Explainable AI and Trust in AI-Driven Security Systems – Modified Delphi Study (Interviews and Questionnaire Data)

This is the data from The Influence of Explainable AI on Trust in AI-Driven Security Systems (Honours Empirical Research; 2025).

Abstract: In South Africa, organisations face rising cyber threats. Low trust in AI-driven security systems can result in significant costs. This modified Delphi study engaged local cybersecurity professionals to examine how explainable AI affects trust in such systems. Explainable AI (XAI) helps untangle complex machine learning algorithms. XAI brings transparency to AI-driven systems. However, few studies have explored the trust relationship between XAI and cybersecurity professionals. Semi-structured interviews revealed key themes about XAI's influence on trust. The findings informed the development of a questionnaire to confirm trust factors, as well as the key benefits and challenges of XAI. Institutional isomorphism guided the study. It is argued that dominant pressures, like those from the cybersecurity profession or organisational imitation, influence XAI adoption. Experts reached consensus that accountability, transparency, and verifiability are primary contributors to trust. XAI serves as a mediating mechanism between organisational pressures and practitioner confidence. These insights support broader human-AI collaboration in cybersecurity and provide practical guidance for responsible AI implementation.

The dataset contains Word transcripts from Round 1 interviews (17 participants) with security professionals as well as Round 2 Questionnaire (Excel documents extracted from Microsoft Forms). The questionnaire used

Interview transcripts were coded and analysed using NVivo. Questionnaire responses were analysed in Microsoft Excel by compiling responses and tallying frequencies to assess whether consensus was achieved. Consensus was defined as 70%.

Period of data collection: 11 July 2025 to 16 October 2025.

Location of data collection: Online