文献类型：期刊文献

中文题名：生成式AI行程规划持续使用意愿的双阶段适配机制

英文题名：A Dual-stage Adaptation Mechanism of Continuous Usage Intention for Generative AI in Itinerary Planning

作者：乔向杰[1];赵子惠[1];刘丁菀[1]

第一作者：乔向杰

机构：[1]北京联合大学旅游学院,北京100101

第一机构：北京联合大学旅游学院

年份：2026

卷号：41

期号：2

起止页码：32-47

中文期刊名：旅游学刊

外文期刊名：Tourism Tribune

收录：;北大核心:【北大核心2023】；

语种：中文

中文关键词：生成式AI行程规划;持续使用意愿;功能适配性;修正可行性;人-AI协作

外文关键词：generative AI itinerary planning;continuous usage intention;functional fit;adjustment feasibility;human-AI collaboration

摘要：生成式AI在旅游行程规划中面临用户认可度高但实际使用率低的认知-行为割裂困境,传统技术采纳模型难以解释人-AI动态协作的复杂性。文章提出“功能适配性修正可行性”双阶段任务适配框架,揭示持续使用意愿的形成机制:生成阶段中,AI生成方案的质量(完整性、准确性、关联性)正向驱动功能适配性,而技术基础风险(如算法缺陷与解释缺失)有显著抑制效应,用户技术熟悉度通过认知归因优化与补偿策略机制缓冲风险负面效应;修正阶段中,操作交互的流畅性提升用户优化方案的可行性,其效应随用户微调能力增强而放大,但修正衍生的情绪损耗与认知负荷构成核心约束。双阶段框架通过任务适配度协同转化价值:功能适配性提供优化基础,修正可行性保障迭代可持续性,二者经任务适配度完全中介驱动持续使用意愿,从而揭示了技术属性需经任务效能评估方能转化为行为意愿的传导逻辑,为生成式AI采纳研究提供新范式。最终,研究提出平台技术优化、用户赋能及风险分治的协同发展路径。
Generative artificial intelligence in travel itinerary planning faces a cognitive-behavioral gap,where high user satisfaction coexists with low sustained usage rates.Traditional technology adoption models fail to explain this paradox due to their neglect of dynamic human-AI collaboration,particularly their inability to capture userscomplete decision-making chains from solution generation to multi-round adjustments.To address this issue,this study proposes a dual-stage task-fit framework of“functional fit–adjustment feasibility”,integrating theoretical constructs and empirical analysis of 605 valid questionnaires to reveal the formation mechanism of continuance intention.In the generation stage,the quality of AI-generated solutions(encompassing completeness,accuracy,and relevance)constitutes the core driver of functional fit,while technological risks(e.g.,content hallucination from algorithmic flaws,trust crises from explainability gaps)significantly inhibit it;userstechnical familiarity buffers these negative effects by forming a critical cognitive protective barrier through cognitive attribution optimization and compensatory strategy activation.In the adjustment stage,operational fluency(manifested in response agility,interface intuitiveness,and error tolerance)empowers adjustment feasibility,with its effects amplified by users'fine-tuning capabilities(e.g.,prompt refinement and feedback interpretation).However,emotional drain and cognitive load arising from revision processes emerge as substantial constraints,particularly triggering behavioral discontinuation when time costs exceed tolerance thresholds,leading to the termination of itinerary refinement processes.The two-stage synergy realizes value conversion through task fit,wherein functional fit provides the initial quality foundation for solution optimization and adjustment feasibility ensures the sustainability of iterative refinement.Together,both factors drive continuance intention through the full mediation of task fit.This mechanism clarifies that usersrecognition of technical capabilities must be translated into task-level efficacy perceptions—the assessment of“whether this tool can genuinely assist in completing ideal itinerary planning”—to bridge the cognitive-behavioral gap.The theoretical contributions manifest as three breakthroughs.First,the innovative construction of a dynamic human-AI collaboration model redefines users as active adjusters rather than passive recipients,thereby transcending the limitation of static scenarios inherent in traditional task-technology fit theory;second,the established risk-partitioning framework distinguishes between technologyinherent risks(requiring algorithmic iteration)and interaction-process risks(requiring human-centered design optimization),outlining differential mitigation paths;third,empirical validation confirms the pivotal role of task fit,revealing that technical attributes must be evaluated through the lens of taskefficacy to be transformed into behavioral intentions,thereby establishing a new paradigm for generative AI adoption research.Based on these findings,the researchers propose synergistic development pathways across three dimensions.At the technological level,platforms should construct dynamic knowledge graphs to enhance information accuracy,develop spatiotemporal logic algorithms to ensure itinerary coherence,and introduce intuitive drag-and-drop interfaces to reduce the cognitive burden of revisions.At the user level,efforts should focus on attribution training to reshape perceptions of technological risks,building a repository of compensation strategies to bolster risk-response capabilities,and providing specialized fine-tuning training to improve operational efficacy.At the regulatory level,it is advised to certify algorithm credibility to standardize foundational technological risks,set emotional load circuit-breaker standards to manage adjustment-derived risks,and establish proximity-based resource matching mechanisms to enhance decision support.This framework provides a theoretical foundation and practical guidelines for scaling generative AI in itinerary planning scenarios.Future research could explore cross-generational user comparisons,validate thresholds across diverse scenarios,examine mechanism evolution through technological iterations,and particularly investigate how multimodal interactions reshape operational fluency.