ChatGPT’s interpretation of contested memoryscapes favours the voice of current governments, capitalism and the far-right

The indexed text (section “1 Active”; no page given) reports that ChatGPT-4o’s interpretations of three contested memorial sites favoured dominant, capitalist, and government-aligned perspectives, supporting the point about privileged narratives. It does not address consumers or the products they see, and the finding is specific to ChatGPT’s heritage interpretations rather than algorithmic systems generally. (LLM, conf=0.85)

The paper’s title identifies a study of ChatGPT interpreting contested memoryscapes and favouring the voices of current governments, capitalism and the far right. That supports the point about dominant narratives, but the passages available from our copy do not establish the claim about simplifying complex cultural categories into tropes or about model-based curation generally (searched: simplified tropes, cultural categories, dominant narratives). (LLM, conf=0.78)

The passage labeled “1 Active” (page unavailable) says ChatGPT-4o’s interpretations of three memorial sites favour dominant, capitalist, and government-aligned perspectives and influence heritage narratives. That supports the claim about evaluative framing, but the available passages do not establish that it constructs category logics or address synthetic content and hallucination. (LLM, conf=0.78)

The retrieved passages contain bibliographic material and an abstract, but no full-text passage addressing whether training data carry cultural assumptions, interpretive frames, or semantic patterns. Searches for those terms and for training-dataset bias did not locate a passage supporting the attributed claim. (LLM, conf=0.70)