Eros Quarta1, Stefano Grasso1, Virginia Papagni1, Irene Lacal1,2, Alexandra Battaglia-Mayer1 1Department of Physiology and Pharmacology, Sapienza University of Rome, Piazzale Aldo Moro, 00185 Rome Italy; 2current address: Cognitive Neuroscience Laboratory, German Primate Center - Leibniz-Institute for Primate Research, 37077 Göttingen, Germany
Introduction.Elucidating the mechanisms driving cooperation is of great interest for advancing knowledge on both biological and artificial intelligence[1,2]. Joint action (JA) paradigms, in which two or more individuals share an intention and coordinate their actions to achieve a common goal, help study cooperation in controlled lab settings in humans and non-human primates [3-5]. However, the computational [6] and neuronal underpinnings [7,8] of JA are largely uncharted. The aim of this work is two-fold: i) to probe if and how macaques monkeys perform cost-benefit computations when deciding between acting alone or together with another agent, and ii) to explore how decisions and movements are encoded in their neural activity. Methods.We developed a novel paradigm that enabled us to investigate both decisions and movements dynamics from two macaques as these were engaged in a behavioral task in which each took turns choosing between individual (SOLO) and joint (JA) actions to obtain varying rewards (n=>10k trials). Using a joystick, each monkey controlled a cursor toward one of two targets which indicated both the reward magnitude and the action type (SOLO or JA) required to obtain the desired reward. During JA, monkeys had to coordinate movements so as to keep their inter-cursors distance below a threshold. Alongside behavioral data, we simultaneously collected single-unit activity (n=366 units) from the prefrontal cortex (PFC, area 9) of the two interacting monkeys.
Results. Logistic models of monkeys' choices revealed the subjective cost of cooperation and spatial biases influencing their decisions. Macaques preferred SOLO, yet opted for JA under favorable payoff conditions, indicating that voluntary cooperation can emerge following cost-benefit computations. Over time, their choice policy integrated JA more frequently, suggesting that macaques dynamically estimate both costs and benefits of cooperation. Most PFC neurons were modulated when deciding between SOLO and JA and when planning and executing either action. Notably, a subset of neurons encoded the partner's probability of cooperating. These findings unveil PFC correlates for costs-benefits analysis and good-to-action transformation in a social context. Discussion. Our findings unveiled that monkeys perform cost-benefit computations and show biases when deciding whether to act alone or in cooperation with a partner. Combined with the preliminary results on the neuronal dynamics when engaging in voluntary cooperation, our work provides a foundational step toward unveiling the rules of engagement for cooperation in the primate brain. In addition, this study offers an experimental basis for developing biologically-inspired computational models, devising artificial agents, and constructing neural networks to investigate cooperation in both physiological and artificial systems.
This study was supported by the Italian Ministry of University and Research (grant Young Researcher 2024 N. B83C24003140006 to EQ and grant PRIN 2022 Prot. N. 2022B9NZC4 to ABM). p { line-height: 115%; text-align: left; orphans: 2; widows: 2; margin-bottom: 2.47mm; direction: ltr; background: transparent }strong { font-weight: bold }a:link { color: #000080; text-decoration: underline }
1 McNally, L. et al. Proceedings of the Royal Society B: Biological Sciences (2012) 2 Bolotta, S. et al. Frontiers in Computer Science (2022) 3 Vesper, C. et al. Front Psychol (2017) 4 Lacal, I. et al. Cortex (2022) 5 Visco-Comandini, F. et al. Cortex (2015) 6 Vicariis, C. D. et al. PLOS Computational Biology (2024) 7 Ferrari-Toniolo, S. et al. J. Neurosci. (2019) 8 Pezzulo, G. et al. Progress in Neurobiology (2022)
