Studies 1 and 2 — Social Presence and Moral Decision-Making (EEG)
The first two studies examined whether being observed by another person changes how individuals make moral decisions. Participants completed a series of moral dilemmas — scenarios requiring a choice between two morally distinct courses of action — while their brain activity was recorded using EEG. In Study 1 (N = 83), the experimenter was positioned behind the participant, creating a passive monitoring condition. In Study 2 (N = 32), the experimenter was positioned beside the participant, increasing the salience and proximity of social observation. Both studies used identical stimuli and acquisition protocols, enabling direct comparison across conditions.
Across both studies, participants showed stable moral preferences regardless of whether they were being observed: the distribution of responses and the time taken to make decisions were not significantly altered by social presence. However, a more fine-grained analysis revealed something subtle and theoretically important: participants spent more time reading the moral scenarios when observed, but not more time actually deliberating. This suggests that social observation heightens attentional vigilance at an early stage of information processing — people read more carefully when watched — without penetrating the core of moral evaluation itself.
At the neural level, EEG analyses revealed that frontal midline theta oscillations — a well-established marker of cognitive conflict and executive control — were significantly elevated when participants made utilitarian choices compared to deontological ones. This finding confirms that utilitarian decisions, which require overriding an immediate emotional response, involve greater cognitive effort. Critically, this neural conflict signal was not modulated by social presence, providing converging evidence that the neurophysiological architecture of moral deliberation is robust to social observation. These results were further corroborated by drift-diffusion modelling, a computational approach to decision-making, which showed that neither the efficiency of evidence accumulation nor the decision threshold was affected by the presence of an observer. Together, the behavioural, computational, and neural findings from Studies 1 and 2 offer a comprehensive and methodologically rigorous account of how social monitoring relates to moral cognition.
A joint empirical manuscript reporting these findings is in preparation for submission to a specialised peer-reviewed journal in social and affective neuroscience.
Study 3 — Joint Moral Decision-Making and EEG Hyperscanning
The third study moved beyond passive social observation to examine genuine social interaction during moral decision-making. An original experimental paradigm was developed specifically for this project: pairs of participants simultaneously viewed identical moral dilemmas on separate screens and were required to reach a joint decision — without any verbal communication, interacting exclusively via keyboard and mouse. This non-verbal, digitally-mediated interaction design allows the study of social influence and consensus-building under precise experimental control, while preserving a meaningful and naturalistic form of joint deliberation.
The paradigm was programmed in Unity and validated across two pilot sessions prior to full data collection. A total of 23 pairs (N = 46 participants) were tested — exceeding the originally projected target of 22 pairs — with simultaneous EEG recorded from both participants throughout each session. At the end of each session, participants completed a semi-structured interview assessing their subjective experience of the interaction, perceived social influence, stress, and perceptions of leadership dynamics during the joint decision process. This mixed-methods design enriches the dataset beyond purely neurophysiological measures. A complete inter-brain coherence analysis pipeline was developed to quantify brain-to-brain synchrony between interacting partners. EEG analyses are currently ongoing.
Ecological Moral Dilemmas Stimulus Bank (EMDS)
In parallel with the experimental studies, a new scientific resource was developed to support more ecologically valid research on moral cognition. A large corpus of real-world moral content was assembled through systematic collection of publicly available text data targeting ethical topics across multiple thematic domains. This corpus was processed through a complete Natural Language Processing pipeline, including text preprocessing, lemmatisation, bigram and trigram modelling, and Latent Dirichlet Allocation topic modelling — an unsupervised machine learning approach that identifies recurring latent themes within large text collections. The pipeline assigns each document a dominant and secondary moral topic with associated probability scores, and is designed to support continuous data ingestion as new content is collected. The full pipeline is publicly available on GitHub, constituting a standalone open-science contribution that any researcher can use and build upon independently of the completed stimulus bank. Scenario construction and psychometric validation with human participants are currently underway.
Theoretical Contribution
Cutting across all three studies, a major theoretical and methodological manuscript was produced during the fellowship: Rethinking Human Morality in Real-Life Social Contexts (Vives & Caspar), currently under review (preprint available at
https://doi.org/10.31234/osf.io/umwvd_v1(s’ouvre dans une nouvelle fenêtre)). This paper argues that dominant models in moral psychology and neuroscience insufficiently account for social context, group dynamics, and ecological validity, and proposes actionable solutions for integrating naturalistic validity into experimental research — including a self-reflection tool for researchers across disciplines. This manuscript provides the theoretical framework underpinning the full research programme and positions its contributions within a broader call for methodological reform in the moral cognition field.
