The Web Science Trust

Webs of Trust and Knowledge: Knowing and Trusting in the World Wide Web

Simon, Judith (2009) Webs of Trust and Knowledge: Knowing and Trusting in the World Wide Web. In: Proceedings of the WebSci'09: Society On-Line, 18-20 March 2009, Athens, Greece. (In Press)

[img] MS Word (preprint) - Repository staff only
[img] MS Word (Poster Description)


A glut of new social software, web.2.0 or even web3.0 applications has come on the market in the last years. Despite the breadth and fuzziness of these terms, the quintessence of these technologies is that social networks or trust relationships between users of a system are associated with the evaluation of the quality of knowledge organization, sharing and retrieval. Based on such observations, I will examine the relationship between knowledge and trust in the WWW from an epistemological point of view. The goal of this paper is twofold: Taking into account the importance of software applications for scientific as well as other forms of reasoning, it should be evident that a thorough analysis of these developments cannot be neglected should philosophical theories of knowledge be relevant and adequate. Moreover and in turn, I am convinced that software development could be improved if an epistemologically informed notion of trust that goes beyond mere track record accounts or game theoretical reductions was employed. The notion of trust has received growing attention in epistemology, i.e. the philosophical discipline concerned with the process of knowing and criteria for knowledge in the last twenty years. An epistemological account of trust would have to analyze the function of trust for knowledge and/or vice versa. In his seminal paper “The role of trust in knowledge” (1991), John Hardwig’s tries to assess the function of trust for knowledge creation in science and links the notion of trust to the question of testimony as one of the major problems in (social) epistemology. He states that in classical epistemology, knowledge and trust are normally conceptualized antithetically: ”[w]e can not know by trusting in the opinion of others; we may have to trust those opinions when we do not know” (Hardwig 1991, p. 693). However, “[m]odern knowers”, he argues “[…] cannot be independent and self-reliant, not even in their own fields of specialization” (Hardwig 1991, p.693). His analysis departs from the observation that the majority of research is nowadays conducted in teams and he presents two examples of major scientific achievements in physics and mathematics as case studies in support of his claims. Cooperation in science is supposedly needed to overcome time restrictions on the one hand and to handle the rising specialization in science on the other. As a consequence, in scientific co-operations scientists have to trust the competency and the honesty of their colleagues, because due to high specialization, they do not only lack the time to perform every subtask of their research on their own, but mostly they also lack the necessary expertise in the respective area of research. In order to successfully operate in science, scientists need to assess their colleagues not only epistemically but also morally. To put it in a nutshell, Hardwig (1991) argues that trust is epistemologically even more fundamental than empirical data or logical argumentation, because one needs to trust these pieces of evidence and their providers to actually use them at all. Thus, the trustworthiness of members of epistemic communities is fundamental to all scientific endeavors and represents the groundwork of our (scientific) knowledge. Let’s turn to the notion of trust in ICT. Being aware that trust in software is another central topic in web science, in the following I will focus exclusively on trust in other people via software and use trust-aware recommender systems (RS) to exemplify my claims. RSs in general are systems that suggest new items to users, which he or she might like and are often used in commercial websites such as In an experimental analysis of data obtained from an internet website deploying such RSs, Massa and Bhattachasjee (2004) demonstrated that classical RS techniques have several shortcomings, such as sparseness (sparsity of useful information for existing users), cold start (difficulty to generate recommendations for users’ who have just registered) and vulnerability of system correctness to attacks. In the following I will concentrate on the so-called cold start problem. “Bootstrapping” in RSs is a term for procedures to meet the cold start problem. When a new user enters a system the system does not “know” anything about this new user and this ignorance makes it difficult to generate appropriate recommendations for her. To counteract this problem, traditionally new users were asked to rate a few items so that the system can “learn” something about the user in order to provide personalized information on interesting items for her. However, especially in large databases necessary correlations are scarce and thus, this procedure often turns out to be quite ineffective. In consequence, Massa & Bhattachasjee (2004) develop an algorithm for “Trust-aware Recommender Systems” and argue that the before mentioned problems “can effectively be solved by incorporating a notion of trust between users into the base CF [A.N. collaborative filtering] system” (Massa & Bhattachasjee 2004). The difference between traditional RSs and trust-aware RSs is quite simple: “[w]hile traditional RSs exploit only ratings provided by users about items, Trust-aware Recommender Systems let the user express also trust statements, i.e. their subjective opinions about the usefulness of other users” (Massa & Avesani 2006). This seemingly minor change proves to be highly effective to remedy the shortcomings of traditional RSs especially with respect to the cold start problem because “it is able to exploit trust propagation over the trust network by means of a trust metric” (Massa & Avesani 2006). Trust, it seems, is indispensable for knowledge creation in science and everyday life. And since the prevalence and use of software applications such as RSs, which technologically implement notions of trust, will probably rise in the next years, it will even gain importance. Assessing the quality of information, deciding whom to trust and whom to distrust is not limited to information obtained on the WWW. However, it becomes all the more obvious in an environment in which information can be exchanged with high speed over long distances, enormously increasing the amount of interactions with people we do not know personally but whom we have to trust – or decide to distrust – nonetheless. Taking these developments into account, a thorough analysis of the relationship between trust and knowledge should be indispensable – for epistemology, web science and software development. Literature Hardwig, J. (1991). "The role of trust in knowledge." The Journal of Philosophy 88(12): 693-708. Massa,P. & Bhattacharjee, B. (2004). Using Trust in Recommender Systems: an Experimental Analysis. Proceedings of iTrust2004, International Conference. Massa, P. and P. Avesani (2006). Trust-aware Bootstrapping of Recommender Systems. Proceedings of ECAI, Riva del Garda, Italy.

Item Type:Conference or Workshop Item (Poster)
Uncontrolled Keywords:Trust, Knowledge, Social Relationships, Networks, Epistemology
Subjects:Web Science Events > Web Science 2009
ID Code:204
Deposited By: W S T Administrator
Deposited On:24 Jan 2009 08:45
Last Modified:25 Oct 2011 16:11

Repository Staff Only: item control page

EPrints Logo
Web Science Repository is powered by EPrints 3 which is developed by the School of Electronics and Computer Science at the University of Southampton. More information and software credits.