Modeling and reasoning about uncertainty in goal models: a decision-theoretic approach

• Published in: Software and systems modeling Vol. 21; no. 6; pp. 1 - 24
• Main Authors: Liaskos, Sotirios, Khan, Shakil M., Mylopoulos, John
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.12.2022; Springer Nature B.V
• Subjects: Automated reasoning; Automation; Compilers; Computer models; Computer Science; Decision theory; Expected utility; Information Systems Applications (incl.Internet); Interpreters; IT in Business; Modelling; Policies; Probability; Programming Languages; Programming Techniques; Reimbursement; Simulation; Software; Software Engineering; Software Engineering/Programming and Operating Systems; Stakeholders; Statistical analysis; Theme Section Paper; Uncertainty; Golog; Markov decision processes (MDP); DT-Golog; Goal modeling
• ISSN: 1619-1366; 1619-1374
• DOI: 10.1007/s10270-021-00968-w

Goal models have been a popular subject of study by researchers in requirements engineering, due to their ability to capture and analyze alternative solutions through which a software system can achieve business objectives. A plethora of analysis methods for automated identification of optimal alternatives have been proposed. However, such methods often assume an idealized reality where all tasks are successfully performed when attempted and all goals are eventually satisfied with certainty when pursued according to a solution. In reality, some tasks run the risk of failure while others produce chance outcomes. In this paper, we extend the standard goal modeling language to allow representation and reasoning about both uncertainty and preferential utility in goals. Tasks are extended to allow for probabilistic effects and preferential statements of stakeholders are captured and translated into utilities over possible effects. Moreover, solutions are not mere specifications (functions, quality constraints, and assumptions), but rather policies, that is sequences of situational action decisions, through which stakeholder goals can be fulfilled. An AI reasoning tool is adapted and used for identifying optimal policies with respect to the value they offer to stakeholders measured against their probability of failure. Evaluation of the approach includes a simulation study and scalability experiments to assess the applicability of automated reasoning for larger problems.