A lot of attention is currently given to the problem of trust on the internet.¹ Trust is for example an essential notion in the vision of the semantic web. Philosophers have also commented on this issue and argued that trust on the web is problematic in different ways. Hardin (2004), for example, argues that it is unlikely that trust will develop in relationships on the internet because of the absence of continuous face to face communication. Pettit (2005) makes similar claims on similar grounds. Hardin does not discuss online trust the as a moral problem. But his conclusions raise questions pertaining to morality. Trust is a valued quality of personal relations, and its absence on the web would therefore mean that the preconditions for an important and valuable basis for cooperation and interaction would be lacking. And also if one disagrees with Hardin’s claim that there cannot be trust on the web, as I do, it seems clear that trust on the internet raise moral issues.

But what exactly is the moral problem here? Helen Nissenbaum (2004) is a philosopher who has clarified this question considerably. She approaches the question by way of a discussion of the relationship between trust and security, and argues that security does not give us trust, Nissenbaum claims, but rather some kind of assurance. Both trust and assurance are bases for co-operation. The difference is the relationship to risk. When trust is the basis there is co-operation despite the risk of opportunism. When there is co-operation on the basis of assurance, in contrast, the perceived risk is instead reduced to an acceptable minimum - through security measures for example. This distinction between trust and assurance seems also to be supported by empirical research by Molm et.al. (2000) who concludes that “ironically, the very mechanisms that were created to reduce risk in transactions – the negotiation of terms and strictly binding agreements – have the unintended consequence of reducing trust in relationships.”²

In contrast to Hardin, Nissenbaum argues that trust is possible online. She argues, however, that we should distinguish between two conceptions of trust: instrumental and non-instrumental trust. Instrumental trust is the kind we typically find in market exchanges, and is according to Nissenbaum, an “imperfect substitute for information”. This kind of trust is deemed to be inferior to assurance. The second conception is non-instrumental trust which applies to situations where trust is an overriding value. “When trust is so conceived, trust-based relationships would not be better if the scope of trust were minimized; rather the opposite holds. In traditional realms, trust among family members exemplifies this conception.”³

This distinction is unclear. First, trust is not in itself a value, I argue. It is only valuable if it is well-founded. Second, I argue that trust is never instrumental but can only develop by indirect means. It is not possible to force someone to trust you, for example, but has to develop autonomously in the person. For these reasons, I believe it is important to develop an account of well-founded trust that does not assume this distinction.

A general theory of well founded trust is proposed by Bacharach and Gambetta (2001, and Gambetta 2006) and explored empirically in Gambetta and Hamill (2006). First, we assume a trust-game, i.e. a situation where there is the possibility for opportunism. But the suggestion of Gambetta is that the problem of trust is the second-order problem of the signal of trustworthiness of the signal of trustworthiness. He suggests that we should look to the theory of signalling games in the analysis of trust. In a signalling game there is a trustee with private information who sends a signal to a truster. The truster has only access to the signal but not direct information to concerning whether the trustee is trustworthy or is an opportunist. In a population where there are both genuine co-operators and deceptive agents who can send signals, the challenge for the truster is to infer the type of the trustee on the basis of the signal. The theory of signalling games says that there will be a sorting equilibrium if, roughly, two conditions hold ⁴: (1) the benefit to an honest signaller of being treated as a genuine co-operator exceeds the cost the cost of sending the signal, and (2) the cost to a non-genuine signaller of sending the signal exceeds the benefits to him of being treated as a genuine co-operator. In a sorting equilibrium a signal of trustworthiness will always carry truth. As an example of a sorting signal, consider the story of the king who was afraid of being poisoned and therefore required that anyone who offered him wine should himself taste the wine first. This is a signal that it is impossible to simulate and that satisfies the two conditions. The extent to which a signal is carrying truth or not depends on a number of things, including the costs and benefits involved, the genuine / non-genuine signaller ratio, and receiver knowledge.

 

This theory can be used to formulate a conception of well-founded trust on the web, I will argue in the paper. To illustrate this idea, I will discuss various forms of fraud on eBay and the eBay-policies to prevent such fraud. One such is the example of fraudulent “second chance offers” ⁵ . In these cases a bidder who has come out second in the auction is contacted with a fake email saying that the winner has withdrawn his offer and that he is given a second chance to buy the item. Naïve users think that the appearance of “eBay” in the sender’s email address, and the appearance of the eBay-logo in the page to which he is directed is evidence that the offer is genuine. A naïve user will be fooled if he is not aware that these are very cheap signals to produce. More sophisticated users, however, who perhaps have read the eBay policy on second chance offers, and know that real second chance offers are never given on email, and that the page where the offer can be accepted is identical to the original page of the auction, but with the second bidder on top. This is a signal that it is much more difficult to fake for a non-genuine seller and therefore closer to a sorting signal. Trust on the basis of sorting signals can thus be seen to be a normative ideal in trust-based systems.

On this analysis, developed in detail in the paper, the theory of signalling games can be used as the basis for an account of well-founded trust and hence provide an important part of the characterisation of the moral problem of trust online. Furthermore, it is argued that this can be done without assuming two different notions of trust (instrumental and non-instrumental trust).

Reference :

Bacharach, M. and D.Gambetta. 2001a. Trust As Type Detection. In Trust and Deception in

Virtual Societies , edited by Christiano Castelfranchi and Yao-Hua Tan. Kluwer.

Bacharach, M. and D. Gambetta. 2001b.Trust in Signs. In Trust in Society, edited by Karen S.

Cook. New York: Russel Sage, pp. 148-184.

Gambetta, Diego and Heather Hamill. 2006. Streetwise. New York: Russel Sage.

Hardin, Russel. 2004. Internet Capital. Analyse & Kritik. No. 1, vol. 26, 122-138.

Jøsang, C. Keser and T. Dimitrakos. 2005.Can We Manage Trust?, in Trust Management.

Edited by Peter Herrmann, Valerie Issarny and Simon Shui, pp. 93-107. Springer.

Kollock, P. 1994. The Emergence of Exchange Structures: An Experimental Study of

Uncertainty, Commitment, and Trust. The American Journal of Sociology, Vol. 100,

no. 2, 313-345.

Molm, L.D. et a: 2000. Risk and Trust in Social Exchange: An Experimental Test of a

Classical Proposition. American Journal of Sociology, Vol. 105, no. 5. Pages 1396-1427.

Rouhaama, S. and L. Kutvonen. 2005. Survey of trust management, in Trust Management.

Edited by Peter Herrmann, Valerie Issarny and Simon Shui, pp 77-92. Springer.

Helen Nissenbaum. 2004. Will Security Enhance Trust Online, or Supplant It?. In Trust and

Distrust in Organizations: Dilemmas and Approaches . Edited by Roderick M. Kramer and Karen S. Cook. New York: Russel Sage.

Philip Pettit. 2005. Trust, Reliance and the Internet. Analyse & Kritik. No. 1, vol. 26, 108-120.

Shauna Wright. 2006. Don’t Get Burned on eBay. Sebastopol: O’Reilly.

1 See, for example, Rouhaama et. al. 2005 and Jøsang et. al. 2005.

2 p. 1398. Earlier work by Kollock (1994) also implies that the distinction between trust and assurance is essential.  

3 Nissenbaum (2004), p. 176.

4 The formulation is adapted from Bacharach and Gambetta 2001a.

5 See Wright 2006.