Enhancing menu upselling through human and robotic recommenders: the role of source and message credibility
研究了机器人餐厅中推荐者类型（人类、人形机器人、非人形机器人）和推荐策略（专家、本地、增长）如何影响顾客对来源和信息可信度的评价及推荐接受意愿，发现人类服务员最有效，并识别出两种最优配置。
Purpose This study aims to examine how recommendation heuristics in menu upselling, including recommender type and recommendation strategy, impact customers’ source and message credibility evaluations and recommendation acceptance in robot restaurant settings. Design/methodology/approach An online scenario-based experimental survey design was adopted. Respondents were randomly assigned to one of nine conditions, in a 3 (recommendation strategy: expert, local and growth) x 3 (recommender type: human, humanoid robot and nonhumanoid robot) between-subject factorial design. The study model was tested via structural equation modeling (SEM), utilizing data collected from 435 restaurant customers with previous service robot experience. Findings Through an extended source-message credibility model, the study revealed that human servers remained most credible and effective in menu recommendation task, followed by humanoid and nonhumanoid robots. Moreover, two optimal heuristics configurations were identified, which were human servers utilizing expert recommendation and humanoid robots utilizing growth recommendation. The authors also found that source and message credibility significantly influenced customers’ willingness to accept menu recommendations at robot restaurants. Practical implications The findings could guide technology vendors in optimizing robot design and communication capability, while robot restaurant managers may leverage this insight for strategic human-robot task allocation that enhances menu upselling effectiveness. Originality/value This research’s novelty lies in the integration of heuristics processing and dual source-message credibility, which delineated the distinct roles of human and robot servers in menu recommendation context. To the best of the authors’ knowledge, it is the first paper to propose dynamic recommendation strategies, tailored to the unique strengths of human and robot recommenders.