Utilising machine learning to investigate actor engagement in the sharing economy from a cross-cultural perspective

Abstract

PurposeRecent literature on customer engagement has introduced the concept of “actor engagement,” which serves as the foundation for this study. The study aims to investigate the formation of engagement and engagement's impact on the performance of sharing economy platforms in an international context.Design/methodology/approachThe study analyses unstructured data from 145,434 service providers and 1,703,266 customers on Airbnb across seven countries (USA, Canada, United Kingdom, Australia, South Africa, China and Singapore). Machine learning techniques are used to measure actor engagement, and the research model is tested using structural equation modelling (SEM).FindingsThe findings suggest that actor engagement, encompassing the reciprocal relationship between customer engagement and service provider engagement, has a significant impact on platform performance. The moderator analysis highlights the role of cultural differences in the relationship between customer engagement and service provider engagement and between actor engagement and platform performance. Specifically, the study reveals that actor engagement exhibits a more pronounced impact on platform performance in Western countries (such as the USA, Australia and the UK), compared to Eastern countries (such as China and Singapore).Research limitations/implicationsThe analysis of the conceptual model is based on the utilisation of behavioural data obtained from the Airbnb website. Due to the nature of the available data, proxies are employed as measures for variables such as platform performance.Originality/valueThis research is amongst the first to provide empirical evidence for actor engagement formation and the function's role in platform performance in the sharing economy. The global nature of Airbnb as a platform facilitates the investigation of country-level factors, specifically cultural values, across seven diverse countries and highlight differences from business to customer (B2C) business models.

Unveiling the dark side of influencer marketing: how social media influencers (human vs virtual) diminish followers’ well-being

“…Therefore, instead of using LIWC to measure followers' negative emotions and relied on an approach developed by Mohammad and Bravo-Marquez (2017) to measure anger, anxiety and sadness. Based on this approach and to classify comments based on specific emotions, we first removed non-English responses (Barari et al, 2023). We then enlisted 200 raters from Amazon MTurk to create a training dataset by providing them with definitions of each emotion and a sample emotion.…”

“…Based on the majority opinions of the raters (i.e. two out of three), we created a 2,000-comment training dataset (Barari et al ., 2023). Naive Bayes and Support Vector Machine algorithms were used to create a baseline for categorizing each comment into one of three negative emotions.…”