AI, Artistic Style, and Intellectual Property: Legal and Ethical Boundaries in Modern Cartography
Aileen Buckley, Esri
Pete Schreiber, Esri
The capacity of artificial intelligence (AI) to replicate artistic style is a hallmark of modern advances in deep learning, machine learning, and generative AI. But it also raises concerns about copying artistic style that predate AI and have long been relevant to cartographers and other artists. At the heart of this discussion is the argument that cartography = art = protectable copyright expression. Several legal principles are central to this discussion, including the idea/expression dichotomy, the protection of artistic forms, the copyrightability of fact-based works, and the copyrightability of “style.” In a world that is being scraped, ingested, and consumed by AI, and AI increasingly automates map creation, mimics styles, and arranges facts in novel ways, understanding these legal boundaries is essential for the ethical and innovative use of AI in cartography.
Cartographic Frameworks Should Inform AI Best Practices
Vanessa Knoppke-Wetzel, Esri
In our industry we should not ignore the unprecedented impact AI on the world at multiple scales, such as environmental destruction and inaccurate code, data, maps, and visualizations.
Our existing cartographic frameworks and collective expertise can inform, push, and grow the industry and our work to ensure that if AI is part of our tools, products, software, production methodology, and/or more... we know that we are doing the most to have the least negative impact.
Join my talk to learn how to start.
I also hope this talk encourages us to come together and collectively improve AI best practices over time.
GeoAI Embracing Place? Earth Embeddings and The Pitfall of Digital Placelessness
Yue Lin, University of Illinois Urbana-Champaign
Bo Zhao, University of Washington (not presenting)
Earth embeddings, learned as numerical vectors from Earth observation (EO) data, are increasingly promoted as enabling GeoAI to embrace place by representing spatial complexity in dense, multi-dimensional form. This paper questions the promise and argues that Earth embeddings embody a design paradox: their current modes can intensify what we term digital placelessness—the reduction of places to computable surfaces that sidelines relational, social, cultural, and historical dimensions of place. Empirically, we examine Google’s Satellite Embedding dataset in the San Diego-Tijuana border region, where shared terrain and built form coexist with sharply different sovereignties, regimes of exclusion, and lived realities. We show how embedding-based similarity renders cross-border places "similar" by privileging EO-visible landscape while obscuring the political and experiential asymmetries. We further argue that embeddings add a post-visual layer of abstraction relative to satellite imagery and sensor measurements, which can weaken interpretive accountability and constrain the shared representational ground often needed for community-engaged geographic inquiry. We conclude by outlining directions for re-situating Earth embeddings through boundary-setting, plural representations, and provenance-aware evaluation, and we invite geographers to rethink what it would mean for GeoAI to embrace place without collapsing place into surface, proxy, and vector similarity.
All you need are words: LLMs and the future of prompt-driven cartography
Ian Muehlenhaus, Esri
Maps used to be expensive: technically, temporally, and institutionally. With large language models, they’ve become incredibly cheap. This talk introduces “prompt cartography,” where the cartographer shifts from map maker to map director – guiding expert assistants through iterative pipelines of data, design, and critique. Rather than replacing expertise, this shift amplifies it, demanding sharper judgment, clearer intent, and new forms of authorship and collaboration. I’ll outline emerging prompt-to-map workflows, common failure modes, and what this means in everyday practice, teaching, and for the future of the field. The goal isn’t to fear inevitable change; it's to ensure today's cartographers help direct it.
Can AI Reason with Cartographic Rules? A Framework for Evaluating LLM-Based Cartographic Decision-Making
Shiyu Zhang, University of Oregon
Generative AI is increasingly being applied to cartographic tasks, yet most current approaches treat map production as a pattern-matching task, offering little insight into whether AI decision-making aligns with cartographic principles. This talk presents a framework for evaluating LLM-based cartographic decision-making. Using the International Specification for Orienteering Maps as an application domain, the study designs a baseline cartographic agent with a traceable reasoning workflow. It then examines how different ways of organizing cartographic knowledge shape reasoning behavior and failure patterns.
