Incompatibilism and “Bypassed” Agency

CONDITION: In Universe A, is it possible for a person to be fully morally responsible for their actions? YES NO Only 14% of subjects in the abstract condition thought that it would be possible for an agent to be fully morally responsible in Universe A, while 72% in the concrete condition thought that Bill was fully morally responsible for his action. Judging from these results, different ways of asking about responsibility in deterministic scenarios can trigger contradictory intuitions.1 For our purposes, this is interesting in several ways. 1 Each condition had a little more than 40 subjects. Another group of subjects were in a concrete condition with less elaborate description of the action in question: “In Universe A, Bill stabs his wife and children to death so that he can be with his secretary. Is it possible that Bill is fully morally responsible for killing his family?” Here, 50% answered “yes”. INCOMPATIBILISM AND “BYPASSED” AGENCY 4 First, since a substantial majority of subjects gave incompatibilist answers to the abstract question, incompatibilist reactions seem to be grounded in a common, non-idiosyncratic, understanding of responsibility and determinism. Second, since answers in the concrete condition seem to contradict those in the abstract condition, it is natural to assume that most judgments in one of these conditions are based on some sort of mistake: perhaps the concrete details in the former remind us of something required for responsibility, or obscure the deterministic character of the scenario or some important consequences of determinism. Third, since incompatibilist reactions are substantially undermined when people are asked about concrete acts of wrongdoing, they are unlikely to rely on what is front and center in this common understanding of moral responsibility and determinism. Though pervasive, the mechanisms by which determinism undermines responsibility judgments seem to be relatively subtle. The latter point is strengthened by variation in the extent to which subjects have been willing to attribute responsibility to agents in various studies: there is considerable variation in responsibility attributions depending both on the ways in which determinism is characterized in the relevant scenarios and on how the questions about responsibility are asked. In particular, descriptions of determinism in terms of how earlier events cause later events or make them predictable seem to undermine responsibility attributions to a much lesser extent than scenarios also stressing that prior events necessitate later events, as in the scenario above (e.g. Nahmias et al. 2006; Nahmias et al. 2007). The non-obviousness or non-centrality of assumptions underpinning incompatibilist reactions is also highlighted by considerable interpersonal variation in responsibility attributions. For example, in a study using the Nichols and Knobe (2007) vignettes and abstract / concrete conditions, I asked subjects to indicate their level of a agreement with a statement saying that in Universe A it is possible for Bill to be fully morally responsible for killing his wife and children (concrete conditions) or for a person to be fully morally responsible for their actions (abstract condition). Answers, which were given on a 1-to-6 scale ranging from strongly disagree to strongly agree, are represented in Figure 1, where the size of each bubble indicates the number of replies at that point, ranging from 1 to 21 (subjects (N=155) were recruited from Amazon Mechanical Turk). The difference between the abstract and concrete conditions is in line with the replies in Figure 1: ABSTRACT, RESPONSIBILITY INCOMPATIBILISM AND “BYPASSED” AGENCY 5 Nichols’ and Knobe’s experiment (M=2.37 and 3.58, respectively), but a striking spread of answers is revealed in the diagram. This is not what one would expect if attributions of responsibility were straightforwardly governed by some easily applied compatibilist or incompatibilist rule. What is clear is that some sort of explanation is needed that allows for significant interpersonal variation, and significant effects of seemingly irrelevant factors, such as the concreteness of the questions asked. 3. The Explanation Hypothesis In earlier papers, partly in collaboration with Karl Persson, I have argued that a wide variety of phenomena involving judgments of responsibility, including those mentioned above, can be given a unified explanation if we understand such judgments as a species of explanatory judgment (see Björnsson 2011; Björnsson and Persson 2009; 2012). More specifically, responsibility judgments see the object of responsibility as explained (in normal ways) by some relevant “motivational structure” of the agent, i.e. a motivational structure of a kind that is generally an appropriate target for practices of holding responsible (for our purposes here, we might think of these as structures that are responsive to reasons). So when we think that an agent is morally to blame for an act or event, we think that it happened because the agent didn’t care enough about morally important matters, or cared about the wrong things. Similarly, when we think that an act or event is to an agent’s moral credit, we think that it happened because the agent balanced morally relevant concerns in a good way. This might seem trite, but ordinary explanatory judgments are known to have a number of interesting features. Most importantly for our purposes, they are selective. If we are thinking about why some event E happened, we will focus only on one (or perhaps a few) events or conditions that were part of the causal prehistory of E, at the exclusion of others. If we are thinking about why a house burnt down, for example, we might focus on the fact that the house was hit by lightning, but not on the fact that the air contained oxygen, or on the fact that the house was built by combustible matter, or lacked a first-class sprinkler system. Though we understand that these other factors were necessary conditions for E, they are part of the explanatory background, as it were, typically because they are more generally expected and so less informative than the factors that we do focus on. Moreover, we naturally focus on factors that have a comparatively straightforward or familiar explanatory connection to E. Though we think that the lightning that hit the house had a causal prehistory—a separation of charges in the neighboring atmosphere, say—our focus will be on the lightning, as the lightning is causally related in a more straightforward and familiar way to the burnt down house than events leading up to the lightning. INCOMPATIBILISM AND “BYPASSED” AGENCY 6 Let us say that to focus on some factors as explaining E is to see these factors as the “significant” explanation of E. Then the following is our proposed account of responsibility judgments: THE EXPLANATION HYPOTHESIS: We take A to be responsible for X if we see some relevant motivational structure of A as (part of) a significant normal explanation of X. Elsewhere we detail how the Explanation Hypothesis and the selectivity of explanatory judgments might account for a number of features of responsibility judgments, including the fact that responsibility judgments display so-called side-effect asymmetries and are closely statistically correlated with explicit explanatory judgments (Björnsson 2011; Björnsson and Persson 2012). Many of these features are relatively disconnected from issues of incompatibilism. But there is a further aspect of the selectivity of explanatory judgments that we suspect explains the seeming force of standard skeptical arguments about moral responsibility as well as the results recounted in the previous section: the selection of explanatory factors is relative to explanatory interests and salient explanatory models. Though you and I might ordinarily focus on the lightning when thinking about why the house burnt down, a fire engineer might instead focus on the lack of a lightning rod, treating the fact that the house was hit by lightning as part of the explanatory background. Similarly, a politician thinking about the same event might focus on inadequate funding for the fire brigade, and a physicist on specific properties of the building materials. Because of different explanatory interests, they might relegate different factors to the explanatory background, and employ explanatory models relating different variables. And because of this, they will think of different things as the significant explanation of the event. Here is how the combination of this interest relativity of explanatory judgments and the Explanation Hypothesis might account for subjects’ general but non-universal reluctance to attribute responsibility to agents in deterministic scenarios (Björnsson and Persson 2012; 2013): First, people ordinarily attribute moral responsibility to agents on the basis of applying ordinary folk-psychological models, explaining actions and outcomes in terms of the beliefs and motivational structures of agents. Second, what deterministic scenarios do is to introduce abstract deterministic explanatory models saying that every event is causally determined by earlier events (back to the beginning of the universe). In such models, human motivational structures, deliberation and decision-making play no privileged role, being mere causal intermediaries and providing no independent input into the general unfolding of events. Given the Explanation Hypothesis, someone looking at things from the perspective of this explanatory model will not see agents as responsible for their actions. This explains the tendency towards incompatibilist judgments. INCOMPATIBILISM AND “BYPASSED” AGENCY 7 Third, although deterministic scenarios introduce abstract explanatory models, folkpsychological models might nevertheless be more salient for particular subjects, especially since the latter are central parts of our everyday explanatory repertoire. This explains why the incompatibilist tendencies are limited. Fourth, questions about responsibility asked about concrete cases are likely to activate folkpsychological models capable of explaining the specifics of such cases, at the expense of abstract deterministic models incapable of explaining any such particulars. This explains why subjects agree more with responsibility attributions in deterministic scenarios when these attributions concern concrete cases. This explanation might itself be accepted by defenders of incompatibilism and compatibilism alike: it tells us that incompatibilist intuitions stem from a certain kind of explanatory perspective rather than another, but does not tell us which perspective is correct. While I think that it ultimately supports a comprehensive error theory for central incompatibilist intuitions (Björnsson and Persson 2012: 345–8; Björnsson ms), the argument needed for such a conclusion are complex and predictably contentious. In comparison, the Bypass Hypothesis offered by Nahmias and Murray is much more straightforward. 4. The Nahmias and Murray Bypass Hypothesis N&M’s hypothesis, recall, is that when subjects take responsibility to be undermined in deterministic scenarios, this is largely because they take agents’ beliefs, desires and decisions to play no role in bringing about actions, i.e. because they take agents’ deliberative or agential capacities to be bypassed. Some early evidence for this hypothesis came from studies by Nahmias, Coates and Kvaran (2007), where subjects were quite willing to attribute moral responsibility when deterministic causation of actions were described in psychological terms, but more reluctant when it was described in neurological terms. In the latter sort of scenario, but not in the former, it would be possible for subjects to conclude that ordinary psychological processes were bypassed.2 The Bypass Hypothesis might also seem to explain why subjects in studies using the Nichols and Knobe paradigm, though prone to understand determinism as involving bypassing, would be less prone to make the mistake when considering a concrete case, and especially one describing the agent’s motivation. After all, our ordinary understanding of such cases takes those to involve deliberative capacities. Apart from having some initial plausibility, the Bypass Hypothesis is potentially highly significant. Since it is generally agreed that determinism does not imply that agents’ beliefs, 2 For my preferred explanation of this phenomenon, in terms of the Explanation Hypothesis, see Björnsson and Persson 2013: 626–32.) INCOMPATIBILISM AND “BYPASSED” AGENCY 8 desires and decisions are bypassed, it would be clear that incompatibilist folk intuitions are based on a mistake. Consequently, to the extent that pre-theoretical hunches and commitments account for stable intuitions and commitments among philosophers, incompatibilist theories of responsibility would also clearly rest on a mistake. To more directly test the Bypass Hypothesis, N&M (Nahmias and Murray 2010; Murray and Nahmias 2012) conducted a survey where subjects were randomly assigned to one of four conditions: the two conditions of the Nichols and Knobe experiment and two further conditions in which subjects read descriptions of deterministic Universe C, descriptions that N&M hypothesized would be less likely to give rise to bypass misinterpretations of determinism, one involving an abstract description on human agency, and another involving an agent, Jill, who steals a necklace.3 All in all, then, there were two abstract and two concrete conditions. After having read one of the four vignettes, subjects were asked to indicate agreement with the statements below on a 1-to-6 scale (strongly disagree, disagree, somewhat disagree, somewhat agree, agree, strongly agree). The three first statements attribute free will, moral responsibility, or desert of blame; the latter four are meant to measure bypass judgments, saying that agents’ beliefs, desires and decisions have no effect or that agents have no control over what they do. Subjects assigned to the abstract conditions read the first version of each statement; subjects assigned to concrete conditions read the version in parentheses: