Self-play constitutes a fundamental paradigm for autonomous skill acquisition, whereby agents iteratively enhance their capabilities through self-directed environmental exploration. Conventional self-play frameworks exploit agent symmetry within zero-sum competitive settings, yet this approach proves inadequate for open-ended learning scenarios characterized by inherent asymmetry. Human pedagogical systems exemplify asymmetric instructional frameworks wherein educators systematically construct challenges calibrated to individual learners' developmental trajectories. The principal challenge resides in operationalizing these asymmetric, adaptive pedagogical mechanisms within artificial systems capable of autonomously synthesizing appropriate curricula without predetermined task hierarchies. Here we present Heterogeneous Adversarial Play (HAP), an adversarial Automatic Curriculum Learning framework that formalizes teacher-student interactions as a minimax optimization wherein task-generating instructor and problem-solving learner co-evolve through adversarial dynamics. In contrast to prevailing automatic curriculum learning methodologies that employ static curricula or unidirectional task selection mechanisms, HAP establishes a bidirectional feedback system wherein instructors continuously recalibrate task complexity in response to real-time learner performance metrics. Experimental validation across multi-task learning domains demonstrates that our framework achieves performance parity with SOTA baselines while generating curricula that enhance learning efficacy in both artificial agents and human subjects.

Virtual agents with social characteristics such as warmth and competence, which are perceived readily from the face, impact human behavior effectively. However, generating such social signals is challenging because faces are complex, high-dimensional information spaces of 3D shape and complexion. We address this challenge using a data-driven psychology-based method that can objectively model socially impactful face features from subjective human perception. Specifically, we modeled the 3D shape and 2D complexion face features that drive four key social perceptions - competence, dominance, warmth, and trustworthiness. Analysis of the face models revealed that the four key traits are systematically structured by a set of latent face features including eyebrow height and chin protrusion, which can thus be used to efficiently generate social trait face features in virtual agents. We anticipate that our psychologically valid generative model of social face signals will enhance the social signaling capabilities of virtual agents.

Equipping virtual agents with attractive face can facilitate human-machine interactions. Given the real-world diversity of human's aesthetic preference and the complexity of 3D face information that comprises high-dimensional shape and complexion variations, synthesizing attractive faces for virtual human is particularly challenging. Here, we addressed this challenge using a psychological data-driven approach that reverse engineers the attractive features perceived by human participants from the randomly generated faces that capture the nature distribution of multivariate 3D shape and 2D complexion variations in human population, separately for Western European and East Asian cultures. Our analysis revealed a set of features that can increase the attractiveness of faces in each culture, including their cultural idiosyncrasies. These human and culturally validated features can thus be synthesized on virtual agents to enhance their utility in practical applications.

Social robots are now part of human society, destined for schools, hospitals, and homes to perform a variety of tasks. To engage their human users, social robots must be equipped with the essential social skill of facial expression communication. Yet, even state-of-the-art social robots are limited in this ability because they often rely on a restricted set of facial expressions derived from theory with well-known limitations such as lacking naturalistic dynamics. With no agreed methodology to objectively engineer a broader variance of more psychologically impactful facial expressions into the social robots' repertoire, human-robot interactions remain restricted. Here, we address this generic challenge with new methodologies that can reverse-engineer dynamic facial expressions into a social robot head. Our data-driven, user-centered approach, which combines human perception with psychophysical methods, produced highly recognizable and human-like dynamic facial expressions of the six classic emotions that generally outperformed state-of-art social robot facial expressions. Our data demonstrates the feasibility of our method applied to social robotics and highlights the benefits of using a data-driven approach that puts human users as central to deriving facial expressions for social robots. We also discuss future work to reverse-engineer a wider range of socially relevant facial expressions including conversational messages (e.g., interest, confusion) and personality traits (e.g., trustworthiness, attractiveness). Together, our results highlight the key role that psychology must continue to play in the design of social robots.