RESEARCH PROJECTS
During the past few years, our lab has produced original findings in the fields of decision neuroscience. We have provided ground breaking neurocomputational accounts of how the brain makes social/moral decisions. In addition, we have identified different brain mechanisms of how hormones influence reward processing and decision making and we have described the brain computations underlying dysfunctions of decision making in impulse control disorders (e.g., hypersexuality in Parkinson’s disease, pathological gambling...).
Axis 1: Social decision making
A) Computations required for group decisions and tracking fluctuating cooperative/competitive intentions
To make decisions in social contexts, humans must predict the behavior of others, an ability that is thought to rely on having a model of other minds, known as “theory of mind” (ToM). This becomes especially complex when one interacts with groups of people and actions are anonymous. We showed that a Bayesian model based on partially observable Markov decision processes outperforms existing models in quantitatively predicting human behavior and outcomes of group interactions to resolve a group decision-making task (Khalvati et al., Science Advances, 2019; Khalvati et al., NIPS, 2019). Our results suggest humans use Bayesian inference to model the "mind of the group," making predictions of others' decisions while also simulating the effects of their own actions on the group's future dynamics. We have also revealed the neurocomputational mechanisms at play during such group decision making (Park et al., Nat. Comm., 2019). We found that ventromedial prefrontal cortex encodes immediate expected rewards as individual utility, while the lateral frontopolar cortex encodes group utility (i.e., pending rewards of alternative strategies beneficial for the group). During social interactions, it is frequently difficult to evaluate fluctuating intentions of others, such as whether they are willing to cooperate or compete. The ToM processes involved in resolving such situations were poorly characterized from a computational point of view. We have described the neurocomputational mechanisms required when it is necessary to arbitrate between unsignaled competitive or cooperative intentions during dyadic interactions (Philippe et al., Nat. Comm., 2024) and we have also identified the impact of competitive interactions on future cooperation (Li et al., Scientific Reports, 2023). Extending our social interactions studies to larger groups/networks of individuals, we are currently developing a new line of research on the psychological underpinnings of the spread of misinformation (Guigon et al., Comm. Psychol., 2024).
B) Neural bases of social dominance
Social dominance is deeply rooted in the biology and evolutionary history of our brains and most social species can assess it in several ways. First, animals can rely on dominance cues (e.g., body size, facial features of dominance ...) to rapidly assess the strength of potential competitors and to avoid costly physical conflict. Second, animals can learn about dominance hierarchies through observational learning. That is, by monitoring the social interactions of other individuals within the group, they can infer dominance relationships. Third, animals can learn about dominance relationships directly through competitive dyadic interactions against rivals (experiencing successive victories or defeats against competitors). We have studied these different types of social dominance representations. In particular, we have demonstrated that the rostromedial prefrontal cortex (rmPFC) computes a signal updating the relative rank between individuals (Ligneul et al., Current Biology, 2016). Building on this original finding, we were the first team in Lyon to use the new simultaneous fMRI-PET scanner, showing that the serotonin transporter regulates rmPFC activity during social hierarchy learning (Janet et al., Neuropsychopharmacology, 2022). fMRI-PET scan also revealed the relationships between serotonin availability and frontolimbic responses to fearful and threatening faces (Janet et al., Scientific reports, 2023). Moreover, we have demonstrated the causal role of the dorsolateral prefrontal cortex in learning social ranks by observation (Qu et al., Comm. Biology, 2024). In addition, we showed that toddlers' are sensitive to dominance traits from faces (Galusca et al., Scientific Reports, 2023) and that the late positive potential is sensitive to dominance levels (Miao et al., Scientific Reports, 2022). We also showed that aversion to social rank reversal depends on whether rank asymmetry is fair or unfair (Foncelle et al., Adaptive Human Behavior and Physiology, 2022).
Axis 2: Moral decision making
We have proposed a new account of neurocomputational mechanisms engaged in moral decision making (Lockwood et al., Ann. Rev. Psychol. 2025; Qu et al., Neurosci. and BioBehav Rev., 2021). Although the brain system engaged in moral decisions has been studied since the early days of cognitive neuroscience, mainly using moral dilemma, the neurocomputational mechanisms describing how the human brain makes moral decisions and learns in various moral contexts are only starting to be established. We have described moral choices at a mechanistic level, extending the conceptual framework of value-based decision making to encompass that of moral decision making. Moral dilemma can be modeled as value-based decisions that weigh self-interests against moral costs/harm to others and different types of prediction errors can be distinguished in various aspects of moral learning. These key computational signals help to describe moral choices and moral learning at an algorithmic level and reveal how these computations are implemented in the brain. For example, we have identified the neurocomputations of corruption in power holders and demonstrated that they can be modulated through transcranial direct current stimulation (Hu et al., elife, 2021; Hu et al., Psychol. Sci., 2022; Qu et al., SCAN, 2020). We have also shown that different brain systems are engaged when deciding whether to earn money by contributing to a ‘bad cause’ and when deciding whether to sacrifice money to contribute to a ‘good cause’, when such choices were made privately or in public (Qu et al., Plos Biology, 2019, Hu et al., J. Neurosci., 2021). These findings reveal how the brain processes three sources of motivation when weighing moral decisions: extrinsic rewards, moral values and concerns for one’s image. We have also studied how intrinsic brain morphology and functional connectivity impinge on the propensity to trust others (Feng et al., Human Brain Mapping, 2020) and the effects of social distance on both altruistic behavior between an individual and their social entourage (Liu et al., Front. Psychol. 2023) and third-party punishment, when an individual (as punisher) observes third parties from their social entourage violating social norms (Tang et al. Scientific Reports, 2023). Such fMRI studies lie at the intersection between social and moral decisions in so far as they reveal conflicts between social ties and moral decisions.
Axis 3: Hormonal influences on reward processing and decision making
We have extended our previous research on gonadal steroid hormones to study the effects of hormones (ghrelin and leptin, known to regulate of food intake) on the brain regions engaged in the regulation of different food rewards. We have identified the cognitive and hormonal regulation of appetite for food presented in the olfactory and visual modalities in the scanner (Janet et al., Neuroimage, 2021). We showed that ghrelin is a neurobiological marker for up-regulation success in the dorsomedial prefrontal cortex in a condition in which participants could indulge themselves (they were asked to smell the food odor and to adopt thoughts that would increase their desire to eat the presented food). In another study on food reward and impulsivity, we developed a new virtual reality system to show that food rewards that are presented in apparent close proximity increase impulsive action (O’Connor, iScience, 2021) and that the supplementary motor area is causally necessary for such spatial impulsivity (Carpio et al., Scientific Reports, 2024). Finally, we extended our previous work on gonadal steroid hormones and decision making to the domain of moral decisions. We showed that endogenous testosterone is associated with increased striatal response to audience effects during prosocial choices (Li et al., Psychoneuroendocrinology, 2020). We compared hormonal responses in behavioral addictions versus substance abuse (Li et al., Progress in NeuroPsychopharmacology & Biological Psychiatry, 2020) and critically evaluated the impact of hormone therapy (HT) on prefrontal structures and functions at menopause (Li and Dreher, Climateric, 2021). Our review suggests that HT, when taken early, may have beneficial effect on reward processing and cognitive control mechanisms.
Axis 4. Neurobiology of diseases
We have studied pathological gambling, Parkinson’s disease and high functioning autism. For example, we have demonstrated altered orbitofrontal sulcogyral patterns in gambling disorder (Li et al., Translational Psychiatry 2019) and revealed hormonal responses in pathological gambling (Li et al., Prog. in NeuroPsychopharmacol. & Biol Psychiatry, 2020). Impulse control disorders include a wide spectrum of behaviors, such as hypersexuality, pathological gambling or compulsive shopping. For example, we have determined the brain systems involved during delay-discounting of erotic rewards in Parkinson’s disease patients with hypersexuality (PD+HS), compared to PD without hypersexuality (PD-HS) and controls (Girard et al., Brain, 2019). Our findings point to reduced delay discounting of erotic rewards in PD+HS, both at the behavioral and brain system levels, and an abnormal reinforcing effect of levodopa when PD+HS patients. Finally, we have opened a new line of research in our group on autism. Previous investigations have found an altered pattern of moral behaviors in individuals with autism spectrum disorder (ASD), which is closely associated with functional changes in the right temporoparietal junction (rTPJ). However, the specific neurocomputational mechanisms at play, that drive the altered function of the rTPJ in moral decision-making, remained unclear. We demonstrated that rTPJ underlies avoidance of moral transgression in ASD (Hu et al., J. Neurosci., 2021). A selectively reduced rTPJ representation of information concerning moral rules was observed in ASD participants. These findings deepen our understanding of the neurobiological roots that underlie atypical moral behaviors in ASD individuals.
The human brain in the digital age: Human-AI interactions in a connected word
Our current work delves into the dynamic interplay between individual choices and the behavior of larger groups or networks. We aim to uncover how emerging collective behaviors influence our personal decisions and vice versa. This includes questions such as:
- What brain mechanisms drive the decision to share information in social networks, including to spread fake news?
- What are the brain computations underlying how we adapt when we need to infer the intentions of others in groups or networks, such as whether they intend to collaborate or compete?
- What are the algorithms used by the brain for strategic interactions engaging different levels of depth of mentalization? How can these algorithms be used to develop artificial agents to solve problems between many interacting agents?
- What learning rules does the brain use to process and integrate information propagating in social networks?
- How can we best use bots in hybrid human-AI social networks to encourage group collaboration?
- How does the morality (or immorality) of those around us shape our own moral behavior?
- What roles do hormones like oxytocin, stress hormones, and gonadal steroids play in how we learn about and navigate relationships within social networks?
To answer these questions, we take a transdisciplinary approach, drawing on models from artificial intelligence, machine learning, game theory, and computational social neuroscience. Our research bridges fields such as behavioral economics, human-computer interaction, psychology, and neuroscience.
We employ a variety of methods, including: behavioral experiments conducted in the lab and online; computational modeling (including machine learning) of decision-making processes; model-based fMRI to understand the nature of the computations performed by specific brain regions; causal manipulations (pharmacological, Transcranial Direct Current Stimulation, Transcranial magnetic stimulation) to temporarily disrupt the function of specific social brain regions; and, intracranial recordings in humans.
Our ultimate goal is to develop computational models that explain the brain’s mechanisms for navigating the reciprocal and dynamic interactions between individuals and the vast social networks they belong to, offering breakthroughs into the neural underpinnings of social behavior in a connected world