Skip to main content
SpringerLink
Log in

The Lesson of Bypassing

  • Published:
Review of Philosophy and Psychology Aims and scope Submit manuscript

Abstract

The idea that incompatibilism is intuitive is one of the key motivators for incompatibilism. Not surprisingly, then philosophers who defend incompatibilism often claim that incompatibilism is the natural, commonsense view about free will and moral responsibility (e.g., Pereboom 2001, Kane Journal of Philosophy 96:217–240 1999, Strawson 1986). And a number of recent studies find that people give apparently incompatibilist responses in vignette studies. When participants are presented with a description of a causal deterministic universe, they tend to deny that people are morally responsible in that universe. Although this suggests that people are intuitive incompatibilists, Eddy Nahmias and Dylan Murray, in a recent series of important papers, have developed an important challenge to this interpretation. They argue that people confuse determinism with bypassing, the idea that one’s mental states lack causal efficacy. Murray and Nahmias present new experiments that seem to confirm the bypassing hypothesis. In this paper, we use structural equation modeling to re-examine the issue. We find support instead for an incompatibilist explanation of the bypassing results, i.e., incompatibilist judgments seem to cause bypassing judgments. We hypothesize that this phenomenon occurs because people think of decisions as essentially indeterministic; thus, when confronted with a description of determinism they tend to think that decisions do not even occur. We provide evidence for this in three subsequent studies which show that many participants deny that people make decisions in a deterministic universe; by contrast, most participants tend to allow that people add numbers in a deterministic universe. Together, these studies suggest that bypassing results don’t reflect a confusion, but rather the depth of the incompatibilist intuition.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

Similar content being viewed by others

Notes

  1. “Abstract” here is to be contrasted with “Concrete”. In “Concrete” cases a specific individual engaging in a particular behavior is described while in “Abstract” cases no particular individual or behavior is described. Some concrete cases tend to elicit more compatibilist judgments (see Nahmias et al. 2007; Nichols and Knobe 2007).

  2. One way this can happen is when M and Y are highly correlated. In Murray and Nahmias’ study 1 (MS, p. 11), for example, they report a very high correlation between Bypassing and MR/FW, r = −.734.

  3. Following both Nichols and Knobe (2007) and Murray and Nahmias (forthcoming), the moral responsibility question was always presented first. The presentation of all the other questions were randomized to control for order effects.

  4. The average inter-correlations between the four items composing Bypassing (Cronbach’s Alpha = .936) and the average inter-correlations between the three items composing MR/FW (Cronbach’s Alpha = .936) were quite high, thereby indicating a high degree of internal consistency among the items composing each variable type.

  5. http://www.phil.cmu.edu/projects/tetrad/

  6. Essentially, GES considers the possible models available given the different variables. Each variable is assigned to a node, and the nodes are used to build the different possible models. GES begins by assigning an information score to the null model in which the nodes are all disconnected. GES then considers various possible arrows (“edges”) between the different nodes. The algorithm will add the edge that yields the greatest improvement in the information score (if there is such an edge). The algorithm repeats the process, adding the next edge that makes for the greatest improvement in the information score. When the algorithm reaches the point where no new edges improve the information score, it proceeds to consider deleting edges. It first finds the edge for which deleting that edge would yield the greatest improvement in the information score (if there is such an edge). It repeats this procedure until no further deletions will improve the score. In all cases, the orientation of the edges is given by edge-orientation rules (Meek 1997). It has been shown (Chickering 2002) that, given enough data, GES will return the true causal model of the data. GES is often interpreted as returning the best fitting causal model, given the data. (For further details, see Chickering (2002) and Rose et al. (2011).)

  7. This model was a very good fit of the data, χ 2 (1) = .9026, p = .3421, BIC = −3.1918.

  8. χ 2 (1) = 10.3052, p = .0013, BIC = 6.2108.

  9. The model returned is a good fit of the data, X 2 (1) = .8, p = .7772, BIC = −4.0143.

  10. X 2 (1) = 18.2393, p < .0001, BIC = 14.1450.

  11. χ 2 (1) = .1453, p = .7031, BIC = −3.9491.

  12. χ 2 (1) = 14.6579, p = .0001, BIC = 10.5636.

  13. X 2 (4) = .6660, p = .9555, BIC = −15.7114.

  14. We began by following the procedure outlined in Baron and Kenny (1986) for testing mediation:

    • Step 1: A regression model with Reasoning as a predictor of Bypassing was significant, t(63) = −2.215, Beta = −.271, p = .030.

    • Step 2: A logistic regression model with Reasoning as a predictor of Nonexistence was significant, Wald X 2 (63) = 8.062, B = −1.821, p = .005.

    • Step 3: A regression model with Nonexistence as a predictor of Bypassing was significant, t(63) = 5.078, Beta = .543, p = .000.

    • Step 4: When Reasoning and Nonexistence were both included in a multiple regression model, the effect of Reasoning on Bypassing was not significant, t(63) = −.671, Beta = −.078, p = .505.

    Note that the Baron and Kenny steps are only the beginning. When using only continuous data and a binary predictor variable X, one can simply test mediation using the standard Baron and Kenny procedure along with a Sobel (1982) test. But if either the mediator variable M or outcome variable Y are dichotomous, one ends up using logistic and linear regression analyses and so the coefficients for the dichotomous and continuous variables are in different scales and thus we can’t simply “mash them into a z-test” (Iacobucci, p.8, 2012). To handle this, the coefficients and standard errors need to be rescaled. We will defer a detailed discussion of this procedure (though the interested reader should see Herr 2013 for details; also see Iacobucci 2012) and simply note that we followed the procedure outlined in Herr (2013). We found that the reduction in the effects of Reasoning on Bypassing when Nonexistence was included in the model was significant Z = −2.3932, p = .016. This suggests that Nonexistence fully mediates the effects of Reasoning on Bypassing.

  15. Again, we began by following the procedure outlined in Baron and Kenny (1986) for testing mediation:

    • Step 1: A logistic regression model with Reasoning as a predictor of Nonexistence was significant, Wald X 2 (63) = 8.062, B = −1.821, p = .005.

    • Step 2: A regression model with Reasoning as a predictor of Bypassing was significant, t(63) = −2.215, Beta = −.271, p = .030.

    • Step 3: A logistic regression model with Bypassing as a predictor of Nonexistence was significant, Wald X 2(63) = 12.625, B = .742, p = .000.

    • Step 4: When Reasoning and Bypassing were both included in a multiple logistic regression model, the effect of Reasoning on Nonexistence was still significant, Wald X 2 (63) = 5.071, B = −1.688, p = .024.

    Since the effect of Reasoning remained significant even when Bypassing was included in the model, step 4 was not satisfied. Nonetheless, we followed the procedure in Herr (2013) for rescaling, finding that the reduction in the effects of Reasoning on Nonexistence when Bypassing was included in the model was not significant Z = −1.8322, p = .067. This suggests that Bypassing does not mediate the effects of Reasoning on Nonexistence.

  16. As is standard, binary variables can be interpreted as continuous when using GES. But, we note that this only works if the binary variables are given values of 0 and 1 when interpreted as continuous. Thanks to Joe Ramsey for guidance on this.

  17. This model was a very good fit of the data, χ 2 (1) = .4635, p = .4960, BIC = −3.6954.

  18. χ 2 (1) = 5.280, p = .0216, BIC = 1.1211.

  19. An important precedent for this result comes from unpublished work by Jim Sias. He described a deterministic universe to participants and asked whether it made sense to say that the agent in the scenario was able to make decisions. Sias found that participants tended to deny that it made sense (Sias unpublished).

  20. We thank a referee for pointing out the asymmetry and for prompting us to run this additional task.

References

  • Baron, R.M., and D.A. Kenny. 1986. The moderator-mediator variable distinction in social psychological research: conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology 51: 1173–1182.

    Article  Google Scholar 

  • Chan, H. forthcoming. Incompatibilist about what?

  • Chickering, D. 2002. Optimal structure identification with greedy search. Journal of Machine Learning Research 3: 507–554.

    Google Scholar 

  • Deery, O., Bedke, M. and Nichols, S. forthcoming. “Phenomenal Abilities: Incompatibilism and the Experience of Agency.” In D. Shoemaker (ed.), Oxford Studies in Agency and Responsibility.

  • Herr, N. 2013. Mediation with Dichotomous Outcomes. http://www.nrhpsych.com/mediation/logmed.html

  • Holton, R. 2009. Determinism, self-efficacy, and the phenomenology of free will. Inquiry: An Interdisciplinary Journal of Philosophy 52(4): 412–428.

    Google Scholar 

  • Iacobucci, D. 2012. Mediation analysis and categorical variables: the final frontier. Journal of Consumer Psychology. doi:10.1016/j.jcps.2012.03.006.

    Google Scholar 

  • Iacobucci, D., N. Saldanha, and X. Deng. 2007. A mediation on mediation: evidence that structural equation models perform better than regressions. Journal of Consumer Psychology 17(2): 140–154.

    Article  Google Scholar 

  • Kane, R. 1999. Responsibility, luck, and chance: reflections on free will and indeterminism. Journal of Philosophy 96: 217–240.

    Article  Google Scholar 

  • Knobe, J. (forthcoming). Free Will and the Scientific Vision.

  • Meek, C. 1997. Graphical Models: Selecting causal and statistical models. PhD Thesis, Carnegie Mellon University.

  • Mele, A. (forthcoming). Free will and substance dualism: The real scientific threat to free will?

  • Murray, D. & Nahmias, E. (forthcoming). Explaining Away Incompatibilist Intuitions. Philosophy and Phenomenological Research.

  • Nahmias, E., D.J. Coates, and T. Kvaran. 2007. Free will, moral responsibility, and mechanism: experiments on folk intuitions. Midwest Studies in Philosophy 31: 214–242.

    Article  Google Scholar 

  • Nahmias, E., S. Morris, T. Nadelhoffer, and J. Turner. 2006. Is incompatibilism intuitive? Philosophy and Phenomenological Research 73: 28–53.

    Article  Google Scholar 

  • Nahmias, E., & Murray, D. 2010. Experimental philosophy on free will: An error theory for incompatibilist intuitions. In New Waves in Philosophy of Action, ed. by J. Aguilar, A. Buckareff, and K. Frankish. Palgrave-Macmillan, 189–215.

  • Nelkin, Dana. 2004. Deliberative alternatives. Philosophical Topics 32: 215–240.

    Article  Google Scholar 

  • Nichols, S., and J. Knobe. 2007. Moral responsibility and determinism: the cognitive science of folk intuitions. Noûs 41: 663–685.

    Article  Google Scholar 

  • Pereboom, D. 2001. Living without free will. Cambridge: Cambridge University Press.

    Book  Google Scholar 

  • Pereboom, Derk. 2008. A compatibilist account of the epistemic conditions on rational deliberation. The Journal of Ethics 12(3): 287–306.

    Article  Google Scholar 

  • Rose, D., J. Livengood, J. Sytsma, and E. Machery. 2011. Deep trouble for the deep self. Philosophical Psychology 25(5): 629–646.

    Article  Google Scholar 

  • Roskies, A., and S. Nichols. 2008. Bringing moral responsibility down to earth. Journal of Philosophy 105(7): 371–388.

    Google Scholar 

  • Sarkissian, H., A. Chatterjee, F. De Brigard, J. Knobe, S. Nichols, and S. Sirker. 2010. Is belief in free will a cultural universal? Mind & Language 35: 346–358.

    Article  Google Scholar 

  • Sias, J. unpublished. Decisions, decisions: A study of folk intuitions.

  • Sobel, M.E. 1982. Asymptotic intervals for indirect effects in structural equations models. In Sociological methodology, ed. S. Leinhart, 290–312. San Francisco: Jossey-Bass.

    Google Scholar 

  • Sommers, T. 2012. Relative justice. Princeton: Princeton University Press.

    Google Scholar 

  • Strawson, G. 1986. Freedom and belief. Oxford: Clarendon.

    Google Scholar 

  • Van Inwagen, P. 1983. An Essay on Free Will. Oxford University Press.

Download references

Acknowledgments

We are grateful to David Danks, Joshua Knobe, Dylan Murray, Eddy Nahmias, and three anonymous referees for helpful comments on a draft of this paper.

Author information

Authors and Affiliations

  1. Department of Philosophy, Rutgers University, Brunswick, NJ, USA

    David Rose

  2. Department of Philosophy, University of Arizona, Tucson, AZ, USA

    Shaun Nichols

Authors
  1. David Rose
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Shaun Nichols
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to David Rose.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Rose, D., Nichols, S. The Lesson of Bypassing. Rev.Phil.Psych. 4, 599–619 (2013). https://doi.org/10.1007/s13164-013-0154-3

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s13164-013-0154-3

Keywords

Search

Navigation