An open and reproducible investigation into the factors that influence judgements of intention and responsibility utilising the Side-Effect Effect

Abstract

One phenomenon in moral decision-making is known as the Side-Effect Effect (SEE; Knobe, 2003), an asymmetry in which participants rate more blame for harmful side effects than they do praise for helpful side effects. Furthermore, participants are significantly more likely to state that harmful side effects are brought about intentionally compared with those that are helpful. The original theoretical explanation for the SEE was that observers utilise the moral valence of actions’ unintended consequences to determine their deserved praise and blame. The SEE has been previously demonstrated to be a robust effect that can be observed across languages, cultures, and age groups. Several theories have been developed to attempt to explain these judgements, including the Norm Violation Model (NVM), which is based on transgressions against social norms, and the Deep-Self Concordance Model (DSCM), which is based on comparisons between a person’s values and the alignment of their actions with these values. This thesis aimed to contribute to the theoretical understanding of the SEE by conducting a series of iterative, within-subjects, multi-item experiments that manipulated factors related to the agent (job role status, gender, power level, and environment). These experiments will elucidate which theoretical models may best explain how the SEE is produced. All experiments were preregistered and employed several open and reproducible research practices. Following a successful conceptual replication (Experiment 1), it was found that the mere mention of the agent’s job role status (Experiment 2) and gender (Experiment 3) were insufficient to modulate the SEE. However, when their level of work-related power was more explicit and salient, more blame was afforded to low-power versus high-power agents for harmful side effects, in opposition to predictions (Experiment 4). A further exploration of work-related power found, contrastingly, participants affording more blame to high-power agents as well as affording more blame in general within the workplace versus outside of it (Experiment 5). There was evidence that participant individual differences may impact the assignment of blame. These mixed results lend support for both the Norm Violation Model and the Deep-Self Concordance Model; however, neither was able to best explain the results in isolation. A new framework for explaining the SEE is proposed that encompasses both the NVM and the DSCM, as well as accounting for the relationship between a participant and the agent. The framework offers a foundation for future research to continue building upon as further elements of the SEE (and moral decision-making generally) are explored. The contributions of this thesis are supported by several methodological and analytical advances, improving the quality, transparency, reproducibility, and credibility of the research.