Abstract
In a recent essay, Deery and Nahmias (Philos Stud 174(5):1255–1276, 2017) utilize interventionism about causation to develop an account of causal sourcehood in order to defend compatibilism about free will and moral responsibility from manipulation arguments. In this paper, we criticize Deery and Nahmias’s analysis of sourcehood by drawing a distinction between two forms of causal invariance that can come into conflict on their account. We conclude that any attempt to resolve this conflict will either result in counterintuitive attributions of moral responsibility or will undermine their response to manipulation arguments.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Though we focus on DN’s view of causal sourcehood in this essay, many of our criticisms apply to Usher’s (2018) account of teleological control as well, which we highlight in footnote 19. We also criticize one of Usher’s objections to DN in footnote 6. And, while we ultimately reject both DN’s and Usher’s accounts, this isn’t to say that no account of causal sourcehood or teleological control can be utilized to overcome manipulation arguments.
DN argue that CAS represents “…the features of an agent’s psychology that compatibilists typically judge as jointly (and minimally) sufficient for free will and moral responsibility” (DN 2017: 1258).
DN allow for more than one variable to bear the strongest invariance relation to an event and hence qualify as the causal source of that event. In such a case, each agent whose CAS is a causal source of an event would be in a position to be morally responsible for that event. They argue: “In cases in which the strength of invariance obtaining between Diana’s decision and Manny’s stealing is equal to that obtaining between the output of Manny’s CAS and his stealing, Diana and Manny may share equal responsibility” (DN 2017: 1263, footnote 8).
Interestingly, Usher (2018), whose response to manipulation arguments also relies on an interventionist approach to causation, defends a hard-line response to the Diana case. He claims that because Diana acts 30 years before the theft occurs, “…if we are to make any intervention on background circumstances (say bringing about rain on the day of the theft), there is nothing that Diana could do to “ensure” that Manny steals the wallet” (2018: 19). But this is to mistake a change in background conditions with an intervention. To assess whether Diana is an actual cause, we consider an intervention—a surgical change in the value of her decision—and ask whether this would change the value of the variable representing the theft. Once we’ve established that Diana is an actual cause, we can assess the invariance of the causal relation by asking under what range of background conditions it will continue to obtain. To do this we consider a possible world in which some non-causal features are different—say, it’s raining at t30—and ask whether the causal relation still obtains. That is, we consider whether an intervention on Diana’s decision in that world successfully predicts whether or not the money is stolen. In the present case, it seems that it would—Diana would simply implement her plan somewhat differently to ensure that Manny isn’t dissuaded by the rain. Thus, to us, DN’s claim that Diana’s decision is the most invariant cause of Manny’s stealing the money is plausible. If this is so, then (given his prior commitments) Usher should agree with DN that Diana’s presence at least mitigates Manny’s responsibility.
In light of DN’s soft-line response, a proponent of a manipulation argument may revise the case such that Diana cannot ensure, or perhaps doesn’t even intend, that Manny will go on to steal the money. Given these revisions, it’s unlikely that Diana’s meddling would relate most invariantly to Manny’s stealing the money. In response to this modified version of the Diana case, DN argue that they would take the hard-line: if Diana’s decision does not relate most invariantly to Manny’s stealing the wallet, then Diana cannot be the causal source of this event and Manny, like Danny, is in the position to be morally responsible for the theft (2017: 1273). However, as DN note, many defenders of manipulation arguments are quick to make such modifications to their manipulation cases, replacing the manipulator with a spontaneously generated machine or force field (Pereboom 2001, 2014; Mele 2005). Given this, it’s hard to see how DN’s account provides a more successful and systematic response to manipulation arguments than other soft-line responses. But, rather than object to DN on these grounds, we will focus on their positive account of causal sourcehood.
Woodward (2003) also runs these together, but in later work distinguishes them.
A causal relation which meets DN’s condition 1 supports many counterfactuals of a certain type (those that hold fixed the background conditions), hence, it is reliable. Here we part company with Woodward (2007), who refers to this simply as invariance. For us, invariance is the genus of which stability and reliability are species.
We’ve rephrased DN’s account both to distinguish between reliability and stability and to make clear that multiple variables can each be causal sources of an event. That is, causal sourcehood admits of ties (see footnote 5).
For ease of exposition, we will write in terms of agents’ decisions as opposed to the outputs of agents’ CASs.
One might argue that the boss’s decision could take a wider range of values than we claim here and still predict the value of the murder. We address this possibility in Sect. 4.3.
It is, of course, difficult to determine the appropriate range of values for variables such as these. The range of values the variables can take will be sensitive to the level of description one adopts, which depends on a variety of contextual factors. While this fact makes it difficult to develop a case in which one agent’s decision clearly and uncontroversially relates more reliably to an event than another agent’s decision, it also complicates matters for accounts of causal sourcehood that invoke reliability, especially those like DN’s that are meant to ground free will and moral responsibility.
A defender of the reliability-centric view might attempt to accommodate this case by arguing that causal sourcehood is only a requirement for direct moral responsibility, and it’s possible to be indirectly morally responsible for events to which one only causally contributes. In this way, the mafia boss, though he isn’t the causal source of the murder, can still be indirectly morally responsible for it. Additionally, the mafia boss would also be directly responsible for his decision to hire an assassin, which could warrant just as much, if not more, blame as the assassin's decision to murder the enemy. This may make the reliability-centric view more palatable, but it fails to address the underlying worry raised by the case, namely: Even if causal sourcehood is only a requirement for direct responsibility, why should we favor reliability over stability in its determination? What is it about reliability that renders the assassin directly responsible for the murder of the enemy while the mafia boss is only indirectly responsible? The problem of determining causal sourcehood simply recurs at the level of direct moral responsibility.
In general, reliability indicates the limits of a causal relationship. Consider, for example, Hooke’s law: F = −kx, where x is the elongation of a spring, k is a constant characteristic of the spring, and F is the restoring force exerted. Hooke’s law expresses a genuine causal relation between elongation and force, but nevertheless breaks down at a certain value of x—one can only stretch a spring so far. By contrast, it is often supposed that the laws of (ideal) fundamental physics should be perfectly reliable (Woodward 2007: 76–80).
Indeed, Usher, who defends his own stability-centric view of causation, characterizes DN’s position as only concerned with stability (2018: footnote 14).
One may deny this and maintain that agents aren’t directly responsible for outcomes like car crashes, but rather, they are only (directly) responsible for their own acts or omissions (e.g., driving distractedly). Though there may be many merits to this view, it will be difficult for DN to accept it, since their framework is designed to determine responsibility for such outcomes. Recall that DN’s central claim is that non-manipulated agents like Danny can be responsible for outcomes (e.g., the stealing of a wallet) while manipulated agents like Manny cannot. Moreover, there is good reason to think the problems we raise in this section will recur at the level of the acts and omissions for which agents can be (directly) responsible. For example, it is in-principle possible for an agent’s distractedness to be the result of two causal variables—one that lies outside her CAS and relates most stably to it and one that lies within her CAS and relates most reliably to it. After all, distracted drivers are often distracted by something, but the mere existence of the thing by which agents are distracted ought not render agents un-responsible for their actions and omissions (provided some feature of their CAS is an actual cause), even if this distracting object relates more stably to the agents’ distractedness than any other causal variable.
The above reasoning counts against Usher’s (2018) stability-centric account of responsibility as well. Usher argues that causal responsibility for an event depends on an agent’s beliefs and intentions relating stably to the event in question. Usher goes on to argue that if an agent possesses the relevant CAS, then the degree to which an agent is causally responsible determines the degree to which she is morally responsible as well. But, as we’ve seen above, the relationship between stability and moral responsibility is not so straightforward. An agent can be morally responsible to a high degree for an action even if her intentions and beliefs don’t relate to it in a particularly stable way. Furthermore, another agent, who possesses the relevant CAS, can be not at all morally responsible for an event even though she relates to it more stably than any other causal variable.
Developing an account of causal sourcehood that can make sense of moral responsibility as coming in degrees may prove to be a promising strategy. Historically, theories of moral responsibility have focused almost exclusively on the threshold of moral responsibility, but this is changing (e.g., Coates and Swenson 2013; Nelkin 2016). Usher (2018) also takes it to be a significant virtue of his interventionism-inspired view of teleological control that it makes room for degrees of responsibility. And, those developing models of causal responsibility in computer science and cognitive science have also focused on modeling degrees of responsibility (e.g., Chockler and Halpern 2004; Lagnado et al. 2014). However, these latter accounts are careful to distinguish between causal responsibility and moral responsibility. In fact, Chockler and Halpern argue that agents can be blameworthy without being causally responsible (2004: 95).
There are two ways this could go. First, one could adopt a stability-centric view of causal sourcehood, which would have as a consequence that the pothole is the causal source of the crash, for the reasons discussed above. Second, one could argue that the pothole is both most stably and most reliably related to the crash, and thus unambiguously the causal source. Perhaps the physical dimensions and the location of the pothole, under a wide range of values, would still predict the occurrence of the crash. In this case, the causal relation between the pothole and the crash has a good claim to being the most reliable.
The same can be said, mutatis mutandis, for a version of the truck driver case in which the truck driver’s behavior is deemed the causal source of the crash.
Perhaps the boss meets the stability condition to a greater degree than the assassin meets the reliability condition or perhaps they each meet their respective conditions to the same degree, but stability is given greater weight in determining overall invariance.
If one finds this scalar approach to causal sourcehood and responsibility promising, one could adopt it to develop a hard-line response to manipulation arguments.
One could stipulate a kind of “additivity” requirement that degrees of causal sourcehood must sum to 1. If this were the case, then the presence of Diana would make Manny considerably less responsible, as she would take up much of the sourcehood, and hence, responsibility. This claim runs contrary to much of the philosophical literature on responsibility. For example, Zimmerman (1985) argues that multiple agents can each be fully responsible for the outcomes of group actions. Additionally, experimental work indicates that the folk do not comply with an additivity requirement when they assess responsibility (Lagnado et al. 2014). Finally, such a principle would make much of DN’s argument unnecessary. The mere presence of a novel cause, with even the most modest degree of sourcehood, would serve to differentiate Manny from Danny. This fails to capture the key feature of the case—that Manny isn’t the source of his stealing the money.
References
Campbell, J. (2010). Control variables and mental causation. Proceedings of the Aristotelian Society,110, 15–30.
Cartwright, N. (1979). Causal laws and effective strategies. Noûs,13(4), 419–437.
Chockler, H., & Halpern, J. (2004). Responsibility and blame: A structural-model approach. Journal of Artificial Intelligence Research,22, 93–115.
Coates, J., & Swenson, P. (2013). Reasons-responsiveness and degrees of responsibility. Philosophical Studies,165(2), 629–645.
Deery, O., & Nahmias, E. (2017). Defeating manipulation arguments: Interventionist causation and compatibilist sourcehood. Philosophical Studies,174(5), 1255–1276.
Fischer, J. M. (2000). Responsibility, history, and manipulation. Journal of Ethics,4(4), 385–391.
Fischer, J. M., & Ravizza, M. (1998). Responsibility and control. Cambridge: Cambridge University Press.
Frankfurt, H. (1971). Freedom of the will and the concept of a person. Journal of Philosophy,68(1), 5–20.
Ismael, J. T. (2013). Causation, free will, and naturalism. In A. Kincaid, J. Ladyman, & D. Ross (Eds.), Scientific metaphysics (pp. 208–235). New York: Oxford University Press.
Ismael, J. T. (2016). Why physics makes us free. New York: Oxford University Press.
Lagnado, D., Gerstenberg, T., & Zultan, R. (2014). Causal responsibility and counterfactuals. Cognitive Science,37(6), 1036–1073.
McKenna, M. (2008). A hard-line reply to Pereboom’s four-case argument. Philosophy and Phenomenological Research,77(1), 142–159.
Mele, A. (1995). Autonomous agents: From self-control to autonomy. New York: Oxford University Press.
Mele, A. (2005). A critique of Pereboom’s ‘Four-Case’ argument for incompatibilism. Analysis,65, 75–80.
Mele, A. (2013). Manipulation, moral responsibility, and bullet biting. Journal of Ethics,17(3), 167–184.
Nelkin, D. (2016). Difficulty and degrees of moral praiseworthiness and blameworthiness. Noûs,50(2), 356–378.
Pearl, J. (2009). Causality. Cambridge: Cambridge University Press.
Pereboom, D. (2001). Living without free will. Cambridge: Cambridge University Press.
Pereboom, D. (2014). Free will, agency, and meaning in life. New York: Oxford University Press.
Roskies, A. (2012). Don’t panic: Self-authorship without obscure metaphysics. Philosophical Perspectives,26(1), 323–342.
Sartorio, C. (2016). Causation and free will. New York: Oxford University Press.
Usher, M. (2018). Agency, teleological control and reliable causation. Philosophy and Phenomenological Research. https://doi.org/10.1111/phpr.12537.
Wolf, S. (1990). Freedom within reason. New York: Oxford University Press.
Woodward, J. (2003). Making things happen: A theory of causal explanation. New York: Oxford University Press.
Woodward, J. (2006). Sensitive and insensitive causation. Philosophical Review,115(1), 1–50.
Woodward, J. (2007). Causation with a human face. In A. Price & R. Corry (Eds.), Causation, physics, and the constitution of reality: Russell’s republic revisited (pp. 66–105). Oxford: Oxford University Press.
Zimmerman, M. (1985). Sharing responsibility. American Philosophical Quarterly,22(2), 115–122.
Acknowledgements
We’d like to thank Derk Pereboom and an anonymous referee for their very helpful comments and suggestions on earlier versions of this essay. We’d also like to thank the participants of the “Free Will and Laws of Nature” workshop, hosted at the University of Cologne and organized by Andreas Hüttemann and Christian Loew, as well as the audience at the University of Sydney’s Current Projects Seminar, for their excellent questions and feedback.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Tierney, H., Glick, D. Desperately seeking sourcehood. Philos Stud 177, 953–970 (2020). https://doi.org/10.1007/s11098-018-1215-3
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-018-1215-3