关键词: 推荐解释线索, 消费者在线多阶段行为, 信息加工, 购前搜索, 购中选择, 购后反馈

Abstract: Algorithmic recommendation research has traditionally emphasized backend techniques, prediction accuracy, and downstream outcomes such as clicks and purchases. This article shifts attention from the algorithm itself to the recommendation explanation cues consumers actually encounter on e-commerce interfaces. We distinguish two front-end cues with different psychological meanings: similar-user cues, which justify a recommendation through the browsing, comparison, or choice behavior of consumers with similar preferences, and product-attribute cues, which justify it through attribute, functional, category, or need-based matches among products. The framework does not treat these cues as exhaustive algorithm categories or interchangeable descriptions of any recommendation outcome. Instead, it separates backend recommendation logic, front-end explanation cues, and consumer psychological processing. To identify the independent effect of explanation evidence, the proposed research design holds candidate products and their substitute/complement relationships constant while varying the evidence source highlighted in the interface.
The central innovation is an integrative process model explaining how these explanation cues shape different behaviors across three consecutive stages of the online consumer journey: pre-purchase search, purchase-stage choice, and post-purchase feedback. Similar-user cues are expected to produce lower perceived information consistency, higher perceived information certainty, and stronger perceived sociality than product-attribute cues. These differences correspond to three stage-specific mechanisms—perceived product differences, perceived product uncertainty, and perceived social connectedness to the product—forming a coherent chain from explanation cue to information processing, psychological response, and behavior.
At the pre-purchase search stage, product-attribute cues organize the recommendation set around similarity in attributes, functions, categories, or usage needs. This makes products appear more homogeneous and reduces perceived product differences, encouraging goal-directed search characterized by narrower search breadth, lower click diversity, and a more concentrated consideration set. Similar-user cues make the same set appear more open and preference-diverse because products are connected through comparable consumers rather than product similarity alone. The resulting increase in perceived product differences encourages exploratory search, broader browsing, greater click diversity, and a more diversified consideration set. Perceived product differences therefore serve as the key mediator at this stage.
At the purchase-stage choice stage, the role of explanation cues shifts from structuring the search space to reducing uncertainty about a focal option. Similar-user cues provide social proof by indicating that comparable consumers have considered or selected the recommended product. This evidence should reduce perceived product uncertainty more effectively than product-attribute cues, which mainly communicate attribute or functional fit. Lower uncertainty should increase consumers’ willingness to choose unfamiliar products while improving choice confidence, decision efficiency, and purchase conversion intention. Perceived product uncertainty is therefore proposed as the central mediator of purchase-stage decisions.
The post-purchase stage extends the framework by treating recommendation explanations as representations that can persist beyond their initial exposure. Similar-user cues can create a mental representation of the recommendation source that connects the consumer’s choice with comparable users. After product use, this representation can strengthen perceived social connectedness to the product and make feedback feel like an informational contribution to similar consumers rather than merely personal expression. The framework therefore predicts greater feedback effort and higher feedback quality, reflected in more specific, diagnostic, and distinctive reviews. The effect may persist over time and may also be reactivated at later evaluation touchpoints. Perceived social connectedness to the product is proposed as the mediating mechanism.
The model also specifies stage-specific boundary conditions. Explanation-cue effects on pre-purchase search should weaken when consumers have highly specific shopping goals because established goals constrain attention and search. Effects on purchase-stage choice should weaken as product knowledge increases because knowledgeable consumers rely more on their own evaluative structures. Effects on post-purchase feedback should weaken under high cognitive load because producing rich and diagnostic reviews requires cognitive resources. These moderators clarify when front-end explanations are most consequential.
The article contributes in three ways. First, it establishes recommendation explanation cues as independently manageable interface variables rather than treating them as proxies for backend algorithms. Second, it integrates search, choice, and feedback into a continuous multistage decision framework that distinguishes immediate processing from delayed effects and later reactivation. Third, it develops stage-specific micro-mechanisms of information processing rather than relying on a generic pathway. Managerially, the framework suggests that platforms should provide truthful, comprehensible, and stage-matched explanations: cues that guide search before purchase, diagnostic evidence that reduces uncertainty during choice, and feedback prompts that emphasize value to similar consumers after purchase. The value of recommendation explanations thus lies not in fully opening the algorithmic black box, but in shaping how consumers interpret recommendation evidence throughout the decision journey.

Key words: recommendation explanation cues, consumers’, multistage online behavior, information processing, pre-purchase search, purchase-stage choice, post-purchase feedback