Blog

CREATe at the Data Science and AI Festival: Workshop on Artificial Intelligence and Interactive Worlds

Posted on    by
Blog

CREATe at the Data Science and AI Festival: Workshop on Artificial Intelligence and Interactive Worlds

By 1 December 2025No Comments

As part of the Data Science and AI Festival (27–29 October 2025), CREATe hosted a series of events at the Advanced Research Centre (ARC) exploring AI’s role in law, creative practice, and regulation. Through interactive workshops, talks, and debates, researchers, policymakers, and practitioners critically engaged with the challenges and possibilities AI presents in the digital world. This blog post is the first in a series highlighting key insights from the festival.

Blog post by Junxiang Feng and Ruiyu Xu, PhD students at CREATe. 

On 28th October, CREATe held a workshop on Artificial Intelligence and Interactive Worlds at the Advanced Research Centre.

Led by Prof. Kristofer Erickson, Felix Rodriguez Perez, and Dr. Amy Thomas, the workshop combined live demonstrations and interactive discussion to explore how AI enables more diverse and flexible forms of interaction in the creative industries. From copyright and text and data mining to agent-based storytelling and creative labour, the workshop highlighted both the innovative potential and legal complexities of AI-driven interactive worlds.

Screenshot of Noodle Bar with some character dialogue

Prof. Kristofer Erickson introduced a Stranger Things fan game he built using large language models (LLM) to illustrate potential copyright issues when using AI for game development, particularly in the text and data mining (TDM) process. This prototype is a text adventure that lets players interact with the Stranger Things universe through retrieval-augmented generation (RAG). Prof. Erickson pointed out that generative AI has pushed the marginal cost of creation close to zero. And as fans’ desire for more content, for seeing their own interests reflected in stories, and for deeper interaction with original characters, may lead to AI becoming increasingly involved in fandom. He demonstrated to attendees how the LLM-based Stranger Things text adventure fan game prototype helps build interactive virtual worlds through RAG, a real-time AI framework that retrieves information on demand from an external knowledge base. Prof. Erickson concluded his presentation by suggesting possible legal uncertainties implicated by his prototype, particularly for commercialization of AI-based interactive worlds. First, when the data used in the LLM comes from third parties, the contractual relationships and database rights associated with those third parties may need to be considered. Second, the RAG retrieves data in real time, unlike other AI training methods that permanently embeds data into model parameters. As a result, since the many TDM exceptions apply to non-repetitive analytical uses rather than data retrieval, RAG may not benefit from this exception, making it legally precarious. Furthermore, it is difficult to follow transparency requirements if the data applied by the agent AI model is retrieved in real-time, as that data is continually updating. Finally, whether the TDM exception for large language models also applies to locally run open-source models is debatable, and the exception may be used to disguise the creation of game clones.

Mr. Felix Rodriguez Perez presented Cyberpunk Noodle Bar, an LLM-driven interactive narrative game he developed alongside Prof. Kristofer Erickson. This prototype is a cyberpunk social simulator where characters pursue evolving goals, respond to news, and form dynamic relationships while the player manages a noodle bar serving as the narrative hub. He explained the game’s design concept and character loops, and highlighted the technical challenges encountered during development. Regarding the future development of AI applications in games and other creative industries, the presenters identified the following challenges. First, mainstream game developers remain cautious about using AI in game production, largely due to the coherence of narrative content, the uncontrollable risk of generating offensive material, players’ moral objections to AI-driven games., hardware burdens, and business models (e.g. the difficulties of pricing the game or maintaining stable revenue streams through DLC or sequels, expenses and liability risks arising from offensive or infringing AI outputs, and possible commercial uncertainties associated with the use of new licences for copyrights and data training Second, the hardware required to create and run AI-based games like Cyberpunk Noodle Bar (which uses an Nvidia 5090 32GB) is expected to become more common among high-end system consumers in the coming years. This means that fan creators and indie developers will also be able to run more capable LLMs locally in the future. However, using generative AI in this way may cause game developers to lose control over narrative pacing and the overall structure of the gameplay experience. The outcomes triggered by player actions also become more unpredictable. The presenters also noted that future applications of AI in the creative industries may raise issues relating to product liability, IP slippage (i.e., AI outputs including recognisable elements like character names or settings deriving from third-party sources, which are protected by copyright or trade mark), and the use in agent-based modelling (a computational model sets goals and behavioural rules for AI agents, allowing them to simulate how individuals and groups interact and make decisions within complex systems.).

Dr. Amy Thomas led the workshop’s final session, in which participants shared their plans for future collaboration. Attendees actively shared project ideas and progress related to AI and interactive worlds. In the final discussion, participants explored three themes underpinning AI-based interactive worlds. First, under AI and creative labour, attendees considered how archives and cultural institutions might document AI assisted creativity to support authorship and accountability in game development. Attendees also noted that, when AI relies on generic materials to populate an interactive world, rather than branded materials which more apparently implicate IP protection, its market performance in the context of the “attention economy” can be diminished: this can be examined through both the labour and time savings that AI provides, against the need for artistic expression and communicative functions. Other attendees pointed out that, in the process of using AI, human creativity expressed through rule-setting and parameter design can become diluted or ignored, as these forms of labour often fall outside traditional IP protection. Second, the theme of responsible AI centred on biases in training datasets, creators’ ability to opt out of AI training, and growing transparency obligations. Participants raised questions about how developers might record AI operations, ensure accountability, and protect minors in platform-like environments. For example, in a platform ecosystem like Roblox, where player-created games form the core content, AI-driven games may generate offensive material that can harm users. Recording when and how such outputs occur is essential for clarifying the respective rights and responsibilities of creators, the platform, and players. The final theme turned to users, including how to equip students to critically engage with AI mediated worlds, and how these environments affect social interaction, marginalised groups, and risks such as dopamine driven design and addiction.

In conclusion, the workshop shed light on both transformative promise and inherent contradictions of AI in creative industries. While Prof. Erickson and Mr. Perez demonstrated how LLMs can reduce the cost of participation and expand storytelling possibilities, their work also exposed unresolved issues around copyright, transparency, and creative control. It is a tricky balance. We are seeing fans and indie developers gain new powers, yet these gains force us to rethink what “authorship” and “artistic integrity” really mean. During the final discussion, with Dr. Thomas guiding the session, participants pointed out that the challenges associated with the use of AI in interactive worlds arise across multiple disciplines. Beyond legal questions, the discussions pointed to larger moral and cultural responsibilities. When the line between a human creator and a software tool is blurred by AI in an interactive world, who takes responsibility? How do we keep young users safe and minimise bias without crushing human creativity? Ultimately, the future of interactive media depends on finding a middle ground between open innovation and responsibility Technical experts, legal scholars, and cultural institutions should work together, to ensure that emerging tools support, rather than erode, diverse creative ecosystems. Future work will continue through CREATe’s ongoing projects on AI-based interactive worlds. We thank all speakers and attendees for their contributions, and look forward to continuing this conversation.