships that broaden requirements representation, improve treatment of nonfunctional requirements, and yield a revised requirements problem, while noting unresolved framework limitations.
- The paper revisits RE’s core concepts because existing ontology and requirements-problem formulations are incomplete and deficient in several respects.
- The ontology uses a restricted DOLCE vocabulary with explicit speech acts and mental states to classify stakeholder communications.
- Its concepts include goals, softgoals, quality constraints, plans, and domain assumptions, linked by optionality, preference, justified approximation, and non-monotonic consequence.
- Goals, quality constraints, and softgoals together cover functional and nonfunctional requirements while distinguishing verifiable requirements from subjective, difficult-to-share softgoals.
- The revised ontology leads to a requirements-problem formulation that the authors argue fits intuition better and avoids pitfalls in Zave and Jackson’s formulation.
- Established RE frameworks still lack capabilities for representing and reasoning about all ontology terms, including approximation justification and attitudes.