Ethical compromise, valuing life, and Fechner's Law

Abstract

Ethical compromise, valuing life, and Fechner's Law Authors: Nigel Harvey, Xiaozhou TanAffiliation: Department of Experimental Psychology, University College London (UCL)Principal Investigator: Prof. Nigel HarveyLicence: Creative Commons Attribution 4.0 International (CC BY 4.0)DOI: https://doi.org/10.5281/zenodo.21940136 Overview This repository contains the behavioural datasets from four online experiments investigating moral reasoning in sacrificial dilemmas. All experiments use a variant of the trolley problem (framed as an incoming-missile scenario) in which participants decide whether to sacrifice a number of people in order to save others. The central manipulated quantity is the kill/save (k/s) ratio — the number of people sacrificed relative to the number saved (e.g. sacrificing 2 people to save 1 corresponds to a k/s ratio of 2). Across experiments the ratio is varied on different scales (linear and logarithmic/geometric) to examine how the magnitude and scaling of sacrifice influence moral judgements. The data allow examination of how moral acceptability judgements change as a function of the k/s ratio, the scale on which the ratio increases, and related response measures including reaction times. Data collection All data were collected online using the Gorilla Experiment Builder (gorilla.sc). Most files are therefore in Gorilla's raw export format, which includes a standard block of session and task metadata columns followed by study-specific response columns. See CODEBOOK.md for full documentation of both. Provenance note: The Experiment 1 data were collected by a member of the research lab who is not an author of this deposit. Experiments 2–4 were collected by the authors. All files are provided as raw exports except where noted; no analytic transformations have been applied to the response data. Repository structure Experiment1/ expt1consent.csv Consent-form task export (one row per participant) expt1demographics.csv Demographics task export expt1linear2.csv Dilemma task, linear ratio schedule (variant "2") expt1linear114.csv Dilemma task, linear ratio schedule (variant "114") expt1log.csv Dilemma task, logarithmic/geometric ratio schedule (2^n) expt1reactionTimes.csv Trial-level reaction-time export (one row per trial) Experiment2/ expt2demographics.xlsx Demographics (custom schema, keyed on Gorilla ID) expt2R1.xlsx ... R10.xlsx Dilemma task, one file per randomiser branch (10 conditions) Experiment3/ expt3data.csv Dilemma task, trial-level (wide format) expt3demographics.csv Demographics task export Experiment4/ expt4data.xlsx Dilemma task, trial-level (wide format) expt4demographics.xlsx Demographics task export Experiment-specific notes Experiment 1. The three dilemma files share an identical column structure and differ only in how the number of people sacrificed scales across items: linear2 — number sacrificed increases linearly in steps of 2. linear114 — number sacrificed increases linearly in steps of 114. log — number sacrificed increases geometrically (2^n). Each dilemma file stores responses as repeating Question 1 / Question 2 / Question 3 object-N Value triplets, one triplet per dilemma item in a single wide per-participant row. Within each item the three questions ask (1) whether the participant would sacrifice the N people to save 1, (2) what percentage of 100 other people would do so, and (3) what a perfectly ethical robot would do — each answered as a percentage likelihood (0–100). This three-question format is shared across Experiments 1–3. See CODEBOOK for the object-N scaling detail. Experiment 2. The dilemma task is split across ten files, expt2R1–expt2R10, one per randomiser branch (condition). The files share an identical column structure and differ only in which condition's participants they contain. Demographics are in a separate workbook using a compact custom schema (pid, age, gender, residence, nationality); the pid field contains Gorilla participant IDs and is the key for joining demographics to the response files. Experiments 3 and 4. Each is a single wide, trial-level file containing Gorilla Spreadsheet: (stimulus/design) columns, Store: (accumulated participant response) columns, and Manipulation: columns. Experiment 3 uses the same three-question percentage-likelihood response format as Experiment 1. Experiment 4 uses a two-step response: participants first choose between sacrifice and inaction, then judge the odds they would choose their selected option versus the alternative (the sacrifice/inaction and …i-suffixed response variables; see CODEBOOK Section E). Joining demographics to response data Within each experiment, demographic and response files are linked by the Gorilla participant identifier (Participant Private ID in the raw exports; pid in the Experiment 2 demographics workbook). Participant Private ID is Gorilla's internal anonymous key and is recommended as the linking variable. Important format note (Gorilla exports) In the raw Gorilla exports, the first data row (row 2) contains the question wording, not participant data. For example, the first value under Age object-9 Value is the literal text "What is your age (in years)?"; genuine responses begin on the following row. Filter out this metadata row before analysis. Ethics and participant privacy The datasets involve human participants and were collected under the relevant institutional ethics approval [insert approval reference]. Data have been prepared for public deposit in accordance with applicable ethics approval and data-protection requirements. To protect participant privacy, the following columns were removed from the raw Gorilla exports before deposit: Participant Public ID, Schedule ID, Participant Completion Code, Participant External Session ID, and the device columns (Device Type, Device, OS, Browser, Monitor Size, Viewport Size). Participant Private ID (Gorilla's internal anonymous key) is retained as the cross-file linking variable. Timing columns and all demographic variables (age, gender, country of residence, nationality) are retained as study data. See CODEBOOK Section F for the complete, finalised list. Citation Harvey, N., & Tan, X. (2026). Ethical compromise, valuing life, and Fechner's Law. University College London. Zenodo. https://doi.org/10.5281/zenodo.21940136 Licence This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) licence. You are free to share and adapt the material for any purpose, including commercially, provided you give appropriate credit, provide a link to the licence, and indicate if changes were made. Contact For questions regarding permissions or use beyond the terms of this licence, please contact Prof. Nigel Harvey, Department of Experimental Psychology, University College London.