Abstract
This research introduces a novel method for uncovering potential causal relationships in tourism literature through artificial
intelligence (AI)-based counterfactual reasoning and big data. Tourism generates massive volumes of device, transaction, and
user-generated data, and these can be leveraged using AI algorithms to better understand tourism-related social phenomena
(Park, Xu, Jiang, Chen, & Huang, 2020). Existing tourism studies have used deductive, fuzzy, inductive, and transductive AI models
(Cevikalp & Franc, 2017) to extract insights from big data, but these often fail to capture potential causal effects (Guidotti, 2022),
which is problematic for two reasons. First, decision-making by tourism stakeholders cannot be improved if AI models mainly rely
on spurious correlations (Law & Li, 2007). Second, the failure of capturing potential causal effects in big data diminishes its perceived value for both tourism scholars and practitioners.
To better capture causality, extant literature has recommended experimental designs (Viglia & Dolnicar, 2020). Although
highly effective, experimental designs rely heavily on previously established theories (Sun, Law, & Zhang, 2020). For example, research first introduces random utility theory, theorizing rational decision-making of tourists' hotel and destination choices, to then
conduct experiments (Zhang, Grisolía, & Lane, 2023). This reliance may impede the discovery of potential causal relationships that
sit beyond established theories. Experiments also have practical limitations (see Podsakoff & Podsakoff, 2019), and cannot combine multiple concepts and then control for all possible factors that may influence the concepts. The simpler the experimental design, the easier it is to establish causality. The availability of big data can overcome several limitations inherent in experimental designs. This research outlines an agenda that explains how tourism research can leverage big data and counterfactual reasoning
AI algorithms to determine potential causal effects.
intelligence (AI)-based counterfactual reasoning and big data. Tourism generates massive volumes of device, transaction, and
user-generated data, and these can be leveraged using AI algorithms to better understand tourism-related social phenomena
(Park, Xu, Jiang, Chen, & Huang, 2020). Existing tourism studies have used deductive, fuzzy, inductive, and transductive AI models
(Cevikalp & Franc, 2017) to extract insights from big data, but these often fail to capture potential causal effects (Guidotti, 2022),
which is problematic for two reasons. First, decision-making by tourism stakeholders cannot be improved if AI models mainly rely
on spurious correlations (Law & Li, 2007). Second, the failure of capturing potential causal effects in big data diminishes its perceived value for both tourism scholars and practitioners.
To better capture causality, extant literature has recommended experimental designs (Viglia & Dolnicar, 2020). Although
highly effective, experimental designs rely heavily on previously established theories (Sun, Law, & Zhang, 2020). For example, research first introduces random utility theory, theorizing rational decision-making of tourists' hotel and destination choices, to then
conduct experiments (Zhang, Grisolía, & Lane, 2023). This reliance may impede the discovery of potential causal relationships that
sit beyond established theories. Experiments also have practical limitations (see Podsakoff & Podsakoff, 2019), and cannot combine multiple concepts and then control for all possible factors that may influence the concepts. The simpler the experimental design, the easier it is to establish causality. The availability of big data can overcome several limitations inherent in experimental designs. This research outlines an agenda that explains how tourism research can leverage big data and counterfactual reasoning
AI algorithms to determine potential causal effects.
| Original language | English |
|---|---|
| Article number | 103617 |
| Number of pages | 4 |
| Journal | Annals of Tourism Research |
| Volume | 101 |
| DOIs | |
| Publication status | Published - Jul 2023 |