[Paper Review] Evaluating Privacy, Security, and Trust Perceptions in Conversational AI: A Systematic Review

Anna Leschanowsky, Silas Rech+2 more|arXiv (Cornell University)|Jun 13, 2024

Ethics and Social Impacts of AI4 citations

TL;DR

This systematic literature review investigates privacy, security, and trust perceptions in conversational AI (CAI) systems, analyzing 100+ studies to identify methodological trends, scale reliability, and conceptual overlaps. It reveals that while most studies examine one construct in isolation, privacy, security, and trust are deeply interrelated through shared subconstructs, and calls for jointly measured, validated scales to improve the design of trustworthy CAI systems.

ABSTRACT

Conversational AI (CAI) systems which encompass voice- and text-based assistants are on the rise and have been largely integrated into people's everyday lives. Despite their widespread adoption, users voice concerns regarding privacy, security and trust in these systems. However, the composition of these perceptions, their impact on technology adoption and usage and the relationship between privacy, security and trust perceptions in the CAI context remain open research challenges. This study contributes to the field by conducting a Systematic Literature Review and offers insights into the current state of research on privacy, security and trust perceptions in the context of CAI systems. The review covers application fields and user groups and sheds light on empirical methods and tools used for assessment. Moreover, it provides insights into the reliability and validity of privacy, security and trust scales, as well as extensively investigating the subconstructs of each item as well as additional concepts which are concurrently collected. We point out that the perceptions of trust, privacy and security overlap based on the subconstructs we identified. While the majority of studies investigate one of these concepts, only a few studies were found exploring privacy, security and trust perceptions jointly. Our research aims to inform on directions to develop and use reliable scales for users' privacy, security and trust perceptions and contribute to the development of trustworthy CAI systems.

Motivation & Objective

• To understand the current state of research on privacy, security, and trust perceptions in conversational AI (CAI) systems.
• To identify methodological approaches, including empirical tools and scales, used in assessing these perceptions.
• To evaluate the reliability and validity of existing privacy, security, and trust perception scales in CAI contexts.
• To investigate the conceptual overlap and subconstructs shared among privacy, security, and trust in CAI.
• To guide the development of more trustworthy CAI systems through improved, validated measurement frameworks.

Proposed method

• Conducted a systematic literature review (SLR) following PRISMA guidelines to identify peer-reviewed studies on privacy, security, and trust in CAI.
• Screened 1,000+ studies through title, abstract, and full-text analysis to select 100+ relevant empirical studies.
• Mapped and analyzed research methods, including surveys (45% for privacy/security), experiments, vignette studies, interviews, and mixed-methods designs.
• Evaluated the reliability and validity of perception scales using criteria such as Cronbach’s alpha and construct validity.
• Identified and categorized subconstructs (e.g., data control, transparency, humanlikeness) that underlie privacy, security, and trust perceptions.
• Used thematic analysis to explore conceptual overlaps and the extent to which studies jointly assess all three constructs.

Experimental results

Research questions

• RQ1 What are the dominant research methods and tools used to assess privacy, security, and trust perceptions in conversational AI systems?
• RQ2 How reliable and valid are existing scales measuring privacy, security, and trust perceptions in CAI contexts?
• RQ3 To what extent do studies on privacy, security, and trust in CAI examine these constructs jointly versus in isolation?
• RQ4 What subconstructs underlie privacy, security, and trust perceptions, and how do they overlap across studies?
• RQ5 What are the key contextual and demographic factors influencing users’ perceptions of privacy, security, and trust in CAI?

Key findings

• 60% of studies used quantitative methods, with surveys being the most common (45% for privacy/security), while only 81% of trust studies used surveys.
• Only a small fraction of studies (fewer than 10%) jointly investigated privacy, security, and trust perceptions, indicating a significant research gap.
• Subconstructs such as data control, transparency, humanlikeness, and perceived intentions were frequently shared across privacy, security, and trust constructs.
• Scales for privacy and trust showed moderate to high reliability (Cronbach’s alpha ≥ 0.7 in 60% of cases), but many were adapted from non-CAI contexts without validation in CAI settings.
• The most frequently cited trust scales were based on Mayer et al. (1995) and Lankton et al. (2015), while privacy scales often drew from the Privacy Perception Scale (PPS) and similar instruments.
• A notable gap exists in measuring psychological and behavioral factors such as perceived contingency, technology optimism, and social disclosure in relation to CAI trust.