From Binary Groundedness to Support Relations: Towards a Reader-Centred Taxonomy for Comprehension of AI Output
2026-04-09 • Human-Computer Interaction
Human-Computer Interaction
AI summaryⓘ
The authors say that when AI uses documents to answer questions, we usually just check if the answer is supported or not, which is too simple. They suggest making a detailed system to show exactly how the AI uses sources, like whether it quotes directly or paraphrases, and how it reasons from the information. This system would help create better tests for AI and clearer ways for users to see where answers come from. They plan to build this system using ideas from language studies and philosophy, and test it with human judgment.
Generative AIGroundednessHallucinationRetrieval Augmented GenerationSyntactic MovesInterpretive MovesProvenanceLinguisticsPhilosophy of LanguageHuman Annotation
Authors
Advait Sarkar, Christian Poelitz, Viktor Kewenig
Abstract
Generative AI tools often answer questions using source documents, e.g., through retrieval augmented generation. Current groundedness and hallucination evaluations largely frame the relationship between an answer and its sources as binary (the answer is either supported or unsupported). However, this obscures both the syntactic moves (e.g., direct quotation vs. paraphrase) and the interpretive moves (e.g., induction vs. deduction) performed when models reformulate evidence into an answer. This limits both benchmarking and user-facing provenance interfaces. We propose the development of a reader-centred taxonomy of grounding as a set of support relations between generated statements and source documents. We explain how this might be synthesised from prior research in linguistics and philosophy of language, and evaluated through a benchmark and human annotation protocol. Such a framework would enable interfaces that communicate not just whether a claim is grounded, but how.