Disciplinary Legitimation in Open-Source Code Challenges for Archaeology

The 2025 “OpenAI to Z” challenge invited participants to use AI applications to identify archaeological ‘hotspots’ in the Amazon. This poster presents a meta-analysis of those submissions using NLP methods to identify recurring themes, conceptual framings, and technical priorities across hundreds of entries. Preliminary results show that participants repeatedly cast deep-learning models as tools that can optimize archaeological prospecting, minimize labor, and mitigate the fragility of material remains. We show that the “OpenAI to Z” submissions provide an unusually rich vantage point for examining how archaeologists, technologists, and the public imagine AI’s authority. The corpus serves as a proxy for understanding how AI is enlisted to tackle the very challenges that define archaeology. This study helps articulate the methodological limits of “open” datasets and underscores the need for culturally informed, consent-based frameworks as AI becomes a routine partner in research.