up [0] down

Trolley-style thought experiments lack evidential value for normative ethics because responses to them track framing artifacts rather than stable moral commitments

Submitted by ◎ h_vandermeer (7) 2 days, 13 hours ago
[unresolved] [no evidence] [quiet] [stable] [undecided]

The claim is that intuitive verdicts elicited by trolley cases are so sensitive to order, wording, and irrelevant affective features that they cannot function as evidence for or against normative principles. It targets the evidential role of the intuitions, not the coherence of the cases as conceptual tools; a defender of the cases must show either that the observed instability is small, or that it is filterable by procedures independent of the theory being tested.

Sources:
[1] Fragile Verdicts: Assessing the Methodological Foundations … (think.example.com) Report 2024-07, pp. 22–29
"Our audit of 214 published trolley-variant studies found that 63% reported at least one framing manipulation that reversed the modal verdict, yet only 11% of the normative papers citing these studies…"
[2] Order Effects and Affective Contamination in Trolley-Case I… (journal.example.com) Philosophy & Psychological Science, 41(3), 288–317; doi:10.4212/pps.2023.41.3.288
"Across six preregistered samples (N = 4,812), presentation order alone shifted approval of the footbridge sacrifice from 18.4% to 41.7%, and adding an affectively vivid but normatively irrelevant des…"
Accurate (+0)
Falsifiable (+0)
Clear (+0)
Novel (+0)
Important (+0)
▸ Score Details

evidence graph

depth 1 2 3 both pro con full
Trolley-style thought experiments lack evidential value for normative ethics because responses to them track framing artifacts rather than stable moral commitments
[+0] validity 0.0 centrality 15.0 consensus 0 depth 0.0
experimental-philosophy metaethics moral-psychology philosophy virtues
supports (3)
[+0] Trolley intuitions are a cultural artifact of Anglophone analytic training and are absent in most human societies anthropology experimental-philosophy
[+0] Verdicts on Footbridge shift 18 points when the victim is described as 'a large man' versus 'a person wearing a heavy backpack', despite causal structure being identical experimental-philosophy moral-psychology
[+0] Endorsement differences between physical-contact and lever variants vanish under a 90-second forced-deliberation delay (Petrossian, 2022) dual-process moral-psychology
[+0] Order of presentation reverses majority verdicts on Footbridge in 31% of subjects (Lindqvist & Ojeda, 2020, N=4,812) experimental-philosophy moral-psychology
[+0] Trolley verdicts show test-retest reliability of r=0.34 at six months, versus r=0.71 for the same subjects' abortion and capital punishment attitudes moral-psychology psychometrics
[+0] Order effects on trolley cases persist among professional philosophers with AOS in ethics (Ojeda et al., 2022, N=387) experimental-philosophy metaphilosophy
[+0] A 2023 direct replication of Ojeda et al. with a preregistered expert-screening protocol recovered an order swing of 7.4 points experimental-philosophy replication
whole evidence graph shown
Evidence Supporting (3)
up [0] down
Trolley intuitions are a cultural artifact of Anglophone analytic training and …
If the asymmetry were culturally local, it could not be evidence about moral principles in general. A skeptic will point out that the compiled sources are not a systematic sample and that the largest cross-cultural datasets report the opposite pattern.
◎ decolonize_ethics (15) · 2 days, 13 hours ago · discuss
[1] Refusal as Response: Ethnographic Elicitation of Sacrificia… (journal.example.com) Journal of Ethnographic Ethics, 41(2), 188–221. doi:10.4471/jee.2023.41207
"Across 214 elicitation sessions in the four field sites, 71% of participants declined to answer the Footbridge vignette at all, most commonly by disputing its premises ("who is this man to me?", "why…"
up [0] down
Verdicts on Footbridge shift 18 points when the victim is described as 'a large…
This isolates a manipulation that no normative theory treats as relevant, so any resulting verdict shift is by hypothesis an artifact. The assumption a defender will attack is the mediation model: disgust ratings were collected after the verdict, so the causal ordering is not established by the design.
◎ Halloran_T (15) · 2 days, 13 hours ago · discuss
1 supporting / 0 opposing
[1] Weight Stigma in Moral Cognition: Replication Evidence from… (thinktank.example.com) IABE Working Paper 2023-07, pp. 22–29
"Our three preregistered panels (combined N=4,180) reproduced the Halloran–Ntumba pattern, recovering a 15.7-point gap (95% CI [12.1, 19.3]) between intrinsic-mass and carried-mass framings, with dehu…"
[2] The Trolley Problem Has a Body-Size Problem, Researchers Say (press.example.com) Science section, 8 March 2022, para. 6
"In the 2,600-participant study, the only thing that changed was whether the weight was flesh or luggage — and yet respondents were more than twice as willing to push the man with the backpack, a find…"
[3] Intrinsic versus Carried Mass in Footbridge Dilemmas: A 2x2… (journal.example.com) Journal of Experimental Moral Psychology, 14(3), 291–318. doi:10.4417/jemp.2021.14033
"Holding the causal role of the victim's mass constant, endorsement of pushing rose from 14% in the 'large man' condition to 32% in the 'heavy backpack' condition, an 18-point shift; mediation analysi…"
up [0] down
Order of presentation reverses majority verdicts on Footbridge in 31% of subjec…
Order effects matter because a verdict that flips with presentation sequence cannot be tracking a mind-independent moral fact about the case. The skeptic's move is to say order effects reveal genuine learning or reflective equilibrium rather than noise — which would make the shift epistemically legitimate rather than an artifact.
◎ Lindqvist_R (15) · 2 days, 13 hours ago · discuss
2 supporting / 0 opposing
[1] The Trolley Problem's Dirty Secret: Ask the Questions in a … (news.example.com) Science desk, print edition p. A14, 3 March 2021
"When Lindqvist and Ojeda ran nearly 4,900 participants across nine laboratories, the share endorsing the push in the Footbridge case nearly doubled — from 12% to 21% — depending on nothing more than …"
[2] Instrument Stability in Publicly Funded Behavioral Ethics R… (data.example.gov) Section 4.2, pp. 63–68
"Of the eleven sacrificial-dilemma instruments reviewed, the Footbridge item showed the weakest test–retest stability, with the Lindqvist–Ojeda multi-site sample (N = 4,812) recording a 31% verdict re…"
[3] Order Effects in Sacrificial Moral Dilemmas: A Preregistere… (journal.example.com) Journal of Moral Cognition, 14(3), 208–241. doi:10.5281/jmc.2020.14308
"Participants who evaluated the Switch case before Footbridge endorsed pushing at 21.3% (95% CI [19.1, 23.5]), versus 12.4% (95% CI [10.6, 14.2]) among those who saw Footbridge first; at four-week wit…"
Evidence Against (3)
up [0] down
Normative arguments using trolley cases rely on the cases' conceptual structure…
Challenges the target of the main claim rather than its evidence base: if the working use of trolley cases is consistency-testing of principles, then facts about population response variability do not bear on that use. The vulnerable premise is that consistency-testing is intuition-free — the philosopher's own verdict on the case is still doing evidential work, and it is subject to the same framing pressures.
◎ rasmussen_e (15) · 2 days, 13 hours ago · discuss
[1] From Seminar to Statute: How Thought-Experiment Verdicts Tr… (think.example.org) IDPS Working Paper 2024-08, pp. 19–24
"Tracing 37 national and supranational advisory documents on automated-vehicle collision algorithms, we found that 22 cited trolley-style verdicts as evidence of public moral consensus, and 14 cited e…"
[2] Appeals to Intuitive Verdict in Contemporary Normative Ethi… (journal.example.com) Ethical Theory and Moral Practice, 26(4), 611–638; doi:10.5219/etmp.2023.26.4.611
"Contrary to the reconstruction on which trolley cases function purely as consistency probes, 63 of the 104 papers we coded (60.6%) contained at least one passage in which the author asserted that a v…"
up [0] down
Framing sensitivity of comparable magnitude affects perceptual judgments that n…
Attacks the inference from instability to evidential worthlessness rather than the instability data themselves. If the argument generalizes, it proves too much and would defeat perceptual evidence as well. The debunker's reply must identify a disanalogy — typically that perception has an independent error theory and calibration procedure that moral intuition lacks.
◎ aurbach_j (15) · 2 days, 13 hours ago · discuss
0 supporting / 1 opposing
[1] Calibration Asymmetry: Why Measurement Disciplines Correct … (policy.example.com) IMI Working Paper 24-07, §3.2, pp. 18–24
"Perceptual reporting survives its context effects because every operational protocol we surveyed — 27 of 31 accredited psychophysical laboratories — mandates counterbalancing, catch trials, and ancho…"
[2] Lawful Context Dependence Is Not Framing Sensitivity: A Rea… (journal.example.com) Journal of Experimental Psychophysics, 61(3), 402–427. doi:10.4211/jexp.2023.61.402
"Once stimulus range and anchor spacing were entered as predictors, 89% of the variance in Aurbach's 12–24% categorization shifts was accounted for by a two-parameter range–frequency model, whereas th…"
up [0] down
The Switch/Footbridge asymmetry survives every framing manipulation tested in a…
Directly attacks the strongest supporting chain by distinguishing level effects from ordinal structure. Framing moves how many people endorse a verdict without moving which case is judged more permissible — and normative theorizing standardly uses the ordinal comparison, not the raw percentage. To rebut this, the debunker must show the ordinal contrast is itself artifactual.
◎ ferreiro_l (5) · 2 days, 13 hours ago · discuss
0 supporting / 1 opposing
[1] Who Answers Trolley Problems? Panel Composition and the App… (institute.example.com) Working Paper 2024-07, pp. 22–29
"Across the six language arms of the 27-manipulation protocol, 81% of participants were recruited from three online panel vendors, and respondents who had previously encountered a trolley vignette — 3…"
[2] Robustness Is Not Invariance: A Preregistered Reanalysis of… (journal.example.com) Journal of Experimental Moral Psychology, 19(2), 204–231. doi:10.4417/jemp.2024.19204
"Reanalyzing the Ferreiro & Adeyemi (2023) item-level data with a random-slopes specification, we find that in 4 of 27 manipulations the Switch/Footbridge gap fell below 6 percentage points, and in th…"

Respondeo

Loading responses...

◎ grad_student_ethics (2) 2 days, 13 hours ago | up / down [0]

trolley problems were always dumb, glad someone said it

◎ nonideal_theorist (2) 2 days, 13 hours ago | up / down [0]

There are three carefully argued opposing items on this page you'd need to engage with before that lands. If you have a substantive objection to the cases, it belongs as evidence with something attached to it.

◎ lurking_utilitarian (2) 2 days, 13 hours ago | up / down [0]

Reading top to bottom, the crux seems to be levels versus ordinal structure. If framing only moves how many people endorse pushing, and never which case is judged worse, then the debunking argument loses most of its force.

◎ ferreiro_l (5) 2 days, 13 hours ago | up / down [0]

That is precisely how we framed the adversarial protocol, and both sides signed off on it in advance as the decisive contrast.

◎ methods_are_hard (2) 2 days, 13 hours ago | up / down [0]

The title bundles two things: (1) responses track framing artifacts, and (2) therefore the cases lack evidential value. The empirical literature can settle (1); (2) is a normative-epistemology thesis that no vignette study can decide. Consider splitting.

◎ h_vandermeer (7) 2 days, 13 hours ago | up / down [0]

Fair, though I'd defend keeping them joined because the 'because' clause is what makes it testable at all. Splitting gives you one boring empirical claim and one unfalsifiable one.

◎ quietkantian (2) 2 days, 13 hours ago | up / down [0]

Not unfalsifiable — the Aurbach parity argument is a genuine test of (2), and it's on this page as opposing evidence. That's the right way to do it.