Towards effective communication of uncertain information

Zhao, Yu

Towards effective communication of uncertain information

Publication Date

December 31, 2026

Creator

Zhao, Yu

University of Nottingham, UK

Abstract

Uncertain information is pervasive in real-world contexts, due to measurement limitations, the complexity and ambiguity of what is described; this is notably the case for human perceptions, which are often shaped by subjectivity and conceptual vagueness. In the social sciences, Likert and numeric scales are the predominant formats in survey-based research for eliciting responses about human perceptions. These responses are often analysed collectively to measure latent psychological constructs—such as mood, trust, and satisfaction—whose understanding is consequential for decision-making. While efficient to administer and analyse, the single-valued response formats of these scales can constrain the authentic expression of uncertainty and impose false precision. Over recent decades, fuzzy sets and intervals have been put forward to represent and communicate uncertainty in human perceptions, from elicitation through analysis to decision support. Existing methods across these stages, however, have largely been developed independently, often resting on incompatible assumptions and pursuing divergent goals—for example, some discard uncertainty in favour of simplicity, while others retain it in forms too complex to be interpreted in practice. Building on the state of the art, this thesis aims to develop and evaluate methods for communicating uncertainty in human-sourced information across an end-to-end workflow, in support of uncertainty-aware decision-making. This thesis further demonstrates the practical value of the proposed methods through their application within service research. This application focus is motivated by the Rail Safety and Standards Board as a sponsor of this research, together with the practical need to communicate uncertainty in rail passengers’ perceptions to stakeholders in support of rail service development. More specifically, this thesis focuses on interval-valued representations as a unifying basis for communicating uncertainty throughout a conventional service evaluation workflow. The thesis contributions are organised around the key stages of this workflow: At the elicitation stage, this thesis articulates the necessity of capturing uncertainty in customer responses, as the intangibility and subjectivity of services make customer perceptions difficult to elicit with precision. A UK rail passenger survey is designed to elicit comparable interval- and single-valued responses on the commonly used service evaluation constructs, producing an empirical dataset for systematic comparison across subsequent stages. Reliability assessment is adopted as the subsequent stage, as it is essential for evaluating whether items intended to measure the same construct exhibit sufficient internal consistency. While Cronbach’s α is the most widely used reliability coefficient and has recently been extended to accommodate interval-valued responses, this extension introduces interpretive assumptions that may be misleading in real-world settings. This thesis therefore introduces a novel extension which avoids these assumptions. The behaviour of the proposed coefficient is then evaluated through simulation against the existing extension. Following reliability assessment, responses are often summarised to provide a succinct overview and facilitate comparison across items. For this purpose, recent research has developed methods to model interval-valued responses as fuzzy sets, which, while containing rich information, can be difficult to interpret given their two-dimensional complexity. This thesis therefore develops and evaluates methods for summarising fuzzy sets as intervals in order to communicate uncertainty while enabling efficient comparison in practice. Decision-support methods typically serve as the final stage of service evaluation, transforming collected responses into actionable insights. Widely used methods such as Importance–Performance Analysis cannot present uncertainty, and may therefore mislead decision makers with a false sense of precision. The thesis therefore extends these methods with interval-valued representations, and demonstrates these extensions for uncertainty-aware decision-making using the UK rail passenger survey data. To facilitate wider adoption in practice among researchers and practitioners from a variety of fields, it is important to make novel methods accessible through software support. An open-source R toolkit is thus developed as a final stage to provide an integrated implementation of the methods reviewed and developed in this thesis, with its use demonstrated through a service evaluation workflow using the UK rail passenger survey data. In summary, this thesis advances the state of the art in the analysis and application of interval-valued representations for communicating human-sourced uncertain information, ultimately strengthening the basis for well-informed decision-making in data-driven contexts.

Item Type

ethesis

Thesis Type

PhD

Supervisors

Wagner, Christian

University of Nottingham

Ryan, Brendan

University of Nottingham

Pekaslan, Direnc

University of Nottingham

Keywords