Skip to main content
SpringerLink
Log in

Agency Laundering and Information Technologies

  • Published:
Ethical Theory and Moral Practice Aims and scope Submit manuscript

Abstract

When agents insert technological systems into their decision-making processes, they can obscure moral responsibility for the results. This can give rise to a distinct moral wrong, which we call “agency laundering.” At root, agency laundering involves obfuscating one’s moral responsibility by enlisting a technology or process to take some action and letting it forestall others from demanding an account for bad outcomes that result. We argue that the concept of agency laundering helps in understanding important moral problems in a number of recent cases involving automated, or algorithmic, decision-systems. We apply our conception of agency laundering to a series of examples, including Facebook’s automated advertising suggestions, Uber’s driver interfaces, algorithmic evaluation of K-12 teachers, and risk assessment in criminal sentencing. We distinguish agency laundering from several other critiques of information technology, including the so-called “responsibility gap,” “bias laundering,” and masking.

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

Similar content being viewed by others

Notes

  1. Note that a number of commentators believe the story makes too close a connection between predictive analytics and pregnancy-related advertising. There are reasons to send such advertising to people who are not pregnant, the advertising may have been based on a criteria unrelated to pregnancy, and others (Harford 2014).

  2. Our account of the structure of responsibility follows closely those articulated by Nicole Vincent (2011) and Chris Kutz (2004). Both Vincent and Kutz recast Hart’s ship captain case to distinguish various facets of responsibility.

  3. Antony Duff calls this “prospective”responsibility (Duff 1998). Here we should note that we are only discussing morally justifiable roles, where the holders of role responsibility are themselves moral agents. Hence, being assigned a role within a criminal organization, or being assigned a role when one lacks the capacity to act morally, cannot confer role responsibility in the required sense.

  4. See (Vincent 2011; Williams 2008; Goodin 1986, 1987)

  5. For the sake of simplicity, we will refer to ‘actions’ in discussing responsibility. However, our account extends to omissions and dispositions. Note, too, that causal responsibility is complicated in over-determination cases. But those cases don’t affect our analysis here.

  6. This, of course, may not absolve the captain completely. See Fischer and Ravizza 2000, 49–51 for an explanation of “tracing” responsibility to prior actions.

  7. There remain some controversial issues, including for example Frankfurt-style cases in which one may be responsible or not regardless of whether she does or does not know how her actions will be causally effective. But the issues in those cases turn on the link between causal responsibility and the ability to do otherwise. That does not affect our arguments.

  8. Within this group of views, there is substantial debate about whether person X is responsible for Y in virtue of Y being attributable to X, of X being answerable for Y, or of X being accountable for Y. Scanlon’s view focuses on attributability (Scanlon 2008). Shoemaker distinguishes between attributability, answerability, and accountability (Shoemaker 2011). Smith (like Shoemaker) distinguishes a thing being attributable to a person and that person being responsible for it; however, she views accountability as a species of answerability. What is important for our purposes is that each of the views in this constellation recognizes that the content of responsibility claims is that responsible agents are those for whom it is appropriate, or for whom it ought to be the case, that they provide an account of their intentions, interests, and reasons for an action.

  9. Fischer and Ravizza provide an accountability view that bridges Strawson’s attention to the social function of holding others responsibility by way of reactive attitudes and accountability views’ attention to reasons. Specifically, they maintain that an agent is responsible if she is an apt target of reactive attitudes. More important here, though, is that being morally responsible for actions requires that agents exercise “guidance control.” That requires that agents be at least weakly reasons-responsive, which is to say that where the agent has access to strong reasons in favor or against an action, she will act in accordance with those reasons. It also requires that the source of actions be the agent, which is to say that the reason-responsiveness is internal to the agent (Fischer and Ravizza 2000, 31–41).

  10. 18 U.S. Code § 1956 - Laundering of monetary instruments.

  11. Other aspects of money laundering are about concealing identities of agents, for example by routing illicit money through shell corporations and bank accounts in permissive jurisdictions.

  12. Two other accounts addressing causal and moral responsibility in the computing context are worth noting here. First, Daniel Dennett (1997) posits that machines may be credited with (i.e., responsible for) some tasks (e.g., Deep Blue beating Kasparov) but cannot be responsible for others (e.g., murdering Kasparov). We would argue that this difference tracks the causal/moral responsibility distinction, though that is not Dennett’s claim.

    Helen Nissenbaum (1994) argues that the increased use of computing systems poses a threat to accountability, based on four key barriers. These include the problem of many hands, the existence of bugs that cause computing failures, the ability to use computers as scapegoats, and the separation of system ownership from legal liability for problems. In doing so she notes that distributed causal responsibility can function to obscure responsibility and blameworthiness (p. 74). Our view of laundering can apply to each of the barriers she discusses, but does not depend on any of them. Consider the example of “blaming the computer,” or pointing to the computer as the sole source of causal responsibility. That considered by itself would not seem to be a case of laundering, but instead just a straightforward denial of responsibility. If instead, it included a process by which a party ensures the computer has causal responsibility, ascribes morally relevant qualities to the computer’s actions, obscures the party’s causal responsibility, and in so doing fails to adequately account for events for which the party is morally responsible, it could be laundering. In other words, merely blaming something else does not rise to laundering. Laundering is, we take it, more insidious in that it forestalls others’ abilities to demand an account of actions within domains of their legitimate concern.

    Thanks to an anonymous reviewer for pointing us to these articles.

  13. Note that agency laundering does not require that one infringe one’s substantive role responsibilities (except to the extent that one’s role responsibility includes being transparent about one’s causal responsibility). In Chair, for example, it’s plausible that the chair was fulfilling her role responsibilities with respect to the department’s curriculum. We return to this point in section 4.

  14. Note that Facebook has recently taken measures aimed at reducing discriminatory advertising (Levin 2019).

  15. Note that this is a possible moral claim that one might make about Facebook and other media organizations. This is a distinct question from what kinds of legal rights and obligations information intermediaries have in light of (inter alia) the United States Communications Decency Act (see 47 U.S.C. section 230), the European Union’s eCommerce Directive, articles 12–16 (Council Directive 2000/31, 2000 O.J. (L 178) 1 (EC)), and the EU’s General Data Protection Regulation (GDPR) (Commission Regulation 2016/679, 2016 O.J. (L 119) 1 (EU)). See, e.g., Keller (2018)

  16. See, e.g., (Facebook 2018)

  17. One can frame this as a question of capacity responsibility. That is, if Facebook did not have epistemic access to the relevant information about the possibility of misuse, it would not have the necessary capacity to be morally responsibility. Note here that epistemic access is not limited to actual knowledge, but the ability to garner it under reasonable conditions. Hence, Facebook’s moral responsibility will turn on the degree to which it could reasonably have known about potential for misuse. And that would define its degree of agency laundering.

    One further complicating issue is mitigation. Facebook or another social media company might use its suggestion system to better understand relations among (for example) racists or purveyors of disinformation to promote anti-racist or epistemically sound information. The degree to which that would mitigate or deepen laundering is a question beyond what we can cover here. Thanks to an anonymous reviewer for making this point.

  18. One tool, named “Greyball,” was developed to surreptitiously ban users who Uber believed were violating the company’s terms of service. Uber eventually used Greyball to surreptitiously ban people Uber believed to be government regulators investigating whether Uber was operating illegally. See (Isaac 2017).

  19. Defendant’s Original Answer and Defenses, Houston Federation of Teachers, Local 2415 v. Houston Independent School District (S.D. Tex.),” n.d., 9 citing (Sanders et al. 2009)

  20. Note that Loomis demonstrates another way one can launder even while fulfilling one’s substantive role responsibilities. Imagine that the trial court had deliberated about its decision, but did not explain its reasoning for the sentence. Suppose instead it merely wrote that it agreed with the COMPAS report’s assessment with no further comment. That would obscure the scope of the court’s causal responsibility and would fail to provide an adequate account of the decision. But in that case, the court would not have violated some other substantive role responsibility.

  21. Thanks to a reviewer for pointing out these possibilities and noting their similarities to “Chair” and “Democratic Chair.”

  22. Thanks to Suresh Venkatasubramanian for pointing us to this talk.

  23. We appreciate an anonymous reviewer raising the question of whether there are cases where one maintains agency but launders accountability instead. Our sense is that any such case would involve minimizing one’s agency. In other words, accountability is the thing that is avoided, and one avoids it by laundering the degree to which one is (morally) responsible, which is in turn a function of a person’s agency in a process. Likewise, money laundering is a way to forestall accountability, and it is the laundering of some other thing (viz., money) that helps avoid the accountability.

  24. Thanks to an anonymous reviewer for pressing us on this point.

  25. Thanks to an anonymous reviewer for this language and description. In this signaling respect, our use of a metaphor here works similarly to the “crumple zone” metaphor in Elish (2019), discussed in section 8.2.

  26. Note that there may be other, non-laundering moral wrongs involved in goal prompts and queuing, as discussed in section 5.

  27. Regulation (EU) 2016/679, of the European Parliament and the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation), 2016 O.J. (L 119).

References

Download references

Acknowledgements

We would like to thank audiences at Eindhoven Technological University, Delft Technological University, Twente Technological University, the Zicklin Center for Normative Business Ethics at the Wharton School of Business,University of Pennsylvania, the iConference, and the Privacy Law Scholars Conference. In particular, we would like to thank Richard Warner, Filippo Santoni Del Sio, Sven Nyholm, Owen King, Philip Jansen, Suresh Venkatasubramanian,and three anonymous reviews for this journal for their enormously helpful comments.

Author information

Authors and Affiliations

  1. University of Wisconsin-Madison, Madison, WI, USA

    Alan Rubel & Adam Pham

  2. Florida International University, Miami, FL, USA

    Clinton Castro

Authors
  1. Alan Rubel
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Clinton Castro
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Adam Pham
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Alan Rubel.

Ethics declarations

Disclaimers

We presented a paper discussing the idea of agency laundering at the 2019 iConference (March 31 – April 4, 2019, Washington, D.C.) which appears in the conference proceedings (Rubel et al. 2019). The account of agency laundering in this paper is expanded and has changed substantially from the account in those proceedings. Most importantly, the conference paper did not tie agency laundering to an account of responsibility, and instead invoked concepts of de facto and de jure power. The conference paper did not include many of the examples used here (e.g., Democratic Chair, Uber, HISD). The conference paper did discuss the Facebook and Loomis examples, but the treatment has changed in accord with the changes to the account of laundering.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Rubel, A., Castro, C. & Pham, A. Agency Laundering and Information Technologies. Ethic Theory Moral Prac 22, 1017–1041 (2019). https://doi.org/10.1007/s10677-019-10030-w

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10677-019-10030-w

Keywords

Search

Navigation