Elsevier

Decision Support Systems

Volume 115, November 2018, Pages 105-116
Decision Support Systems

A computational framework for understanding antecedents of guests' perceived trust towards hosts on Airbnb

https://doi.org/10.1016/j.dss.2018.10.002Get rights and content

Highlights

  • A computational framework is proposed to understand antecedents of perceived trust.

  • Reputation plays the most important role in affecting perceived trust.

  • Accumulating reviews and striving for superhost badge almost mean half a success.

  • Focusing on interactions and services in self-description would be more helpful.

  • Positive sentiment in self-description or profile photo can help build trust.

Abstract

Several studies have researched the antecedents influencing the perceived trust of guests towards hosts on Airbnb typically relying on survey data. However, the contribution of these antecedents to trust building in a practical context remains unclear. To fill this gap, we focused on the antecedents within the manageable information about hosts and proposed a computational framework for understanding the antecedents influencing perceived trust. Specifically, perceived trust was proxied by the growth rate of bookings and the validity of the proxy method was proved through comparing with human labeled data. From the snapshot information about hosts, the antecedents were quantified through text mining and face recognition methods. The least square regression was applied to analyze and compare the influence of these antecedents. We found that the contribution of reputation is not less than the summation of all the other antecedents. Additionally, in terms of self-descriptions, it is worthwhile to pay more attention to interactions and services. Expressing positive sentiment in either self-descriptions or profile photo is also helpful. The response behavior pattern and the number of verifications also matter. At last, several effective trust prediction models were built by using deep neural network and the ensemble method. The findings shed light on the working of the antecedents in trust formation and can provide instructions for the transaction partners, designers and managers of online services in the sharing economy.

Introduction

In the sharing economy, the trust between providers and consumers forms the basis of a successful resource-sharing transaction. Airbnb is a good example, acting as an online lodging marketplace for short-term peer-to-peer rentals, where guests seek low-cost accommodations and direct interactions with the local community [1]. The necessity of face-to-face interactions implies that the hosts may experience severe damage to their properties or theft of personal belongings and the guests may also be faced with risk of unreliable hosts or even personal security [2,3]. What's more, the quality of the accommodation service is highly dependent upon the hosts. Therefore, trust, as an efficient mechanism for lowering transaction costs of social exchange [4], is absolutely necessary in the sharing economy.

From the guests' perspective on Airbnb, the main sources of information for inferring a host's trustworthiness is the online profile provided by the platform, which contains important trust-related cues, such as reviews, rating score, verifications, self-descriptions, profile photos, and so on. According to the characterization of Mayer et al. [5], guests can assess the trustworthiness of hosts with the criteria of ability, integrity, and benevolence, based on the information provided. Indeed, a lot of studies have been performed according to the subjective measurement criteria. However, the subjective measurement methods are not applicable to large scale data and research based on actual trusting behavior in the sharing economy remains scarce, although it can help show the actual working of trust mechanisms. Therefore, it is necessary to develop a computation method to measure perceived trust based on real world data.

For convenience, we will use “perceived trust” to represent “Airbnb guests' perceived trust towards hosts” in the following sections. It should be noted that perceived trust is different to trustworthiness. In detail, perceived trust can be considered as the reflected trustworthiness of the trustees and trustworthiness is subjectively entertained in the judgment of the trustors [6]. From the microscopic perspective, a particular trustee's trustworthiness can be taken as a constant and the perceived trust towards him may differ among trustors. But from the macroscopic perspective, it is generally acknowledged that the more guests trust a trustee, the more trustworthy the trustee probably is. Although trust is not a behavior (e.g. cooperation) or a choice (e.g. take a risk), it is an underlying psychological condition that can cause such actions [7]. And when trusting behavior occurs, it generally means that trust has been built [8]. Accordingly, we assume that perceived trust on Airbnb can be measured based on the accumulated trusting behavior of guests.

Besides of ability, integrity, and benevolence, Sztompka [6] claims that people employ three other criteria in estimating the trustworthiness of their transactional partners: reputation, performance, and appearance. The majority of information online about hosts, presented by Airbnb, can be generally categorized into the three criteria. Thus, by managing the antecedents such as self-description, profile photo, service behavior, hosts can accumulate reputation and earn trust from the guests.

At present, reputation is often regarded as the panacea for establishing trust [2], and a lot of research has been devoted to understanding the inner workings of reputation mechanisms [9,10]. However, trust in the sharing economy is much more complex and extends far beyond reputation [2]. In fact, Airbnb's reciprocal review system enables both hosts and guests to review one another and the reviews tend to comprise a restricted set of highly positive commentary [11]. Except for the reputation system, the impact of linguistic features within self-descriptions and personal photos on trust has also been investigated by manually analyzing limited survey data [[12], [13], [14], [15], [16], [17]]. But, majority of the studies are focusing on isolated antecedents in fabricated settings, and the contribution of these antecedents in real-world trust related behavior remains unclear. Consequently, from the perspective of hosts and platform managers, it is important to gain better understanding about the contribution of the aforementioned antecedents. Haas and Deseran [18] assert that the transaction partners have the burden of not only creating trust but also maintaining it and this process involves the duty of presenting themselves as trustworthy persons. Meanwhile, the self theory suggests that people are constantly engaged in managing and controlling the impressions they make on others to attain their goals [19]. Therefore, we contend that it would be valuable to use a more complete framework that incorporates all possible antecedents within hosts' online information to study the full spectrum of trust-building mechanisms.

In this paper, we propose a computational framework for understanding antecedents of guests' perceived trust based on real-world trust related behavior data of Airbnb. For each host, the perceived trust was proxied by the growth rate of bookings in three months. Then, the antecedents were extracted and quantified from the snapshot of host information. Regression analysis on datasets with both human labeled perceived trust and quantified perceived trust proves the validity of the proxy variable. And our study generated several interesting findings. Among all the antecedents, reputation plays the most important role in building trust, while the rating system was proved to be a failure. In respect to self-descriptions, it is worthwhile to pay more attention to guest's concern, such as interactions and services and use positive words or phrases. Quick response speed and high response rate can represent a hospitable attitude, which can also significantly improve perceived trust. Additionally, more verifications and photos with positive facial expression are also helpful. Based on these antecedents, four predictive models were proposed to identify hosts with higher/lower perceived trust. The findings provide the hosts with a comprehensive set of instruction for presenting themselves effectively and further help earn trust of guests. According to the antecedents influencing guests' perceived trust, the service platforms can focus on building a better trust evaluation system.

The rest of the paper is organized as follows. Section 2 presents related work and relevant theory. In Section 3, the experimental data and methods are introduced. Section 4 introduces the data analysis, empirical result and prediction model. The final section provides a discussion of the findings and concludes with limitations and implications of this study.

Section snippets

Trust in sharing economy

Although trust has been defined in many different ways, a widely held definition of trust is as follows: trust is a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behavior of another [7,20]. Across disciplines, there is agreement on the conditions that must exist for trust to arise: risk and interdependence. Risk is the perceived probability of loss as interpreted by a decision maker [21]. Interdependence means that the

Datasets and methods

The computational framework for understanding antecedents influencing perceived trust is outlined in Fig. 1. The framework mainly consists of snapshot data processing, perceived trust quantification, feature engineering, regression analysis, comparison of feature contribution and prediction model. In general, we firstly cleaned and aligned the series data. Then, perceived trust was proxied by the growth rate of bookings. Thirdly, seven group of features, i.e., reputation, response behavior

Data analysis and results

After the feature engineering, the antecedents containing 11 variables were obtained from the dataset with 3801 listings and the summary statistics of these variables can be seen in Table 3. In the following section, we will first prove the validity of the proxy method of perceived trust by comparing with human labeled data. Then we will examine how these antecedents corporately influence the perceived trust through the multivariate linear regression by ordinary least squares (OLS). Lastly,

Discussion and conclusion

A lot of researchers have focused on the antecedents influencing trust formation, such as profile photos [12,15,17], linguistic features [13,14], reputation [10,29]. However, these antecedents are usually analyzed in isolation by using questionnaires within fabricating settings and the role that different antecedents play in trust building within the real world remains unclear. This paper proposed a computational framework for understanding antecedents of perceived trust based on real-life data

Acknowledgments

This work was supported by a grant from the National Social Science Foundation of China [No. 17AGL026].

Le Zhang is a Ph.D. candidate in the School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing, China. Her research interests include trust in the sharing economy, social computing, natural language processing.

References (60)

  • N. Hu et al.

    Manipulation of online reviews: an analysis of ratings, readability, and sentiments

    Decision Support Systems

    (2012)
  • F. Li

    Annual report readability, current earnings, and earnings persistence

    Journal of Accounting and Economics

    (2008)
  • Z. Liu et al.

    Identifying and predicting the desire to help in social question and answering

    Information Processing & Management

    (2017)
  • D. Guttentag

    Airbnb: disruptive innovation and the rise of an informal tourism accommodation sector

    Current Issues in Tourism

    (2015)
  • M. Huurne et al.

    Antecedents of trust in the sharing economy: a systematic review

    Journal of Consumer Behaviour

    (2017)
  • T.A. Weber

    Intermediation in a sharing economy: insurance, moral hazard, and rent extraction

    Journal of Management Information Systems

    (2014)
  • D.H. McKnight et al.

    What trust means in e-commerce customer relationships: an interdisciplinary conceptual typology

    International Journal of Electronic Commerce

    (2001)
  • R.C. Mayer et al.

    An integrative model of organizational trust

    Academy of Management Review

    (1995)
  • P. Sztompka

    Trust: A Sociological Theory

    (1999)
  • D.M. Rousseau et al.

    Not so different after all: a cross-discipline view of trust

    Academy of Management Review

    (1998)
  • W.B. Pearce

    Trust in interpersonal communication

    Speech Monographs

    (1974)
  • J. Bridges et al.

    If nearly all airbnb reviews are positive, does that make them meaningless?

    Current Issues in Tourism

    (2016)
  • F. Hawlitschek et al.

    Trust in the sharing economy: an experimental framework

  • X. Ma et al.

    A computational approach to perceived trustworthiness of airbnb host profiles

  • X. Ma et al.

    Self-disclosure and perceived trustworthiness of airbnb host profiles

  • D.F. Haas et al.

    Trust and symbolic exchange

    Social Psychology Quarterly

    (1981)
  • E. Goffman

    The presentation of self in everyday life

    American Journal of Sociology

    (1949)
  • A.C. Costa et al.

    Trust in work teams: an integrative review, multilevel model, and future directions

    Journal of Organizational Behavior

    (2017)
  • T.H. Chiles et al.

    Integrating variable risk preferences, trust, and transaction cost economics

    The Academy of Management Review

    (1996)
  • R. Botsman et al.

    What's Mine is Yours: The Rise of Collaborative Consumption

    (2010)
  • Cited by (94)

    View all citing articles on Scopus

    Le Zhang is a Ph.D. candidate in the School of Economics and Management, Beijing University of Posts and Telecommunications, Beijing, China. Her research interests include trust in the sharing economy, social computing, natural language processing.

    Qiang Yan is a professor in the School of Economics and Management at Beijing University of Posts and Telecommunications. His researches focus on E-commerce and information systems. He has published more than 90 papers in peer-reviewed journals.

    Leihan Zhang is a research associate in the Institute of Computer Science and Technology, Peking University, Beijing, China. His research interests include data mining, complex network, and natural language processing.

    View full text