A Telic Theory of Trust

As legend has it, the Swiss folk hero William Tell was forced to shoot an apple off the top of his son Walter’s head with a crossbow to save both their lives. Walter trusted his father, so he confidently stood in front of a tree with the small target on his head. Being a great marksman, William made an excellent shot, hitting the small apple from a distance. And, crucially, Walter was right in trusting his father. He trusted well. In A Telic Theory of Trust, Adam Carter defends the view that William’s shot and Walter’s trust are the same kind of thing: Both are performances, aimed attempts. Hence, as with any other performance, the standards by which we judge that William’s shot was a good shot are the same standards by which we judge that Walter’s trust was good trust. This is a provocative idea, for it runs against views that equate trust with states like belief, hope, or desire and thus equate the normativity of good trusting with that of good belief, good hoping, or good desiring. Carter aims to show how seeing trust as a performance dissolves many of the key problems these views face and yields a detailed framework that illuminates the intricacies and forms of trust.Chapter 2 begins by setting up the core strategy of the book: deploying the conceptual machinery originated in virtue epistemology to answer the question ‘What is good trust?’ Crucially, trust is not essentially an epistemic phenomenon. While trust is a vital aspect of many of our epistemic practices, there’s a plethora of cases where the outcome of trust is nonepistemic. Trust is ubiquitously present and required in our cooperative interactions. Indeed, one of the most exciting aspects of Carter’s book is its demonstration that the telic framework of normativity promisingly extends well beyond epistemology.In particular, the telic framework involves three interacting norms by which we evaluate aimed performances: those of success, competence, and aptness. William’s shot was a good shot because it was successful (it’s what William aimed to do), competent (it stemmed from William’s well-placed disposition to make successful shots), and apt (it was successful because it was competent). Lesser shots would not exhibit this coordination of telic features: Perhaps they are successful but incompetent, or competent but unsuccessful, or competent and successful but not apt. The telic theory of trust says that Walter’s trust that his father would shoot the apple on his head is also an aimed attempt, subject to three parallel telic norms. An instance of trust is successful when the trustee actually performs the action they were entrusted with as it was entrusted by the truster. An instance of trust is competent when it stems from the trusting agent’s disposition to reliably trust successfully in the right shape and situation. And an instance of trust is apt when its success is owed to the deployment of the disposition that made it competent. The rest of the book aims to develop, specify, and clarify this view.Carter points out that this basic picture already avoids key problems faced by extant views of trust. Take the doxastic view, which holds, roughly, that to trust someone to φ is to believe they will φ. Among its problems is a tension between good believing and good trusting. The quality of a belief increases along with evidence; the more evidence one has for one’s belief, the better the belief. But this is not true of trust. While I can come to trust badly if I’m too gullible, trust requires a degree of vulnerability to betrayal that is not present in belief formation. If I carefully verify whether my doctor prescribed the right medicine until I confirm by my own means that they did, I’m undermining my trust in them. Similar problems arise for nondoxastic views that equate trust with affective or conative states. The norms of good affection or good conation are either in tension with or orthogonal to the norms of good trusting.The telic model of trust is clearest in cases of deliberative trust—when we explicitly deploy our discerning capacities to decide whether we should trust another person. However, the telic model easily extends to implicit trusting. All that is required for trust to be subject to telic norms is that the performance of trust is an aimed attempt, even if that attempt stems from the functional or teleological, rather than intentional, features of the systems that perform it. Yet Carter holds that apt deliberative trust is of a higher quality than apt implicit trust. This is because implicit trust merely aims at success, while deliberative trust has a stronger aim: to trust if and only if it would be apt to trust. In other words, deliberative trust aims to trust only when success would be ensured by the agent’s disposition to successfully trust. These distinct aims also explain how the telic framework accounts for particular cases of distrust. Insofar as an agent aims to successfully and/or aptly trust, they will also aim to forbear trusting in cases where trusting would be unsuccessful (in the case of implicit distrust) or inapt (in the case of deliberative distrust).Chapters 5 and 6 develop in more detail what is at play in deliberatively apt trust, or convictively apt trust, as Carter also calls it. This involves first-order and second-order competence. First-order competence obtains when trust stems from the right skill (the disposition to successfully trust), deployed by an agent in the right shape (in good cognitive function and without ex ante manipulation), and in the right situation (when betrayal is neither too easy or convenient for the trustee, nor too hard to recognize by the truster). An instance of trust involves second-order competence when the agent explicitly assesses whether their trust would be apt. In other words, second-order trusting competence requires agents to properly and explicitly assess the risk that the conditions of their trusting will not lead to the proper deployment of their skills in a way that leads to successful trust. Crucially, an instance of convictively apt trust occurs when the fact that an agent’s trust is first-order apt is based on the fact that the agent’s trust is second-order apt.Given that deliberately competent trust must involve some openness to betrayal but also some assessment of the risk of betrayal, the next natural question is, ‘What kind of risk is permissible to ignore when deliberately trusting?’ To answer this question, Carter applies his de minimis account of risk, developed elsewhere. In short, a trusting agent may only ignore risks that would not be addressed by adhering to one or more cooperation-sustaining rules. Put otherwise, agents are entitled to ignore minimal risks that would only be addressed by precautions that are independent of the cooperation practices that are enabled by their trust. In turn, this account yields a nice derivation of the kind of risk that is fundamental to trust and the kind of monitoring that is incompatible with trust. Trusting agents, even when trusting well, are inherently vulnerable to (1) risks that are so modally distant that they can’t be addressed in any cooperative-conducive way and (2) risks that can only be addressed in ways that would diminish or disable the cooperative interaction enabled by trust. For example, I can permissibly ignore the risk that a once-in-a-lifetime fluke on my doctor’s computer causes them to prescribe the wrong medicine. This is because I cannot guard against said risk by adhering to the practices that enable the trust-based doctor-patient relationship. In fact, if I tried to eliminate such a risk entirely, I would likely undermine that very relationship.Another extension of the telic model lies in its promise to illuminate not only trust but trustworthiness as well. Carter suggests that being properly trustworthy amounts to responding well to good trust. This leads to telic norms of trustworthiness that map onto the norms of trust. To respond well to trust, one must do so successfully, competently, and aptly. Good trust and good trustworthiness amount to the proper realization of two symmetrical sides of a cooperative exchange; the truster’s achievement in trusting matches with the trustee’s achievement in responding to trust. This “achievement matching” is the gold standard of truster-trustee interactions against which lesser instances of trust are evaluated. Importantly, adopting a telic picture of trustworthiness allows it to be analyzed in tandem with trust, and not as an independent, fundamentally dispositional property, as is conceived by several other views. This yields a unified framework of analysis of trust and trustworthiness, poised to address further programmatic questions with which Carter concludes the book, such as ‘What’s collective trust?’ ‘What kind of outsourcing is compatible with good trust?’ and ‘Can we trust nonhumans—specifically, AI models?’Carter has given a cogent, exemplarily detailed and thorough account of what good trust amounts to. However, a metaphysical question lingers: What does trust simpliciter amount to? The telic norms proposed by Carter have ‘trust’ in their content, so they assume that one already has a notion of trust at hand. This is fine when we have an intuitive grasp of the performance to which these norms apply—as with William Tell’s shot. It’s not so clear to me what, in Carter’s view, Walter did what amounted to trusting his father’s shot. What, specifically, happens in the world that makes it the case that A trusts B to φ? We know two things from Carter’s success norm: B φ-ed and they φ-ed as A entrusted them. It would seem, then, that whatever trusters do is encapsulated in the ‘as entrusted’ clause. But saying ‘trusting is entrusting’ is not too helpful. Should we think that the performance in question is the formation of a distinctive cognitive state that somehow represents the way in which A expects B to φ? Perhaps the performance of trust is more than that. After all, Walter moved his body in a way that not only aligned with but enabled the success of William’s shot. Yet not all trust seems to involve this kind of action coordination. Alternatively, then, perhaps trusting requires honestly signaling our expectations of how others should act? Carter mentions that he wants to defend an ecumenical picture of what ‘as entrusted’ means. However, without details about what trust and entrusting amount to, one may worry that while the telic theory of trust elides the core pitfalls of extant theories of trust, it does so by also eliding the core questions that they were meant to answer.Other open questions concern Carter’s claim that trust occurs ubiquitously in cooperation. It’s unclear whether Carter holds that every instance of cooperation involves trust. But here’s a natural way of understanding cooperation that may suggest so: Roughly, to cooperate is to act mutually and responsively with others to bring about an outcome in a way that would not occur if either party were to default in their efforts. If A and B are indeed cooperating, A must trust that B will do their part in what they’re doing together. If B fails to act accordingly, or if A takes over B’s part, cooperation will not ensue. This notion, however, is silent on what motivates cooperation. A and B may cook together because they’re friends, or because doing so will bring them closer to their own selfish goals, or because A has threatened B with punishment. Presumably, even if motivated by threat or selfishness, their cooperative actions can still be mutually dependent, require skillful performance, and be vulnerable to default or betrayal. What’s more, selfish or threatened motivational profiles of cooperation occur ubiquitously if we think that conventions are often adopted by agents whose selfish motives happen to align (à la David Lewis) and that social norms are sets of implicit social threats (à la Christina Bicchieri). Carter claims that trust and trustworthiness are defective or absent when agents fulfill what they’re entrusted with under threat—which suggests that motivation matters for trust. But this is hard to parse if cooperation can genuinely and ubiquitously ensue selfishly and under threat. Perhaps the ‘as entrusted’ clause involves a goodwill condition. Or perhaps the relevant sense of cooperation is thicker than the one I suggested. This, again, requires delving into the question of what is trust simpliciter.Even with these open questions, Carter’s book is a refreshing and rigorous contribution to the literature on trust and an exemplar of systematic and provocative philosophical thought. Its content is richer than this review can capture and deserves a thorough engagement with the entirety of the book.