A Revision of the Buechner–Tavani Model of Digital Trust and a Philosophical Problem It Raises for Social Robotics

information Article A Revision of the Buechner–Tavani Model of Digital Trust and a Philosophical Problem It Raises for Social Robotics Jeff Buechner 1,2 1 2 Department of Philosophy, Rutgers University-Newark, Newark, NJ 07103, USA; The Saul Kripke Center, CUNY, The Graduate Center, New York, NY 10016, USA Received: 16 December 2019; Accepted: 13 January 2020; Published: 16 January 2020



 Abstract: In this paper the Buechner–Tavani model of digital trust is revised—new conditions for self-trust are incorporated into the model. These new conditions raise several philosophical problems concerning the idea of a substantial self for social robotics, which are closely examined. I conclude that reductionism about the self is incompatible with, while the idea of a substantial self is compatible with, trust relations between human agents, between human agents and artificial agents, and between artificial agents. Keywords: artificial agent (AA); trust; self-trust; Buechner–Tavani model of trust; self-identification; reductionism about the self; substantial self; personal identity; normative expectation; diffuse default trust 1. Introduction The Buechner–Tavani [1–3] model of trust is one of the first models of trust that accommodates both human agents and Artificial Agents (AAs)—otherwise known as digital agents (see Taddeo [4]). That is, within the model the following three trust relations can be defined: (i) trust between human agents, (ii) trust between human agents and AAs, and (iii) trust between AAs. Because the model can define these three different kinds of trust relations, it is a model of digital trust, since digital agents—AAs—can have trust relations. However, the model is incomplete in one way—it does not say anything about self-trust. In this paper, the Buechner–Tavani model of trust will be amplified by two additional conditions, each involving self-trust. The addition of these two conditions creates a philosophical problem for social robotics—notably, the problem of a substantial self vs. reductionism about the self. This specific problem is an instance of a more general problem for social robotics: there are philosophical concepts necessary for making certain arguments in social robotics and these concepts are controversial—there are opposing sides as to whether they are or are not viable concepts. In such cases, how do we proceed? Do we simply bracket the point of making an argument to establish some claim in social robotics, or do we attempt to refute one view as to whether the concept in question is or is not viable? I will examine the more general problem by examining the specific problem, which is an instance of the more general problem. This is the problem of whether, and how, human agents and AAs could have a substantial self (and the attendant concept of a substantial self) in virtue of which they could have (or fail to have) self-trust. I will argue that even though there might be computational realizations of reductive views of the self (such as Parfit’s relation R) in AAs, there is an incompatibility between reductionist views of the self and trust. Additionally, even though there might be no computational realization of a substantial self in AAs, a substantial self is needed for trust relations. The main points of this paper are to emend the Buechner–Tavani model of trust and to show that trust is incompatible Information 2020, 11, 48; doi:10.3390/info11010048 www.mdpi.com/journal/information Information 2020, 11, 48 2 of 24 with a reductive view of the self and compatible with the existence of a substantial self. It is left to another paper to show how AAs can have trust relations with human agents and other AAs even if there is no computational realization of a substantial self in AAs. 2. Adding New Conditions to the Buechner–Tavani Model of Trust New conditions need to be added to the Buechner–Tavani model of trust [1–3]. The new conditions incorporate the concept of self-trust. The motivation for these new conditions is that human agents must trust themselves in order to trust others. Without trust in oneself, a human agent cannot meaningfully trust other human agents. There are several different reasons for the necessity of self-trust in the Buechner–Tavani model of trust. 2.1. Why Self-Trust Is a Necessary Condition for Trusting Others 2.1.1. Changes in Personal Identity over Time One reason why self-trust is necessary for trust is that we change over time, often to such an extent that we appear to be a different person at tn than the person we were at tn-k . We exhibit one form of self-trust when we trust that the person we will later become is one that we will want to become. Similarly, we trust the person we now are to be someone that will take reasonable steps to become the person we will later become. If we were unable to have these relations of self-trust between earlier and later versions of one’s self, we would not be able to trust another human being, nor would we understand what it means to trust another human being. Why is that? If we cannot trust our own self to take the steps necessary to become the person we wish to become, how could we trust another human being to take steps to do what we wish them to do? And if we could not do that, how could we understand what it is to trust someone else to take steps to do what we wish them to do? 2.1.2. Self-Trust and Self-Competence The following argument occurs in a different form in Lehrer [5], where he argues that self-trust is a necessary condition for any human agent to have competence of a certain kind—reasonableness, and joining in the life of reason. Anyone—in Lehrer’s view—not worthy of their own trust cannot enter into the life of reason. I am not making that argument here. Rather, I am using the argument to argue for self-trust as a necessary condition for trust (see also Wright [6]). Suppose that A successfully teaches B how to use modus ponens (MP) and that B accepts what she has learned from A. That is, B conforms to the rule of MP. B then reasons in accord with MP. In this way, B is worthy of the trust of A, for A normatively expects it of B that B will correctly use MP in reaching conclusions in chains of reasoning involving MP. Indeed, it is reasonable that A accepts the conclusions that B draws using MP. Now suppose that—for whatever reason—B does not understand MP. Although B uses it correctly, B does not have any reason to believe that she is using MP correctly—even though she does use it correctly. B is ignorant of the justification of the rule and thus of its merits. If so, the conclusions that B draws from the rule are not reasonable for B to accept. B is not worthy of her own self-trust because the reasoning that B performs is not reasonable for B to accept, even though it is reasonable for A to accept. A knows the rule MP and also knows the justification of MP and so the merits of MP. A observes B using the rule MP and knows that B has (...truncated)