Games represent significant arenas for testing artificial intelligence (AI) technologies. Open world games, characterised by free environments and non-linear gameplay, present the next challenge for AI due to the central role of exploration, adaptability and contextual understanding. This paper investigates the relation between AI and open world games, by analysing human interaction within these environments to guide AI development. Drawing on a methodology based on human-computer interaction, we use a talk-aloud protocol to gather gameplay data from participants engaging with the game “The Outer Worlds”. Through observation, manual and GPT-3.5 clustering, we investigate three main aspects of open world gaming: planning, decision-making, and interaction. Our findings reveal the complexity, reliance on generalization and necessity of coordination of open world games. While recent advancements in artificial general intelligence offer promising solutions to address computational challenges and improve coordination, the issue of generalization remains a significant hurdle for AI. Our research highlights the role of human experience and knowledge in interacting with open world games. By defining the challenges faced by AI and giving a central role to human game play, we lay the foundations for future developments of AI capabilities.