Survey of 1,150 Chinese researchers links perceived AI usefulness to adoption plans

Researchers who believe generative AI is useful for their work are more inclined to plan to use it, according to a survey analysis published on 26 October 2026 in Humanities and Social Sciences Communications. Among 1,150 scholars and researchers in China, perceived usefulness was associated both with favourable attitudes and with stronger intentions to adopt the technology in research.[1]

The analysis distinguishes perceived practical benefit, general attitudes and intended use. The investigators analysed complete records from 1,283 responses gathered in March 2025. Their cross-sectional design captures how these measures relate at one point in time; it does not track what participants later did or whether their research improved.[1]

Usefulness mattered in two ways

The researchers drew on the Technology Acceptance Model, a framework often used to study why people say they will adopt a technology. The survey asked participants about perceived usefulness in research work, favourable or unfavourable attitudes and intentions to use generative-AI systems. The model estimated a positive relationship between perceived usefulness and attitudes, with a coefficient of 0.489. Usefulness also had a direct reported relationship with usage intention, coefficient 0.385. Attitude was positively associated with intention, coefficient 0.323. All three relationships were statistically significant in the reported analysis.[1]

The authors additionally estimated an indirect pathway from usefulness through attitudes to intentions, with an effect estimate of 0.157 and a 95% bootstrap interval from 0.120 to 0.199. That gives the hypothesised model some internal statistical support, but a cross-sectional survey cannot rule out reverse influence. Researchers who already want to use a tool may be more likely to say they find it useful; existing experiences may shape both answers. Other factors, including institutional incentives, job roles and discipline, can influence the same choices.[1]

The largest of the three direct coefficients linked usefulness with attitude. The separate usefulness–intention relationship also remained positive in the model. Together, these findings suggest that a tool's perceived research value matters both to a scholar's general view of it and to their stated plans.[1]

What adoption means depends on the kind of research

Research work contains several distinct tasks. Brainstorming terminology, finding papers, cleaning data, writing code and interpreting experimental results place different demands on a tool. A scholar may welcome assistance with routine formatting while avoiding it for statistical interpretation. That makes task-level usefulness a more informative starting point for support than a single campus-wide question about whether researchers like AI.[1]

For a university deciding what to offer, the practical implication is to connect access with a clearly specified research task. Training that demonstrates how a tool fits a real workflow can address a different need from a general awareness session. The survey did not compare those interventions, so this is a testable implication of the findings rather than an established effect of training.[1]

The survey was conducted by Kantar through an anonymous online email questionnaire in China. Its March 2025 observation period also matters: the paper appeared in October 2026, after the responses were collected. Institutional rules, disciplines and researchers' prior experience may shape both perceived usefulness and intended use, limiting how far the estimated relationships can be transferred to another setting.[1]

From positive attitudes to useful research support

The authors recommend responsible-use guidance, training and awareness programmes. Their findings give those programmes a specific question to address: which research tasks do scholars believe AI can help them perform? Answers at that level could make institutional support more responsive to the work researchers are trying to do.[1]

Report an error or suggest a correction