Monday, August 29, 2005

Comments on Paul Dumouchel: "Trust as an Action"


Mary Cassatt, The Bath, 1891-2.


Gloria Origgi
Comments on Paul Dumouchel : « Trust as an Action »

Trust is one of the most intractable notions of philosophy and social science. That is because a variety of human interactions are pulled in under the heading of trust. Trust is involved in any asymmetrical situation in which one party has to take a risk that depends on the performance of the other party. And risky businesses concern as many different relations as one can imagine, such as commercial transactions between parties who don’t know each other, temporally asymmetrical transactions (I pay you today for a good that I will receive in a week or a month), love affairs, reliance on experts for important decisions about one’s own life and health, occasional conversational exchanges among people in the street, and so on.
The phenomenological intuitions about trust are as confused as its theoretical definitions, and I agree with Paul Dumouchel that a contrast between “cognitive” vs. “behavioural” (even if he doesn’t use this word) approaches has at least the merit of drawing a line between two main regions of this tangled conceptual space. Surprisingly, as Piotr Sztompka has emphasized in his comment, Dumouchel proposes his distinction as it were original, when it has already been discussed at length in the literature on the subject (see for example R. Hardin, 2002, p. 58). Still, Stompka’s comment denies any interest to this dichotomy, by arguing that most approaches to trust take into account both dimensions, whereas I think that Dumouchel does refer to a clearcut distinction between two kinds of approaches, approaches in the rational-choice tradition that focus on the set of rational beliefs that determine the course of actions an agent prefers given how he or she represents the causal structure of the situation, and approaches to trust that see it as the outcome of an action that was not motivated by a rational assessment of the odds. In the first part of his article, Dumouchel criticizes cognitive approaches, while in the last part, he puts forward his behavioural interpretation of trust. In this comment, I briefly consider Dumouchel’s criticism of cognitive approaches and then concentrate on his view of trust as an action.

In criticizing the cognitive view, Dumouchel makes two points that don’t seem to fit together. He first says that cognitive theories of trust “cast their net too wide” by failing to distinguish trust from any other kind of expectation about one’s future action. He then goes on to say that cognitive perspectives on trust boil down to the assessment of other people’s trustworthiness. Not only this last assertion is unfair, given many ongoing debates in the rational-choice tradition about the difference between trusting and assessing trustworthiness (see Hardin 2002; Bacharach and Gambetta, 2001), but also it contradicts the first criticism about the impossibility to distinguish in this tradition trust from other kind of expectations about one’s future action, given that expectations of trustworthiness are a very special class of expectation. I don’t assess my partner’s trustworthiness while playing tennis against her: I act on the basis of rational expectations about her future behavior in a competitive game. Dumouchel argues that a cognitive perspective should consider the tennis interaction as a case of trust: But how can he consider this as a case of assessment of trustworthiness? It is difficult to understand why Dumouchel in criticizing the ‘cognitive view,” targets a definition of trust as subjective probability or expectation that nobody really defends.

In his proposal to consider trust as an action, some of Dumouchel’s formulations such as: “To trust is to act in such a way as to give another agent power over us” or “When I trust I increase my vulnerability to another agent through an action on my own” strongly echo already existent definitions, such as, for example, Annette Baier’s definitions of trust as “as the accepted vulnerability to another’s possible but not expected ill will toward one” (cf. A. Baier, 1986). He goes on to say: “From the point of view of trust as an action, it is because we trust that we have a normative expectation”, in contrast with cognitive approaches according to which it is the expectation that makes us trust and not the other way round. But this claim too sounds familiar, and it is not far from Otto Lagerspetz characterization of trust: “trust is not the fact that one, after calculating the odds, feels no risk: It is feeling no risk without calculating the odds” (cf. Lagerspetz, 1988). Dumouchel might be tempted to argue that motivational approaches to trust, such as Baier and Lagerpsetz’s, view trust as a feeling or an emotion, that is, a non-cognitive state that causes a certain action, whereas his own proposal is to view trust directly as an action, without further inquiry about what kind of internal state (cognitive or non-cognitive) could have motivated that action. Indeed, one of the immediate advantages that Dumouchel claims for his view is that “actions can be observed in the world”, thus relying on actions instead of expectations or emotions helps to understand why trusting can influence the future course of the events between two parties: Actions have a causal power on the world that is at least partly independent of the intentions for which the action has been performed. Relying on Anscombe’s distinction between doing something intentionally and doing something with an intention, Dumouchel argues that we can intentionally perform the action of trusting someone, and thus giving her or him power on us, even if our intention is not to give her or him power on us. For example, I pay before delivery for a good in order to take advantage of a 10% discount, thus giving the seller the power of cheating me, even if my intention is not to trust him or her, but to save the 10% of the money. Thus, trust is acting cooperatively in a situation when the interests of the parties are neither totally divergent, nor totally convergent.

At a first glance, Dumouchel’s account of trust as an action seems to face the same objection he raised in the first part against the “cognitive” approaches to trust, that is, it fails to explain why trust should be distinguished from other types of cooperative actions (as for example letting your car pass before mine at an intersection) in the same sense as cognitive approaches fail, according to him, to distinguish trust from other kinds of expectation about the future actions of the other agent. If it is the action that counts, and not its motivation, it would be important to understand what is the feature of the action that differentiates it from other classes of actions. Dumouchel’s idea is that what is specific of trust as an action is that, in order to succeed, it must be recognized by the trustee. I cannot trust another person without her awareness that I am trusting her. And it is this acknowledgement of my trust that creates a bond in the other agent, that is, the bond to reciprocate, and not to betray my trust. But this doesn’t explain exactly what particular ingredient of trust as an action creates a moral bond. If I pay a book on Amazon.com before receiving it, I’m doing it because I cannot do otherwise, but Amazon.com doesn’t send me the book because it has recognised my trust, but because I have paid for it. Or if I go to the doctor because I am ill, it is not the fact that I trust him or her that creates a moral obligation to heal me, but the fact that I am ill. Or else - to take the final example provided by Dumouchel - if a child gives power to an abusive parent, it is not because he or she trusts the parent, but because she or he cannot do otherwise. And it should not be the parent’s recognition of his or her child’s trust that creates a bond to reciprocate. What Dumouchel reintroduces in the end of his paper in order to explain why people act trustfully is that “the trusting agent believes that some good will follow from it”, which, combined with the condition of the recognition of the trustful intention by the trustee, gives us back the classical view of trust in terms of expectations about the other agent’s actions towards ourselves in a context in which our own action can influence the course of events. Dumouchel thus seems to admit in the end that actions are motivated by rational expectations about the course of events.

It is unclear also how dependence creates the need to act trustfully. Dependence often causes us to grant power to people whom we do not trust. We can also act altruistically and trust other people without being dependent. I can be trustful because I am in a privileged moral, social or emotional disposition that leads me to trust despite the lack of evidence about the other.
Finally, I can’t see how Dumouchel’s view of trust as an action can accommodate “the specificity of trust in different forms of cooperative behaviour”, as he claims it does in conclusion. Many situations of dependence and imbalance of power that he wants to treat within his approach are not “cooperative” in the standard sense. Clearly my trustful action towards my surgeon won’t much influence - or so I hope - his or her performance. If Dumouchel’s point is just the general and well known point that acting trustfully, even in absence of any motivation, can sometime positively reinforce reciprocity, it would be interesting to understand why it is so and to explore further the dynamics of trustful interactions instead of concentrating on the individual’s action.