AI in Research: Benefits, Risks, and the Next Generation of Scholars

Artificial intelligence (AI) is rapidly transforming the research landscape, offering powerful tools that can both streamline and complicate the investigative process. For educators guiding the next generation of scholars, it is essential to understand not only the opportunities but also the limitations that AI presents when it comes to research and learning.

Students are increasingly turning to AI for brainstorming, information gathering, and even content creation, yet these uses can conflict with professional-level research practices that lean on accuracy, reproducibility, and methodological precision. It is important to identify the distinctions between generative and assistive AI, as there are benefits, risks, and ethical considerations that counselors will need to help students navigate. The goal is to acknowledge that AI is here to stay and can be beneficial while creating a balanced awareness of how to use AI responsibly in the academic world.

Use of AI Research in Modern-Day

It is nearly impossible to conduct any work, including research, without being exposed to a tool claiming to use AI to make things better, easier, and faster. For the most part, these tools improve the quality of the writing, create graphs and figures quickly, and improve the research workflow. But the use of AI can also reduce the chances of publishing, create errors, or even jeopardize future careers.

It is essential for high school students researching to use AI with caution. The importance of using AI in research was underscored by a survey conducted by Nature in 2023. They surveyed over 1,600 scientists to gather their opinions on AI. More than half of them felt that AI would soon be essential to conducting research. The most important application in research, they felt, was a faster way to process data, including speeding up computations, saving time or money, automating data acquisition, processing new kinds of data, and writing programming code.

How AI is Used Today: A New Generation

Today, we can see AI making strides in dozens of sectors. A great example is healthcare, where AI can comb through massive datasets to look for and identify anomalies or patterns; think of mammograms, tissue samples, drug development, and medication interactions. By utilizing AI, the healthcare sector, in particular, has been able to provide patients with faster care by identifying outliers early on.

By contrast, a recent study by the Harvard Graduate School of Education (2024) reported that teenagers use AI primarily for getting information (53%) and brainstorming (51%) but also use AI to generate pictures or images (31%), make sounds or music (16%), and to write code (15%). It is becoming an avenue for entertainment as much as education.

But young researchers may not see it this way. What the professional researchers see as valid and valuable uses of AI (processing data) and what teenagers see as valid and valuable uses of AI (getting information, brainstorming, and generating pictures and images) are not necessarily the same.

A young researcher might easily use AI in a way that is familiar to them, only to later discover that it invalidates years of their hard work! Take, for example, a researcher who used AI to compile references for their papers submitted to academic journals, only to learn later that the AI generated a reference that does not exist, therefore making all of that research suspect.

Generative vs. Assistive AI

Some AI tools are merely more powerful versions of existing tools, like spell checkers, grammar checkers, translation engines, and graph wizards. Branding these tools with AI simply makes them seem more attractive. But their role in research is unchanged. Fundamentally, these tools do not generate content; they only assist the researcher in generating quality content. These are assistive AI tools, and it helps to remember they are simply part of a broader research toolkit.

Generative AI, on the other hand, refers to computer programs that generate content. The program is provided with data, prompts, context, and instructions, typically using natural language (or large language models) and generating text, images, graphs, tables, analyses, and more.

To take a deeper look at the juxtaposition, let us consider how they may be used more specifically in a research setting. A graduate student preparing a paper on protein folding might use assistive AI to improve the structure, clarity, and grammar of their paper. The content would be her own, but AI has assisted with refining the language.

Generative AI, on the other hand, might be used by a data science researcher to analyze large sets of patient records in a hospital setting. AI would be implemented to create code that runs analyses on that dataset and potentially to produce a summary of the statistical findings.

The difference between a system that helps the researcher create content and a system that generates the content is not black-and-white. The most important difference between the two is, in fact, the expectation of the researcher.

When the AI only helps generate a graph, the researcher assumes that the graph must be checked carefully for errors. When the AI analyzes the data and generates the graph, it can be too easy to assume that the AI model has completed the work correctly and no confirmation of accuracy is needed. That assumption creates a problem.

Content generated by an AI model cannot be assumed to be unbiased, accurate, or even true. AI has been known to generate references that do not exist, analyze data using statistical techniques that do not exist, and, perhaps most importantly, completely misinterpret results. The risk of biased, inaccurate, or invented results means that researchers must be extremely cautious when using generative AI. One of the best things we can teach our students is to check their work.