汽車共乘可算是友善社會環境的永續機制，因為可以藉此減少碳排放量、降低交通阻塞的機率、以及停車空間的需求，還有節省燃油成本等。然而，由於某些原因，例如，可能因為不熟悉共乘夥伴、彼此缺乏信任感、以及處於狹小汽車空間感覺不舒適等，讓人們參與的意願大大降低。因此，本研究提出一個以社群互動為基礎的共乘推薦模式，希望透過以人際關係權重高低做為信任因子以促使共乘行為，並且提供使用者相似路徑資訊，以改善共乘系統的可用性與易用性。關於人際關係權重的計算，我們使用語意相似度的概念來設計屬性相似度，同時考慮線上社群的互動頻率以及內在感受的關係因子。對於陌生的共乘者，透過共同的朋友產生間接關係，以提升信任因子。在共乘系統部份，我們分析使用者的路徑資訊，找尋擁有相似路徑的使用者。本研究所定義的人際關係權重，可以使用關係網格的伴隨係數得到合理的驗證。最後，我們將使用者的人際關係權重與可以共乘的路徑，透過天際線運算，產生具適地性服務的共乘推薦對象。本研究使用一個小型範例來說明整體模型的可行性，對於共乘服務平台或服務商來說，未來將可以從會員資料中，收集其參與共乘的歷史記錄，進行整體大數據分析，以提供更佳的共乘服務，促進更多人的共乘意願。

Purpose - Carpooling is an environmental-friendly and sustainable way to reduce carbon emissions, traffic congestion, parking spaces and fuel costs for travelers. However, due to some reasons, (e.g., people may not be familiar with each other, lack of trust, and feel uncomfortable in a small space inside the car), the willingness of participating in carpooling still needs to be enticed. In this paper, we propose a carpooling recommendation model based on the interpersonal relationships derived from social networks. Design/methodology/approach - We believe the interpersonal trust is a critical factor to inspire the ride behavior in a carpooling recommendation system. By using the concept of hierarchical semantic network for calculating the attribute weights based on the frequency of interaction in online social networks, we can reasonably evaluate interpersonal relationships. For route analysis, we use Web GIS and spatial database methods to search similar travel routes by the paths generated from users' travel profiles, and then we reduce the obtained similar routes and store them in the spatial database. Findings - By taking into account recommendations from common friends, new relationships can be established, and the degree of trust can be promoted, which implies the trust weights of interpersonal relationships can be rationally verified. Our approach collects the similar itineraries and generates the recommendation result as a location-based service by using the skyline operation. Research limitations/implications - The experiment considers only a small set of participants in a carpooling scenario to show the feasibility and illustrate the concept of our model. Practical implications - We suggest that carpooling platform/service providers may consider collecting their big data from their membership and daily riding profile to apply their model for offering better service and engendering more willingness. Originality/value - We have designed a new approach to promote the willingness of carpooling based on interpersonal trust relationships derived from social networks. Our model enhances the degree of trust for users and in turn encourages more ride-share behaviors, which creates a core value of carpooling in terms of the user experience and trust relationships.