Tailoring explanations in conversational recommendations: The impact of decision contexts and user interfaces

Abstract

Explainability is crucial for building trust in traditional recommendation systems, yet its role in conversational settings is underexplored. Across three experimental studies (N = 1,429), we used between-subjects designs featuring diverse product categories (cameras, smartwatches, headphones) to examine the interactive effects of post hoc explanations (expert validation-based vs. consensus validation-based) and decision-making domains (hedonic vs. utilitarian) on consumer responses to conversational recommendations. We further examined how consumer decision-making styles (intuitive vs. rational) and user interfaces (text-based vs. voice-based) moderated these effects. Results show that post hoc explanations enhance perceived transparency and interpretability, thereby increasing consumer trust in conversational recommendations. In text-based interfaces, consumers making hedonic decisions preferred consensus-based explanations, whereas no clear preference emerged for utilitarian decision-makers. In voice-based interfaces, utilitarian consumers favored consensus-based explanations, while no significant preference was observed for hedonic decisions. Furthermore, intuitive consumers preferred consensus-based explanations for hedonic decisions and expert-based explanations for utilitarian decisions. Rational consumers consistently favored consensus-based explanations across both decision-making domains. These findings provide valuable insights for designing conversational recommendation systems on e-commerce platforms. By tailoring explanations to decision domains, user interfaces, and consumer decision-making styles, businesses can foster greater trust and engagement, driving more favorable purchasing behaviors and improving business outcomes.