Future ready: Elsevier's AI in higher education newsletter

Claim Radar helps users explore scholarly assertions from multiple perspectives. A researcher, for example, might use it to test the robustness of a hypothesis or evaluate how original or controversial an idea might be. Claim Radar can also be used to identify emerging topics and to complement analytics-based tools like SciVal, helping to guide strategy or identify funding opportunities. Similarly, policymakers and corporate strategists can use the feature to understand whether scientific consensus around a topic supports the case for action.

Scientific consensus is notoriously hard to demonstrate, as shown by the extensive public discussions around the need to “follow the science” (but which science?) during the COVID-19 pandemic. An individual researcher evaluating a single claim may need to undertake multiple searches before developing a clear sense of where these ideas are situated in the scholarly landscape. Claim Radar simplifies this process, making it easy to undertake a focused literature review around any given claim and providing rapid orientation within the research ecosystem.

At that time, the practice of using a scribe to compose letters was widespread, and correspondents could choose between requesting a word-for-word transcription or offering rough suggestions that could then be developed further. In this respect, the process was like using modern AI tools, which build on user‑supplied ideas to produce more detailed written communications. In some ways, the dispute around the authenticity of Margherita’s voice is strikingly similar to the kinds of conversations tutors are currently having with students about the role that AI tools play in their work.

The legacy of the scribes

Before the printing press spread in Europe, literacy was synonymous with power. From ancient Egypt and Mesopotamia onward, professional scribes often formed a prestigious social class, usually exempted from taxes and manual labor. By recording events, writing messages, or mastering specialized legal or religious vocabularies, they could wield the power of language itself, shaping narratives in a way that would be familiar to today’s politicians or advertising agencies.

In academia, the natural heirs of the scribes are arguably researchers in the humanities, where natural language is both the subject of study and the primary medium of expression, although social scientists also examine the role of words in the development of social structures and individual identities. In educational settings, the goal has been to cultivate critical thinking, build ethical reasoning and prepare individuals to both navigate and enhance society. Intellectually, there has been a preoccupation with ensuring cultural continuity and, more elusively, understanding what it means to be human.

Pushing AI off a cliff

With this in mind, it is perhaps unsurprising that the encroachment of non-human AI technologies in these domains should be so traumatic. A recent newspaper interview with humanities researchers highlighted this feeling with the memorably blunt headline: “I wish I could push ChatGPT off a cliff.” While the source of this remark remains tactfully anonymous, other interviewees such as Dora Zhang, Associate Professor in the Department of English at the University of California, Berkeley, are more nuanced and eloquent, but no less grave in their appraisal of the situation: “I now talk about AI with my students not under the framework of cheating or academic honesty but in terms that are frankly existential.”

This crisis is driven largely by the way some students are thought to be substituting AI outputs for independent critical thought – a process of cognitive offloading – ultimately casting doubt on the wider value of a university education. Other concerns include an unwillingness to concede the power of language to machines – thereby leaving governments and corporations to act in a criticism-free vacuum – but there are also deeper anxieties around a perceived loss of self.

Strikingly, the situation across campus in the Science and Technology faculties is very different. While natural language is still used in the physical sciences to frame journal articles or define ontologies, these fields also emphasize objective and non-verbal communication methods such as mathematics, data and visual models. Although English remains the main language of scientific communication, mathematics has become the language of scientific truth following the decline of Latin from the 17^th century onwards. For science and technology (S&T) researchers, AI tools tend to be a means to an end – a way of accelerating literature reviews or helping with the writing that showcases a set of findings, often without impacting the findings themselves. While there are still caveats around hallucinations, potential bias, the traceability of sources etc., there is a sense that these risks are far outweighed by the opportunities that AI presents.

A house divided

In short, the advent of mainstream AI has deepened the existing rift between the two cultures of science and the humanities, adding to the list of more familiar differences like research behavior and values, funding levels, career outcomes and gender representation. Is this a new digital divide?

Of course, AI has already exacerbated an existing divide based on unequal access to digital technology, often resulting from a lack of resources, education or connectivity. Indeed, some AI-sceptic humanities researchers have begun to advance a variation on this theme whereby only wealthy students will one day benefit from a largely tech-free liberal arts education. However, from the university’s perspective, the curricular discrepancy between the sciences and the arts may be almost as fundamental as the access-based division, especially for large institutions that derive their historical reputations (and rankings status) from excellence in both areas.

While in the twenty-first century there may seem little choice but to allow better-funded S&T subjects to subsidize embattled humanities disciplines – they often face declining enrollments, difficulties reporting societal impact and mounting public scepticism – the two cultures division cuts deep and can lead to organizational paradoxes. For example, university librarians, key players in most institutional AI transformations, usually have backgrounds outside S&T, leading to a situation where staff predisposed to be wary of AI may well be spearheading its rollout.

Rebellion or partnership?

Of course, this is often a very good thing – the pragmatism and flexibility of librarians has strengthened innumerable university AI adoptions over the last two years. Meanwhile, humanities-based researchers are among the most insightful critics of the new technologies, even if their focus is predominantly on risks and limitations rather than prospective benefits.

The fact is, however, that it is far too late to dream of pushing AI off a cliff. While some arts-based researchers will continue to fight a rearguard action, perhaps reminiscent of the Romantic revolt against the first Industrial Revolution, a better option might be to work to ensure the AI used in academia is as good as it possibly can be. This means accepting the fact that not all AI systems are created equal – and that some solutions can acknowledge the importance of independent thought and creativity and work to amplify rather than stifle them.

Transparency tooling

Elsevier’s LeapSpace is one AI tool that does not set out to convince users of an “answer,” but instead provides insight into a research space, deliberately avoiding the synthesis of viewpoints into a falsely fluent response that might blur specific perspectives. The new Claim Radar feature, which highlights where sources agree or diverge on specific claims, is a prime example of the kind of transparency tooling that distinguishes responsible AI from systems that simply enable cognitive offloading. Of course, if an AI user steadfastly refuses to engage their own mind there is little that can be done – just as the most vivid prose may do little to convert a reluctant reader – but LeapSpace is designed to support and repay cognitive effort. The aim is to challenge critical thinking around a problem or field and enable more distinctly “human” qualities like creativity, empathy, moral judgement and emotional intelligence in the context of research.

What it means to be human

Interestingly, AI tools also raise the tantalizing prospect of more dynamic connections between humanities and the social and physical sciences. While the best tools are often trained on bibliographic databases where arts disciplines can be underrepresented due to a focus on journal abstracts, this situation has recently been improved by more inclusive indexing. Another encouraging trend is the rise of publishing partnerships that offer broader coverage in some AI tools (LeapSpace now covers humanities content from leading publishers like Sage and Oxford University Press). At the same time, AI offers new ways to track the societal and cultural impact of humanities and social science outputs that have not always been well represented by traditional citation-based analytics, potentially helping to address problems in an education and funding landscape that has historically prioritized quantifiable, short-term metrics.

While it is possible to see this type of evaluation as a constraint on the independence of non-S&T disciplines, the benefits of working to reunite the two cultures into a more holistic view of human knowledge could be significant. Challenged to redefine their relationship with society, universities will likely be stronger if their approach to knowledge generation embraces that society more comprehensively. At the same time, as AI technologies become more powerful, the humanities can play a critical role by placing these capabilities in context and helping to ensure they augment rather than replace human skills. Like Margherita Datini arguing with her money-making husband, they need to champion and defend an authentic human voice and protect the sense of self that lies behind it. If ever there were a time when society needed a practical concept of “what it means to be human,” then it is now.

This lack of enterprise-level coordination was a key theme of Inside Higher Education’s fourth annual Survey of Campus Chief Technology/Information Officers in 2025. Indeed, only 35% of the CTOs/CIOs then surveyed thought their institutions were handling the rise of AI adeptly, while just 11% said their institutions had a comprehensive AI strategy.

While some universities have begun to come to terms with AI, with integrations increasingly focused on the adoption of secure internal systems rather than relying solely on public or vendor AI tools, many are still grappling with the legacy of organic rollouts. For this reason,Your AI strategy checklist for leading a future ready university, a new guide to institutional AI integrations, is extremely timely.

A practical roadmap for AI progress

Focused on the needs of university leadership, this checklist offers a structured way to think through the challenges of AI adoption, providing a catalyst for discussion and reflection rather than detailed instructions. Flexible, empowering and easy to absorb, the checklist provides an eminently practical roadmap for AI progress – whether you’re starting out, reorganizing your efforts, or occupying that state of “maturity” that simply means you are well-placed to assimilate future technological innovations.

Perhaps the most concerning development, however, is the growth of mis- and disinformation, whether caused by flawed methodologies, blatantly fabricated data, or AI hallucinations.

Even while it offers significant opportunities to researchers, AI is widely believed to have accelerated the spread of misinformation, making the production of convincing fake content faster, cheaper and more accessible. Moreover, AI tools and systems are already embedded in most of our working lives – so what can we do to safeguard the accuracy, authority and transparency of research and the integrity of the processes that underpin it? This question is discussed in Misinformation at machine speed: Why trusted content matters more than ever, a new article that focuses on the issue of misinformation used within and generated by AI tools.

Avoiding the false economy of unverified sources

This distinction between what AI tools are trained on and the outputs they produce is less meaningful when we consider the well-known IT axiom, “garbage in, garbage out.” As the article makes abundantly clear, any AI tool, whether built or bought, is only as good as the quality of its underlying data. This is shown through a detailed comparison of general AI tools that pull in content from non-academic sources and dedicated scholarly offerings that are more suited to the needs of researchers. “Research-grade AI” emerges as less of a strapline than a warning to avoid the false economy of AI trained on unverified sources such as news stories or blogs.

4 tips for optimizing your LeapSpace prompts is a new spin on an old format, updating the familiar “write better prompts” guide for the age of agentic AI. While it was once critical to get prompts right first time, research-focused tools like LeapSpace increasingly give users the opportunity to shape outputs in flight by qualifying the original query, or by asking follow-up questions that elaborate on specific points. This means it is now less important to produce ultra-precise or prescriptive prompts, giving users more control over their AI responses.

6 ways LeapSpace can help you move from curiosity to discovery faster offers an engaging survey of the core LeapSpace use cases, from executing a literature review to finding a funding opportunity, and includes a tantalizing glimpse of the LeapSpace product roadmap.

• Read 4 tips for optimizing your LeapSpace prompts

• Explore 6 ways LeapSpace can help you move from curiosity to discovery faster