Simply having advanced digital capabilities in an organization aids decision quality, but it does not enhance trust in the same way. These findings were published in Scientific Reports.
The Need for Explainable Infrastructure AI
Infrastructure is one of the toughest testing grounds for AI adoption. What sets this field apart from most other AI use cases is the combination of pressures, outcomes tied to public safety, regulatory oversight, and personal professional liability, therefore making it imperative to determine whether an AI system can actually be trusted in this sector.
Much of the prior work in this space has centered on why people adopt new technology and how well models predict outcomes, pointing to factors such as a system’s utility and ease of use, as well as an organization's readiness to adopt it. Separately, research that specifically considers explainable AI has tended to focus on making models interpretable, fair, and technically well-documented.
No time now? Download a PDF and finish later.
What's been missing is empirical work that connects explainability to professionals’ interpretation of how decision-support tools actually perform in an infrastructure setting. To close that gap, this research tests a socio-technical model that maps how transparency, explainability, accountability, and digital readiness each shape trust in explainable AI, and in turn, perceived decision quality.
Survey Design and Data Collection
The researchers gathered their data with a standardized questionnaire aimed at professionals whose jobs touch infrastructure planning and engineering, rolling out digital tools, or making day-to-day operational calls at infrastructure-focused organizations.
Beyond basic demographic questions, it captured algorithmic transparency, decision accountability, organizational digital readiness, perceived explainability, trust in explainable AI, and infrastructure decision quality, each rated on a five-point scale from strongly disagree to strongly agree.
Before rolling it out widely, the researchers had three academics and two industry practitioners check the survey for validity, then piloted it with 30 infrastructure professionals and tweaked some wording based on that feedback.
The team drew a careful distinction between two related ideas: algorithmic transparency captured how visible the AI's internal workings were, whereas perceived explainability captured whether users could actually follow the reasoning behind its outputs.
The target pool spanned engineers, project managers, technical specialists, planners, and consultants working across government bodies, private infrastructure companies, engineering consulting firms, and technology vendors. Each person surveyed needed some hands-on background with digital tools, AI-driven systems, or data-based decision support in their work.
Out of 600 surveys sent out, 283 came back usable, a response rate of 47.2%. To check for non-response bias, the team compared early respondents against late ones using a wave-analysis approach; t-tests found no meaningful differences between the two groups.
For the analysis, the researchers used partial least squares structural equation modeling (PLS-SEM) in SmartPLS 4, with a bootstrap procedure (5000 subsamples) to assess significance. This happened in two steps, firstly checking the measurement model's reliability and validity (both convergent and discriminant), then moving on to test the structural model itself.
Findings and Structural Model Outcomes
Most of those surveyed were men in the 25–34 age bracket who hold a bachelor's degree. Infrastructure engineers and project managers made up the biggest chunks of the sample.
None of the common-method-bias checks turned up any meaningful distortion. In Harman's single-factor test, the leading factor accounted for just 32.4% of the variance, comfortably below the 50% danger line.
Reliability and validity checks on the measurement model held up well too. Applying the Fornell-Larcker criterion, every construct's diagonal value came in higher than its correlations with other constructs, supporting discriminant validity.
Model fit was deemed acceptable, with a standardized root mean square residual (SRMR) of 0.062. As for explanatory power, the model accounted for 43.8% of the variance in trust in explainable AI (R2 = 0.438) and 69.1% of the variance in infrastructure decision quality (R2 = 0.691).
Three factors, namely, algorithmic transparency, perceived explainability, and decision accountability, each came out with a positive effect on both trust and perceived decision quality. Organizational digital readiness was a partial exception: it boosted decision quality on its own, but its link to trust didn't reach statistical significance.
Of everything tested, trust had the single biggest effect on decision quality. When the researchers ran a mediation analysis, trust statistically mediated part of the effects of transparency, explainability, and accountability on decision quality, though not the effect of digital readiness, whose indirect path wasn't significant.
Taken together, the results point to algorithmic transparency, perceived explainability, decision accountability, and trust as meaningful drivers of better infrastructure decisions. The fact that digital readiness didn't significantly influence trust implies that simply being technologically well-equipped isn't enough to earn people's psychological confidence in a system.
Beyond Digital Readiness to Real Trust
The study set out to identify what makes explainable AI actually work as a decision-support tool in infrastructure settings. It suggests that transparent algorithms, clear explanations, and accountable decision-making processes meaningfully increase both trust in the system and the quality of the decisions people make with it.
Digital readiness still plays a role, as it is associated with improved decision quality, but it isn't what builds trust. Trust itself, meanwhile, emerged as one of the strongest predictors of decision quality, and it amplifies how much transparency, explainability, and accountability end up mattering for the final outcome.
Journal Reference
Waqar, A., Ali, A.S., Alrasheed, K.A. et al. (2026). Socio-technical determinants of explainable artificial intelligence for infrastructure decision support. Scientific Reports. https://www.nature.com/articles/s41598-026-60853-8.
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.