Back to search

Athens University Economics and Business (AUEB)

Recommendations and conversational AI: the impact of trust and persuasive explanations in user adoption

Abstract

dc:description

Recommender systems have evolved from simple ranking tools into interactive environments that actively shape user decisions. Today, these systems do not only suggest items, but also influence how users interpret information, form preferences, and make choices. As they become more conversational and agent-like, the focus gradually shifts from evaluating the recommended item to evaluating the system itself, raising important questions about trust and user adoption. This dissertation aims to examine how persuasion mechanisms embedded in recommendation environments influence user behavior, and how these mechanisms evolve as systems transition from traditional recommender systems to AI-based recommendation agents. Building on the Elaboration Likelihood Model (ELM), the study adopts a process-oriented perspective, where recommendation cues operate through central and peripheral routes of information processing, shaping both user evaluations and behavioral responses. The dissertation draws on three empirical studies. The first two studies focus on traditional recommender system environments and examine how persuasive explanations, user motivation, and personality traits influence recommendation acceptance. The findings show that persuasive strategies are particularly effective under low motivation conditions, where users rely on peripheral cues, while their impact is limited or even negative under high motivation, where users engage in more deliberate evaluation. In addition, the results reveal that persuasion is not uniform across users, as personality traits significantly shape responsiveness to different influence strategies. The third study extends this perspective to AI-based recommendation systems, where the system itself becomes the main object of evaluation. The findings show that perceived accuracy and perceived explainability influence trust formation, while trust operates as a dual and closely interconnected mechanism. More specifically, Emotional Trust strongly predicts Cognitive Trust, a pattern that is consistent with an affective-first trust formation process. This suggests that emotional reassurance may facilitate users’ cognitive evaluations of the system. However, this relationship does not establish that Emotional Trust develops temporally before Cognitive Trust. Importantly, the dissertation distinguishes between two forms of adoption: the use of systems as Decision Aids and as Delegate Agents. The results show that these forms of adoption involve different psychological processes, with decision-aid use being supported by both Cognitive and Emotional Trust, whereas delegation is driven primarily by Emotional Trust. Overall, the dissertation demonstrates that persuasion alone is not sufficient to explain user behavior in modern recommendation environments. As systems become more interactive and autonomous, the focus shifts from persuasion-driven recommendation acceptance to trust formation toward the system itself, which becomes a key mechanism underlying adoption. The study contributes to theory by integrating the Elaboration Likelihood Model with trust formation in AI environments and by proposing a unified framework that explains how system-related cues are associated with different forms of trust and, ultimately, with distinct modes of adoption. Furthermore, it highlights the strong relationship between Emotional and Cognitive Trust and clarifies how different levels of system autonomy are associated with different trust configurations. From a practical perspective, the findings provide guidance for designing recommendation systems that are not only effective, but also transparent, psychologically aligned with users and capable of supporting both decision-making and delegation.

Degree

thesis:*
Grantor dc:publisher
Athens University Economics and Business (AUEB)
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Gkika, Sofia
  • Γκίκα, Σοφία

Subjects

dc:subject × 24

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/62282

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Gkika, Sofia; Γκίκα, Σοφία. Recommendations and conversational AI: the impact of trust and persuasive explanations in user adoption. Athens University Economics and Business (AUEB), 2026. http://hdl.handle.net/10442/hedi/62282