Argument schemes for reasoning about trust

Abstract

Trust is a natural mechanism by which an autonomous party, an agent, can deal with the inherent uncertainty regarding the behaviours of other parties and the uncertainty in the information it shares with those parties. Trust is thus crucial in any decentralised system. This paper builds on recent efforts to use argumentation to reason about trust. Specifically, a set of schemes is provided, and abstract patterns of reasoning that apply in multiple situations geared towards trust. Schemes are described in which one agent, A, can establish arguments for trusting another agent, B, directly, as well as schemes that A can use to construct arguments for trusting C, where C is trusted by B. For both sets of schemes, a set of critical questions is offered that identify the situations in which these schemes can fail.

Keywords:

• argument representation
• formal models of argumentation
• social influence

Acknowledgements

The authors would like to thank the anonymous reviewers of previous versions of this paper. The comments of these reviewers were very helpful in improving the paper. We would also like to thank the participants at the 2012 Conference on Computational Argumentation (COMMA), especially Trevor Bench-Capon and Tom Gordon, for their comments on an earlier version of this paper.

The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Army Research Laboratory, the National Science Foundation, the National Security Agency, or the US Government. The US Government is authorised to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation here on.

Funding

This research was funded under Army Research Laboratory Cooperative Agreement Number W911NF-09-2-0053 by the National Science Foundation [Grant #1117761] and by the National Security Agency under the Science of Security Lablet grant (SoSL). Additional funding was provided by a University of Liverpool Research Fellowship and by a Fulbright-King's College London Scholar Award.

Notes

^† This is a revised extended version of a paper that appeared at the International Conference on Computational Argumentation (COMMA), 2012 (Parsons et al., 2012).

1. Subjective probability having a natural interpretation as a propensity to make bets at particular odds (Jaynes, 2003, p. 655).

2. As we shall see, this is just one possible pattern of trust propagation.

3. ‘Comping’ is food industry slang for providing food and drink at no cost, typically to friends or family of the providers. The Urban Dictionary, http://www.urbandictionary.com, suggests the term is derived from ‘complimentary’.

4. The MTA is the body that runs the public transportation system in New York City.

5. http://www.yelp.com.

6. Recall that propagation is when A takes the step that says A trusts B and combines it with information that B trusts C to infer something about the relation between A and C. Reputation, in contrast, is a mechanism by which A may establish trust in B from information given to A by D and E.

7. This is not necessarily a very strong argument for trusting, and the cinephile reader may recognise in this example a minor plot device from the classic 1973 movie The Sting (http://www.imdb.com/title/tt0070735/) where (spoiler alert!) the failure of the scheme to predict trust was important.

8. Where it can be a winning strategy to start out trusting others to be cooperative.

9. There may, of course, be other schemes that capture aspects of the same relationships. Indeed, this paper contains a number of schemes that were not present in Parsons et al. (2012), even though that set of schemes was also conceptually complete in the same way as this one is.

10. In the play (spoiler alert!), Cyrano is in love with his cousin Roxanne, but believes that his appearance means that Roxanne can never love him. Roxanne, meanwhile, has a crush on Christian, one of Cyrano's comrades in the Gascon Cadets, and Christian is also in love with Roxanne. Christian, who is unable to express his love for Roxanne, enlists Cyrano to help him, and Cyrano composes letters and speeches that Christian can use to woo Roxanne. It is as a result of these communications, ostensibly from Christian, but actually from Cyrano, that Roxanne falls in love with and marries Christian. Much later, once Christian has died a heroic death, the deception is revealed, and Roxanne realises that it is Cyrano that she has loved all along. Sadly Cyrano is dying by the time that this occurs. One might, of course, argue about Cyrano's motives. Perhaps he is acting nobly because he thinks that his love for Roxanne can never be requited. Perhaps he is just using any means he can to be able to interact intimately with her.

11. Formally, Simpson's paradox Blyth (1972) is that but Less formally, to take the widely quoted example, Simpson's paradox is how the baseball player David Justice can have a higher batting average than Derek Jeter for each year in the period 1995–1997, and yet have a lower batting average over the three years combined (Pavlides & Perlman, 2009; Ross, 2004). To see how this can be the case, it is necessary to know that a batting average in baseball is the number of ‘safe hits’ that a player attains in some period divided by the number of ‘at-bats’, that is, the number of chances to have a safe hit. Justice's batting averages are 0.253, 0.321, and 0.329 for the three years, and 0.298 when aggregated over the three years, Jeter's are 0.250, 0.314, 0.291, and 0.300, respectively. The paradox arises because the number of at-bats varies from year to year. Justice had many at-bats in 1995 and 1997, but many fewer in 1996 as the result of a shoulder injury. Jeter had only a few at-bats in 1995, when he was a rookie, and many more in 1996 and 1997.

12. We prefer the term ‘direct propagation’ since if one considers the context in which trust is being applied, we can view this as inferring a functional trust relation between A and C on the basis of a functional trust relation between B and C and a referral trust relation between A and B. We discuss this more below.

13. As parents of teenagers will know, the fact that your children fail to do the chores that you ask them to do will not stop them from relying on you to do things like providing them with regular meals.

14. More correctly these schemes typically involve two different trust contexts since it is perfectly possible that the context of the functional trust that is being considered is the ability to make referrals, so that the entire propagation scheme is relating to referral trust.