eing part of academia, I have pondered upon the rise of artificial intelligence (AI) and its impact on research. I have attended various seminars and spoken with professors and research students to learn about their perspectives on the prospects of AI. There is still considerable scepticism about the use of this new technology. While some adopt it as a pragmatic tool for efficiency, many have developed an uncritical dependence. Technology is changing so rapidly that it has been challenging for anyone to keep pace with its advance. As a researcher and faculty member, I believe that we need to reflect more deeply on how researchers should work with this technology.

Evaluating student knowledge is a core task for faculty members. End-of-semester research projects are assigned to students to deepen their understanding of the subject and provide an opportunity to apply the knowledge acquired during the semester in real-world contexts. Recently, however, a troubling pattern has emerged. I frequently come across research reports that are well written with all the bells and whistles of academic citations but seriously lack depth. They are coherent in form but hollow in substance. When I ask fundamental questions about the submitted manuscript, I often receive poor responses. This has led me to consider whether the proliferation of AI tools specifically for research is beneficial or harmful.

“Research is a whole mind activity,” one of my PhD professors had told me during my early days of doctoral studies. She was a passionate researcher with a degree in agricultural economics from a leading US university. I say the same thing to my research students, but how they understand the whole mind activity is hard to tell. As AI is changing the way we work, it is also significantly impacting the conduct of research. Being part of the research fraternity for the past 20 years, I often reflect on various aspects of research and the key traits that a researcher must possess. In the post-AI era (since 2022) I have often reflected on how the core skills of a researcher are affected by its use.

Is there a need to change how future researchers are trained? How should researchers interact with AI tools for personal and societal benefits? As AI tools offer deep research, deep thinking and learning models, how should humans harness the power of Large Language Models)? The LLMs are meant to perform research tasks effectively and efficiently. Researchers in the pre-AI era had to endure significant pain to do so.

There are many traits that a researcher must possess. The most crucial ones include observation, identification and understanding of prior research; critical analysis; data collection and analysis; experimentation; and, finally, communication of research outcomes. With AI tools’ ever-increasing capabilities, it is time to analyse how these skills are affected, both positively and negatively. Humans are blessed with five senses through which they learn about their surroundings and the world. Since the emergence of smart phones with large, colorful touch screens and engaging apps for socialising and entertainment, people spend substantial time on their screens. In markets, on sidewalks and behind the wheel, heads are bowed and eyes locked onto screens.

We have become aloof from our physical reality. Few people glance up to observe the architecture, the changing sky or the chaotic interactions of daily life. This is a crisis for science because a true researcher must be a witness first. Research begins with observing nature, capturing unscripted interactions and finding intriguingly invisible connections that an algorithm cannot see. Now, with AI-enabled smart phones, the eyes are even more fixated on the screens. The connection with the real world is getting weaker by the day. Attention spans have grown shorter over the past three to five years. Many researchers and scholars are now found staring at the screens. Some of them rely on the computer vision even for their awareness of the world around them.

The next important skill is search and review of prior research or literature on the research area. AI tools for searching, reviewing and analysing literature are abundantly available. Libraries with physical books are now rarely consulted. The time spent searching for material via search engines and scrolling through links provided by large publishers (such as JSTOR, Taylor & Francis, SAGE and Emerald) is also becoming less common.

As the performance of AI search engines improves rapidly, researchers have begun mostly using those. The skill of sifting through the large bulk of offline documents is diminishing rapidly. I have seen several literature reviews by young researchers who were able to generate manuscripts in response to prompts as short as a single line. Now, AI tools claim to provide a comprehensive synthesis of literature, an activity in the pre-AI era that required a higher level of critical thinking. Reliance on rapidly evolving AI tools, driven by training, threatens to undermine critical thinking skills. Young researchers need to understand that their own brains are rewired as they review, synthesise and critique existing literature. Although there is nothing wrong with using AI tools for literature searches, synthesis should not be left to them. The brain is an intelligent machine that seeks to conserve energy; when AI tools offer an easy option for reviewing literature, the natural tendency to conserve energy compels researchers to rely on the output with minimal intellectual effort. Researchers ought to think deeply to analyse, compare and contrast prior findings, draw conclusions and carve out new paths.

Current AI models offer sophisticated reasoning and exhaustive research capabilities, promising to uncover gaps that might elude human detection. Advertisements for these tools are persuasive enough to convince many young researchers of their capabilities. As a result, their use is increasing fast. These tools can undoubtedly assist researchers, but they cannot replace the ingenuity of the human mind. Humans are blessed with the gift of creativity. Their cognitive capacity encompasses creative, analytical and critical thinking. Appropriately honed, these skills are way beyond current and near-future AI capacity. I have seen students present AI-generated research reports without even bothering to verify their soundness.

AI models are currently developed to please users. Any connections or relationships of entities proposed or dictated by the user prompt are not only appreciated by the AI but also amplified by generating a lengthy, whimsical document. This is a trap so convincing that many young researchers ignore common sense checks and fall prey to the LLMs’ so-called advanced capabilities.

The next skill at risk for researchers is data analysis. Computers are fast because they can process information much more quickly. In the pre-AI era, researchers generally collected, curated and sorted data for analysis. During the process, they picked the techniques to be used for detailed analysis and analytics. Because all methods and tools have limitations, a more careful approach was generally adopted. In the post-AI era, the tools offer flexible ways to analyse datasets and produce multiple results from the same data file. The tools can undoubtedly analyse large datasets and detect noteworthy trends or abnormalities that are difficult for humans to detect in a short time.

There are examples of AI use in cancer detection and in understanding complex protein structures. However, over-reliance on AI tools among young researchers is not a positive sign. Powerful AI-based computer vision tools detect and analyse graphical data more efficiently; however, researchers’ increased reliance on such tools without recognising the importance of their own cognitive abilities in interpreting figures and shapes is not a happy sign. The data analysis skills of young researchers are at risk.

Communicating research outcomes is essential; researchers must present their findings effectively in oral, written or graphical form. The proliferation of social media apps has added new dimensions to communication, as people increasingly use them to express their feelings and findings. At the same time, excessive use of social media has also altered how researchers communicate their scientific findings.

In the post-AI era, tools have matured to the extent that it is challenging for a non-expert to distinguish between a human-generated manuscript and an AI-generated one. Generating a full-fledged academic report or thesis has never been easy. Academic writing has a distinctive style that AI tools have mastered through rigorous training. The flip side of these capable tools is that researchers have begun using them extensively and have become so dependent on them that, in my observation, the majority no longer spend time honing their communication skills.

I have come across researchers generating schematics, diagrams, sketches, photographs and graphical abstracts - even videos - using AI tools without their own conscious input beyond a cut-and-paste prompt. Visual communication is an art in danger in the post-AI era. Because researchers are always under time pressure, academic writing tasks are increasingly delegated to AI tools. Overly decorative graphic designs in presentations and long sentences in manuscripts have become the norm.

AI is a technology that has emerged from decades of arduous research and experimentation. It is here to stay and flourish. As academics, we must ensure that young researchers are taught its proper so that they can develop their own capabilities. Here are a few ideas to preserve and develop the human capacity for research and investigation:

– Observation is important. Researchers must stay away from their digital devices for at least for some time. They must move out to observe the natural world and interact with people in manufacturing or service industries to understand processes. They should walk without their mobile phones.

– Researchers must develop a habit of reading scientific papers without the help of AI and try to select, sort and organise the literature on their own. They should use AI in a way that preserves their own judgment, enabling them to guide AI tools effectively.

– Researchers may use AI to understand the complex or difficult-to-understand ideas or concepts, but also spend time creatively analysing the facts, connections and relationships to find gaps and future directions of research. Researchers must also engage in deep thinking and cultivate their reasoning skills as they encounter diverse ideas.

– Researchers should design their data analysis workflows conceptually. They should practice more with non-AI tools, then use AI-enabled tools to compare and refine the analysis.

– AI writing tools are practical, but their overuse has started to affect the communication skills of the researchers. Effort should be made to prepare the document without the use of AI-enabled tools, and subsequently use such tools to edit the manuscript. Researchers must learn the fundamentals of graphic design to produce effective figures and diagrams. This will help them when using AI models to generate and refine their output.

AI is an undeniable asset, but a tool can never replace the architect. We must remind young scholars that the human brain, with its capacity for chaotic, creative and non-linear thought, is far more potent than any predictive model. We must protect and hone our cognitive distinctiveness. It is the one skill a machine cannot replicate.

The writer is a professor at the National University of Science and Technology, Islamabad.