Maier, Maximilian; (2025) Mapping the Moral Mind: Computational & Formal Approaches to Understanding Prosocial Behaviour. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text PhD_Thesis_MM.pdf - Accepted Version Download (6MB) | Preview

Abstract

Many challenges in the 21st century (e.g., addressing climate change) are related to (failures of) moral and prosocial decision-making. In this thesis, I investigate how humans make these decisions, drawing on data from over 25 million individuals (5,934 from experimental studies and the rest from meta-analyses) and tackle two key methodological bottlenecks: publication bias and (lack of) computational modelling. In Part I (Chapters 1--3), I use novel statistical techniques, which I developed to adjust for publication bias in meta-analysis, to critically evaluate the evidence for Construal Level Theory, the Identifiable Victim Effect, and nudging. I document strong publication bias in all three of these areas, with meta-analytic mean effects not distinguishable from zero after adjusting for this bias. Part II of the thesis offers a more constructive counterpoint by developing novel paradigms and computational models. Chapter 4 examines scope insensitivity, evaluates the Unit Asking Method as an intervention to address it, and proposes an improved Sequential Unit Asking technique. Chapter 5 investigates decisions under extinction risk (e.g., death or human extinction) in individual and collective decisions using a novel experimental paradigm. I derive optimal strategies for these types of decisions and use (dependent) mixture modelling to describe individual-level behaviour. I find that, while participants are relatively good in terms of the qualitative strategies employed, their decisions are nevertheless affected by behavioural biases. Chapter 6 studies how people learn to make moral decisions using a novel reinforcement learning paradigm. Drawing on research on strategy selection learning and comparing different reinforcement learning models, I show that learning about strategies (moral rules or cost-benefit reasoning) rather than specific behaviours predicts generalisation to incentive-compatible donation decisions. Overall, this thesis demonstrates limitations in existing approaches while highlighting how formal and computational modelling, and publication-bias-adjusted meta-analysis can advance the understanding of moral and prosocial decision-making.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item