TrustNet - Trust-based Moderation:

Our lives and societies are built on the foundations of trust. Trust in institutions not to swindle us, in governments and politicians not to waste our money, and, most importantly, in our personal relationships. We trust the people closest to us, and extend that trust to people whom they deem trustworthy. Importantly, we also disregard the advice from those whom have accrued our distrust, regardless of what our friends may say about them. Trust is foundational for the proper functioning of chat systems. Trust in the fact that the person you are chatting with is who they claim to be, and that they are not swapped out mid-conversation and replaced by some other intervening third-party. Trust in the permanence of actual people behind online monikers is what allows relationships, amorous and platonic, to blossom in the digital landscape. The digital infrastructure we use daily is also built on trust. Trust of the root and authoritative name servers to honestly answer domain name queries, on the signing authorities of digital certificates, in the companies that are hosting our personal correspondence and private information, and in the tamper-free distribution of software packages and binaries. In the digital sphere, however, there is a pervasive lack of the personal kind of trust we employ in our relationships==. It is, of course, possible to place trust in people and their accounts. To verify fingerprints of secure messaging services, hashes of binaries and so on. This amounts to verifying the authenticity of things—this device belongs to that person, that binary has not been exchanged in transit—but speaks little as to any actual degree of trust. ==There are few translations into the digital realm of the trust and mechanisms which we rely on day to day; that of delegating responsibilities to people that are close to us and whose abilities and confidence we treasure.

5.1 Definitions

The definition of trust this work relies on is one which has been put forward by Bhuiyan, Jøsang & Xu: Trust is the extent to which one party is willing to depend on something or somebody in a given situation with a feeling of relative security, even though negative consequences are possible. [Bhuiyan et al., 2010] Bhuiyan et al. label the above definition as decision trust. The definition beautifully captures the notion of dependence and delegation which is inherent to trust, as well as the fact that trust can be both misplaced and have unanticipated consequences. The measurement of the extent to which one party trusts another is known as a trust metric.

Trust and reputation Trust is also connected to, but separate from, reputation. Abdul-Rahman and Hailes provide the following, somewhat dry, definition of reputation, […] reputation is an expectation about an agents behavior based on information about or observations of its past behavior [Abdul-Rahman and Hailes, 2000] Reputation is an aggregate. It is an opinion on something derived from a quantity of prior opinions and judgement calls by, chiefly, other entities. Reputation can be used to inform one’s trust. If, however, you trust someone, that entails the possibility of maintaining that trust irrespective of any bad reputation the person may have accrued. Trust captures something personal, while reputation is a utilitarian value, condensed from repeated observations by the crowd regarding a particular entity. Bhuiyan et al. define the essential difference of trust and reputation in two commonplace statements, I trust you because of your good reputation. I trust you despite your bad reputation. The first shows trust being informed by someone’s accrued reputation. An unknown becoming welcomed into the fold. The second, that the person remains trusted because of prior personal knowledge; a relationship being maintained in stormy weather. Reputation and trust have their own inherent characteristics which make them suitable for different purposes. Reputation has, for instance, chiefly been used in marketplace scenarios such as ecommerce sites or ensuring the authenticity of files in fleeting peer-to-peer networks [Kamvar et al., 2003]. Whereas personal trust has so far been relatively underused in respect to computer systems.

Trust facets There are a few dimensions==, or facets, which make up that which we call trust. ==Trust is subjective==. That is, our own trust for a particular entity may differ from that of another person’s trust in the same entity. ==Trust is, to some degree, measurable==. Alice may trust both Bob and Carole, while at the same time she may trust Bob more than she trusts Carole. ==Trust is scoped==. Alice may trust Bob to take care of her plants while she is away, but she does not trust Bob for financial advice. Using the above definition of trust in combination with the just described ==facets of trust, that trust is subjective, scoped, and (to some degree) measurable, we can begin to envision a system of trust able to be mediated by computers.

Trust metrics A trust metric is defined as the way in which trust across entities is measured. The term also extends into the territory of trust propagation, or how new trust relations can be discovered given an initial set—trust propagation is discussed further in Section 5.3. Ziegler and Lausen, authors of the Appleseed trust metric which is the topic of Chapter 6, describe trust metrics in the following manner: Trust metrics compute quantitative estimates of how much trust an agent a should accord to its peer b, taking into account trust ratings from other persons on the network. [Ziegler and Lausen, 2005]

Computational Trust