The Social Cost of Unverifiable AI Citations in Research Collaboration

Abstract

A plausible academic reference can waste considerable time when its title, publisher or claimed finding cannot be independently located. The difficulty is particularly visible when a system is transferred to a new setting. The central research objective is to estimate the verification burden introduced by unreliable AI-assisted citation suggestions in collaborative writing. Previous work provides context for sampling, input validation and study boundaries [3–7]. To answer the question, the study proposes a controlled educational exercise using mixed verified and intentionally inconsistent reference records. Infrastructure research helps define how technical services can constrain information access [1] and continuity [2], but it cannot establish the organisational or behavioural outcomes examined here. Under matched evaluation conditions, the protocol compares ordinary manual checking against guided DOI and publisher verification workflows. Methodological sources are used to specify transparent comparison and reporting practices [8–12]. The design prioritises precise operational definitions and a straightforward baseline before introducing more complex alternatives. Performance will be judged using time per checked claim, missed bibliographic inconsistencies and agreement between independent assessors. Interpretation will account for recognition bias, uneven database access, investigator-blinding and the ethics of simulated source material. Primary analyses would be reported by the main deployment or participant subgroup so that averaged results do not conceal systematic failures. The supplementary technical literature is used to identify plausible failure mechanisms [13–17]. No study result is assumed in advance; the analysis is designed to show both when the method helps and when it fails, with reporting context from the remaining sources [18–23].