1 Introduction Volunteered Geographic Information (VGI) [5] is an approach to crowdsource information about geospatial features around us. Recently, VGI is gaining increasing attention and (web) services relying on it are becoming ubiquitous. Thanks to the greater engagement of contributors, coverage and precision of VGI data is quickly approaching the level granted within professional Geographic Information Systems (GIS), as shown by several comparative studies with official national datasets [6, 14]. However, while in professional GIS data quality is granted by certified authorities, the assessment of VGI data quality remains an open challenge [11, 7, 13]. A basic method to assess the quality of a VGI dataset consists in comparing it against a professionally–generated ground–truth dataset. However, this approach suffers major drawbacks. First, it requires access to professional datasets that, in the best case, is expensive and, in the worst case, is not possible at all. Moreover, it does not provide a quality assessment procedure that is universally valid (e.g., think of cases where a ground–truth dataset is not available at all). As suggested in [1], a different approach consists in assessing the quality of VGI data through a proxy measure: trustworthiness. Trustworthiness is defined [12] as a “bet about the future contingent action of others”. In this sense, trustworthiness is strictly related to the concept of (others’) reputation. This paper presents ongoing work on an evaluation model of volunteers’ reputation and data trustworthiness that derives the coveted information from VGI data, without requiring a comparison with external sources. We draw inspiration from the work in [7] and extend it by (i) relating data trustworthiness and user reputation and (ii) accounting for the relevance of data editing. (iii) Finally, our model accounts for atomic editing operations, rather than for composite editing patterns. 2 Related work Quality assessment of VGI is still rather new research topic. As suggested by Flanagin and Metzger [3], there is a critical need for identifying methods and techniques to evaluate the VGI quality. One rather standard approach is to compare VGI datasets to authoritative, ground-truth datasets, as done, for example, by Mooney, Corcoran and Winstanley [11] who analysed characteristics of polygons contributed by OpenStreetMap users. According to the results, volunteers seem to be able to more easily trace outlines of water features compared to forest features. A different approach is undertaken by Bishr and Janowicz [1] that promote the notion of informational trust to be used as a proxy measure for quality. Their proposal was one of the first examples for using trustworthiness for quality assessment in VGI. Keßler, Trame and Kauppinen apply the Bishr and Janowicz proposal to use trust as a proxy measure, in [8] they used trust and provenance for studying contribution patterns in the case of OSM. An extension of this work is [7] in which, Keßler and De Groot, provided a few indicators that influence trust and that were basically derived from data provenance. The work presented in [7] has the purpose to build a model that depends mostly on provenance data; so that there is no need of a reference comparison dataset; trustworthiness is associated to each feature and represents the proxy value of data quality. In this work Keßler introduce the user reputation issue and leave it for future refinement. Keßler used five parameters for trustworthiness evaluation. (1) Versions, they are an important source of provenance information. (2) Users, the higher is the number of users that works on a feature the higher is the trustworthiness value. (3) Confirmations, all the revisions that were made in the neighbourhood of a feature are taken into account. (4) Tag corrections, a semantic change over a feature decreases the feature trustworthiness. (5) Rollbacks, restoring a feature’s previous state also decreases the feature trustworthiness. VGI Edit History Reveals Data Trustworthiness and User Reputation Fausto D’Antonio University of L’Aquila Via G. Gronchi 18, 67100 L’Aquila, Italy fausto.dantonio@gmail.com Paolo Fogliaroni Vienna University of Technology Gusshausstr. 27-29, 1040 Vienna, Austria paolo@geoinfo.tuwien.ac.at Tomi Kauppinen Aalto University School of Science Department of Media Technology FI-00076 Aalto, Finland tomi.kauppinen@aalto.fi Abstract Volunteered Geographic Information (VGI) is an approach to crowdsource information about geospatial features around us. People around the world are engaged with typing in their observations about the world (like locations of shops, cafeterias), or to semi-automatically gather them with mobile devices (like hiking paths or roads). In this process people might make mistakes, for instance assign misleading tags to features or provide over simplistic boundaries for features. In this paper we study what kinds of things might contribute to assess trustworthiness of data, and reputation of contributors for VGI. We present a model for analysing the different factors, and a method for automatically creating the trust and reputation scores.