Biological intelligence emerges from interactive learning, where organisms actively engage with their environments and one another to derive optimal decisions in situations where the best choices are unknown, and traditional generalization to new scenarios falls short. Personalization exemplifies such challenges. Reinforcement Learning (RL) lies at the heart of interactive learning, underpinning decision-making in biological entities from simple brain-based organisms to monkeys and humans. The same principles apply to artificial agents. However, RL manifests differently in biological and artificial systems due to two fundamental contrasts. First, the solo-social dichotomy: RL thrives in social contexts for humans and animals, yet artificial agents predominantly employ RL in isolation. Second, the interplay between RL and high-level knowledge and reasoning in biological systems gives them a significant edge, an advantage often absent in artificial counterparts. In this talk, we will explore promising directions for developing the next generation of social and intelligent RL methods, bridging these gaps to advance interactive learning for artificial cognitive agents.