I'm interested in the concept of trust in the context of secure autonomous agent computing. Before trying to think about what properties computer trust should have, I decided to see what philosophers are saying about trust between people.
In this brief book (100 small pages with lots of white space) O'Neill discusses not so much what trust is, however, than the necessary conditions for it to exist in society, and the way many current practices ostensibly designed to increase it may paradoxically be reducing it. The discussion covers the question of human rights, accountability and audit, transparency, and press freedom.
The passive demand by people to be delivered their rights can lead to a vicious spiral of reducing trust and increasing fear. O'Neill starts from Kant's argument that we are all moral equals, and so that the principles we live by should be principles for all, that we should act only on principles that others may also act on, ruling out deception, coercion and intimidation, and requiring us to actively uphold the rights of others. This view moves the citizen from passively demanding their own rights to actively building a society where everyone's rights are upheld.
Next O'Neill tackles the ever-increasing accountability and auditing legislation, ostensibly designed to ensure that institutions are behaving as they should, and hence demonstrate that they are trustworthy. But in fact the accounting practices are so onerous, and conflicting, that they impose perverse incentives on the staff (to shorten waiting lists by treating "easy" problems, rather than urgent ones, for example), and so end up distorting the very services they were designed to improve, and make citizens more suspicious. The problem is that the required accountability isn't to the public, to the people supposed to benefit from the services, but to the regulators, the government, the paymasters, and the law.
The next practice to come under the spotlight is the move to ever greater transparency, which is supposed to reduce secrecy, and hence enable trust. O'Neill argues that it isn't secrecy that is the problem, however (we quite happily trust our colleagues without knowing all the details about their income and their private lives, for example), but rather deception. We ought to require measures that reduce the opportunities for deception, that enable us to check the information on which we base our decisions. Yet transparency can increase the opportunities for deception, as we drown in masses of uncheckable and irrelevant data. And it can increase the occurrences of self-censorship and evasion, as people become less willing to state opinions that will be made public.
Finally we come to freedom of the press. O'Neill argues that reason for this freedom is to give citizens uncensored information on which to base their decisions. And that implies that the press should not be "free" to disseminate lies, rumours, and gossip as if it were fact.
This is a good read about some very important topics. By going back to first principles, asking what we want to achieve, and what is necessary to achieve that, O'Neill has pointed out many flaws and contradictions, and some downright stupidities, in current policies. It might not address my original reason for reading it (my problem, not its), but it has certainly given me food for thought.