Chatbots May Foster Co-Creation in Collaborative Holiday Planning, Study Finds

Anyone who has tried to organize a holiday with a partner or a group of friends knows how quickly enthusiasm can dissolve into friction. One person wants beaches, another wants museums; one has seen a cheaper flight, another insists the hotel reviews are terrible. The information needed to make a good decision is scattered across booking sites, blogs, and messaging threads, and reaching consensus takes time, patience, and social effort. A new study published in Information Systems Frontiers suggests that artificial intelligence chatbots, when placed directly inside group conversations, may be able to ease this coordination burden and even encourage people to co-create their travel plans together rather than retreat into unilateral decision-making.

The research, conducted by Mohammad Amin Kuhail of Zayed University together with Saifeddin Alimamy, Sana Khan, and Jihene Mrabet, addresses a notable gap in the existing literature on conversational agents. Most prior work on travel chatbots has examined dyadic interactions, meaning a single user conversing one-on-one with an AI assistant. The reality of holiday planning, however, is polyadic: multiple people negotiating within a shared chatroom, each with their own preferences, constraints, and personalities. Whether a chatbot can be useful in that messier, multi-party setting, and what psychological factors determine whether group members actually embrace it, had remained largely unexplored.

To investigate the question, the team grounded their study in expectancy-value theory, a well-established motivational framework that explains behavior through two intertwined components: people’s beliefs about how likely an action is to produce desired outcomes, and the subjective value they attach to those outcomes. In this context, the researchers modeled travelers’ reasons for using a chatbot, including the quality of interaction, the degree of customization, the quality of information provided, and the chatbot’s capacity for perspective-taking. They also modeled reasons against use, namely perceived low-quality advice and privacy concerns. Crucially, these factors were treated as formative predictors of attitudes toward the chatbot, which in turn were expected to influence travelers’ willingness to co-create holiday plans with the technology.

The experimental design was unusually concrete for research in this area. Rather than asking participants to imagine using a chatbot, the researchers built a live multi-user chatroom that integrated a working AI-powered chatbot. One hundred and twelve participants were organized into fifty-six pairs, and each pair was tasked with collaboratively planning a holiday inside the chatroom while the chatbot generated options, tailored recommendations, and attempted to accommodate both members’ preferences. After the planning session, participants reported their perceptions through questionnaires, allowing the team to measure how their experiences during real collaboration shaped their attitudes and intentions.

The findings, analyzed using structural equation modeling techniques appropriate for formative constructs, paint a nuanced picture of when and why people embrace AI assistance in group decision-making. Interaction quality, customization, information quality, and especially perspective-taking all strengthened participants’ reasons for using the chatbot. Perspective-taking emerged as a particularly powerful driver: when the chatbot appeared to recognize and represent each traveler’s viewpoint within the group discussion, users’ motivation to rely on it increased markedly, which in turn enhanced their overall attitudes and raised their willingness to engage in collaborative planning. This result makes intuitive sense in a polyadic context. A tool that simply answers questions is useful, but a tool that seems to understand that one partner cares about budget while the other cares about nightlife is performing a genuinely social function, acting as a kind of neutral broker among competing human interests.

On the negative side of the ledger, privacy concerns and perceptions of low-quality advice did strengthen participants’ reasons against using the chatbot. Sharing preferences, budgets, and travel habits with an AI system embedded in a group conversation can feel intrusive, and the well-documented personalization-privacy paradox suggests people are often ambivalent about trading data for convenience. Yet the study found something surprising: these negative reasons, while real, did not significantly diminish participants’ attitudes toward the chatbot. In other words, the motivational pull of a competent, empathetic assistant appears strong enough that concerns about privacy and occasional errors register as reasons for hesitation without substantially souring users’ overall evaluations. The positive experience of successful collaboration seems to outweigh the abstract worry about what happens to one’s data.

The researchers also examined the role of personal values, and found that they further amplified reasons for using the chatbot. Travelers whose broader value orientations aligned with collaboration and exploration were more responsive to the chatbot’s capabilities, suggesting that individual differences shape not just whether people use AI tools but how much benefit they perceive in them. This finding connects expectancy-value theory to a longer tradition in tourism research showing that personal values steer travel decisions, and it hints that the same AI assistant may be experienced very differently by different users depending on what they bring to the encounter.

Beyond its empirical results, the study makes a theoretical contribution by extending expectancy-value theory into the domain of human-AI group collaboration. EVT was developed to explain achievement motivation in educational settings, and its application to polyadic chatbot use demonstrates the framework’s flexibility. The authors argue that reasons for and against using a technology are not merely symmetric opposites; in their data, positive reasons translated into favorable attitudes and co-creation intentions, while negative reasons created resistance without necessarily destroying goodwill. This asymmetry has practical implications for how designers and companies should prioritize their efforts.

Those design implications are perhaps the most immediately actionable part of the work. For developers building travel-planning chatbots for groups, the message is clear: invest in perspective-taking. A chatbot that explicitly acknowledges each member’s stated preferences, summarizes points of agreement and disagreement, and proposes compromises that reflect the group’s internal logic is likely to earn trust and encourage shared planning. Customization matters too, but in the polyadic case it means adapting to multiple users simultaneously rather than a single profile. The findings also caution designers to manage privacy transparently, even if privacy concerns alone did not overturn positive attitudes, because they still feed the reservoir of reasons against adoption that could become decisive in other contexts or populations.

The study also resonates with a growing body of research on AI in collaborative work. Earlier studies have shown that chatbots can improve efficiency and participation in group chat discussions, moderate deliberative conversations, and support emotion management in distributed teams. The present research extends this line into leisure and tourism, a domain where the stakes are emotional and social rather than strictly professional. Holidays are, for many people, a key source of shared memory and relationship bonding, and the planning process itself is a rehearsal of the trip’s social dynamics. If an AI assistant can reduce the conflict and information overload of that rehearsal, it may indirectly improve the experience of the holiday itself.

The authors acknowledge limitations typical of an experimental study of this scale. Fifty-six pairs planning a hypothetical holiday in a controlled chatroom differ from real friend groups with long histories and high emotional stakes, and the one-session design captures initial reactions rather than sustained use. Still, the live, multi-user setup represents a meaningful advance over hypothetical vignettes, and the sample’s task engagement, evidenced by analysis of chatbot and user response themes reported in the study’s appendices, lends credibility to the self-reported measures. The datasets generated during the study are available from the corresponding author upon reasonable request, and the research was approved by Zayed University’s Research Ethics Committee and funded through a university grant.

As generative AI assistants become woven into everyday messaging platforms, questions about how they behave in group contexts are becoming urgent. This study offers one of the first rigorous, theory-driven answers for the travel domain: chatbots can indeed foster co-creation in collaborative holiday planning, provided they interact well, tailor their recommendations to multiple people, supply high-quality information, and, above all, demonstrate that they can see the situation from each traveler’s point of view. The friend who never quite agrees on the destination may not be persuaded by another human voice in the thread, but an AI that patiently reflects everyone’s preferences back at the group may be exactly the mediator modern travel planning needs.