Playing with Trolleys: Intuitions About the Permissibility of Aggregation

• Published date: 01 June 2014
• DOI: http://doi.org/10.1111/jels.12039
• Date: 01 June 2014

Playing with Trolleys: Intuitions About the

Permissibility of Aggregation

Mark Kelman and Tamar Admati Kreps*

We explore when experimental subjects think aggregation across persons—deciding that

some parties should be worse off, so that others might gain more, compared to an alternative

option—is permissible through subjects’ responses to trolley problem vignettes. Two classic

vignettes are: (1) whether to divert a runaway trolley on to a spur track, killing one, to save

multiple potential victims on a main track, and (2) whether to push an overweight person off

a bridge to block the trolley from hitting the potential victims. Subjects typically judge

diverting in the first problem permissible and pushing in the second problem impermissible.

We first show that respondents reject Taurek’s claim that the number of victims is irrelevant.

Our central novel finding is that the intuition to distinguish the two cases is weaker than the

extensive literature emphasizing its significance has implied. Responses to the two problems

often converge to a significant extent on the position that aggregation is impermissible

because the intuition to divert is both more weakly held and more subject to revision than

the intuition against pushing. We show this convergence when the problems are presented

together. Moreover, while the strictures against pushing are typically seen as mandatory (and

hold even when following the strictures disadvantages kin or more identified parties), the

obligation to divert is rarely unqualifiedly mandatory, and even its permissibility is sensitive

to potential harm to kin and identifiable people.

I. Background and Introduction

A person is considering which of two distinct courses of action to take (call them A and B).

If she chooses A, some individual X will benefit while some other individual or group Y will

not. If she chooses B, then losses to X will be traded for larger gains to Y: compared to A,

X will be worse off, but Y will receive more benefits than X would have received from A.

What is the moral status of aggregating the gains and losses that X and Y will experience in

deciding which course of action to take? Is the subject permitted to follow course of action

B, trading off losses to X for gains to Y, so long as the aggregated gains to Y are greater than

the losses to X—or is action B barred or A affirmatively mandatory, meaning we should

ignore Y’s interests? Or is course of action B, perhaps not only permissible but desirable or

even obligatory? Deontologists and utilitarians diverge in their responses to these types of

problems: deontologists often believe it is impermissible to choose B because X has some

*Address correspondence to Mark Kelman, James C. Gaither Professor of Law and Vice Dean, Stanford Law School,

Crown Quadrangle, 559 Nathan Abbott Way, Stanford, CA 94305; email: mkelman@stanford.edu. Kreps is Ph.D.

candidate at Stanford Graduate School of Business.

bs_bs_banner

Journal of Empirical Legal Studies

Volume 11, Issue 2, 197–226, June 2014

197

inviolable side-constraining right, and utilitarians believe it is obligatory to choose B, so

long as the welfare gains to Y are greater than the losses to X.

For more than 30 years, psychologists and philosophers have carefully considered one

particular variant of this kind of problem. They have studied both how people in fact react,

and how they ought to react, to a series of trolley problem vignettes in which they judge

whether people are morally permitted (or even obliged) to take distinct actions that would

result in the death of a single party, but also save multiple lives. Descriptive psychologists have

tried to explain why people do not and many normative philosophers have tried to explain

why they should not invariably follow the most straightforward act-utilitarian dictate to

minimize the number of lives lost, and why they sometimes feel it is either permissible or

obligatory to take steps that sacrifice some individual’s interests to minimize the aggregate

loss of life. We predominantly explore these intuitions in the trolley problem settings in

which they have most typically been raised—settings in which the party is considering actively

harming an already-identified party to save other already-identified parties from harm.1

Conventional trolley problems were first discussed by normative philosophers who

explored the following issue: Is it—and, if so, why is it—more permissible for a bystander to

divert a runaway train from a main track where it will kill five people to a spur track where

it will kill but one (Spur Track), than to push a fat man to his death off of a drawbridge

overhanging the train track if doing so will block a runaway train and prevent five deaths

(Drawbridge)? Most philosophers addressing this issue claim they are trying both to

describe and to justify the widespread moral intuitions that surveys using these two prompts

have revealed2(e.g., Kamm 1989:227; Montmarquet 1982:440; Thomson 1985:1395–96).

1We make relatively brief reference as well to experiments that we discuss in much more detail in Kelman and Kreps

(2012). What we found in these additional studies is that while the intuition to aggregate in diverting cases relative

to pushing cases is weaker than the literature implies, we further find that intuitions against aggregating in both

pushing and diverting cases are far less firm and unbending than most philosophers and psychologists recognize. A

more robust distinction than the pushing/diverting distinction is that between both of these “killing” dilemmas and

ones in which subjects contemplate saving more rather than fewer people. Saving problems, whether considered ex

post (when considering whether to save more or fewer people already in peril) or ex ante (when considering adopting

safety measures designed to prevent either more or fewer subsequent fatalities), elicit subjects’ most basic and durable

intuition by far: in these savings cases, aggregation is plainly permissible, often desirable, but rarely completely

mandatory. When subjects confront these savings cases alongside killing cases, they are no less prone to aggregate in

the savings cases but markedly more prone to aggregate in the killing cases.

2Even setting aside the problem that people may not react to hypothetical paper-and-pencil moral conundrums in

the same way that they might react to “similar” problems they confronted in real life, many have noted that these

hypothetical cases are “impure” in any number of ways, even when experimental subjects are directed to ignore or

wish away the relevant confounds. Subjects may not really believe the fat man can block the trolley so they may believe

the person in Drawbridge will end up killing one more person than would otherwise die, rather than reducing the

number who die (Greene et al. 2009:366–69); subjects responding to the Drawbridge prompt may not really believe

that the person who is pushed can do a better job blocking the train than the person who pushes, so may wonder if

it is permissible to sacrifice another person rather than oneself (Huebner & Hauser 2011:74); little attention is paid

in either problem to the possibility that the five people on the main track are blameworthy in being there (e.g.,

Thomson 1976:210, 212; Wood 2011:74–77); all the problems are posed as if outcomes are certain to occur even

though it is more plausible that outcomes described as certain to occur are better pictured as more or less probable,

thus making the problems seem more distinct than they arguably should from commonplace social choice problems

(e.g., whether to permit production techniques that increase cancer risks for those near factories) in which we

regularly risk others’ lives for gains far more trivial than the chance to save another set of lives (Fried 2012a, 2012b,

2012c; Norcross 1997).

198 Kelman and Kreps