Guided Research Oleksandra Klymenko

Analysing the influence of uncertainty on architectural decision making

Motivation

Many of the software development stages are highly communication-intensive activities that involve developers, architects, testers, end users and other stakeholders [1]. It is especially common in the implementation and maintenance phases that a system or its parts have to be constantly upgraded, fixed, or replaced. Task management tools help to structure this process by implicitly capturing design decisions [2]. 

Being documented in natural language text, tasks descriptions often lack clarity and completeness, causing more time and effort to be spent on their resolution.  
To address this issue, this guided research focuses on a specific kind of natural language phenomenon - linguistic uncertainty.  

First of all, studies of the phenomenon in various application domains are reviewed. Analyzing different existing categorization approaches, we aim to derive new original categories of uncertainty, supported with the evidence collected from a task management system in the domain of Software Engineering. 

Secondly, a series of expert interviews are carried out to understand current challenges caused by uncertainty and get feedback on the proposed categorization. Conducted interviews reveal how experts handle awareness of uncertainty queues and their relation to overall effectiveness in task accomplishment. 

Research outline

1. Literature review: categorization of uncertainty in different research domains (biomedical, Wikipedia).
2. Data exploration: Applying simple rule-based pattern matching to limit the amount of data to explore. 
3. Criteria definition: Defining domain-specific categories for uncertainty expressions.  
4. Validation: expert interviews.

References

[1] Shradhanand, A. Kaur,  S. Jain, “Use of fuzzy logic in software development” in Issues in Information Systems, Volume VIII, No. 2, 2007.

[2] M. Bhat , K. Shumaiev , A. Biesdorf , U. Hohenstein , F. Matthes, “Automatic Extraction of Design Decisions from Issue Management Systems: A Machine Learning Based Approach”.