The Risks and Benefits of AI for High School Students Doing Research

Artificial intelligence (AI) has gone from being something new to something we see almost every day in research. For students, especially those in high school, learning how to use AI in the right way is important. The rise of AI in high school research is changing how students approach their high school research paper, background reading, writing, and data analysis.

While professional researchers use AI to save time on things like analyzing data, checking papers, and writing, students usually use it for smaller tasks such as making summaries, fixing grammar, or looking into related subjects.

These uses can be very helpful if AI is treated as a support tool, but it can cause problems if students depend on it too much.

In this blog, we'll look at the most common ways high school students can use AI to their benefits and the inherent risks.

The Importance of Reviewing Academic Work

Many academic publishers have implemented AI tools to identify reviewers, check the consistency of manuscripts, summarize findings, verify compliance, and detect plagiarism. Some publishers are even testing AI systems to assess manuscript quality, though final editorial decisions are still made by humans.

In academic settings, it should be assumed that submitted research paper will be reviewed by AI tools for plagiarism and compliance. Since these systems are often more efficient than humans at detecting issues, students are encouraged to conduct preliminary checks before submission.

Submitting work to plagiarism detection tools such as iThenticate can help identify unintentional similarities. Even if plagiarism is accidental, detection by automated systems can negatively affect publication chances. Additionally, it can be useful to input a manuscript (excluding the abstract) into an AI model such as ChatGPT or Gemini and request a summary. Comparing the generated summary with the original abstract helps ensure that the paper accurately conveys its intended message to reviewers.

AI for Background Research

It is often the case that researchers end up encountering other areas of research, for which they need only a passing knowledge. For example, when selecting a flask for mixing, multiple types of glass and plastic can be used.

The bench researcher doesn't need to know everything about the origins, chemical makeup, or even an exhaustive list of specification comparisons for every material. Often, a superficial understanding of the differences is sufficient to choose a flask. In these cases, AI web searching is tremendously valuable.

In a survey conducted by Nature, many experienced researchers felt that the introduction of web searching and online journal sources with AI analysis had made it easier to stay on top of these adjacent fields. In fact, sometimes the search for adjacent knowledge can be a significant time sink. AI can play a helpful role in making these, hopefully, less important research decisions.

AI for Technical Writing

Perhaps the most powerful use of AI for the researcher is in the writing task, especially for novice writers and writers for whom English is not their first language. Many free and low-cost AI tools can help with simple grammar improvements (Grammarly) or complete rewrites of paragraphs (ChatGPT, Gemini, Microsoft Copilot, scite.AI, Wordtune, Quillbot).

One of the major downsides to these tools is that the writing is often terrible. Even if carefully prompted to provide writing that is appropriate for a scientific or technical document, the AI often produces prose that is excessively wordy and overloaded with adjectives.

More dangerously, AI can often invent content to fill in what it perceives to be gaps in the supplied knowledge. Even the AI tools that are claiming to only advise on spelling and grammar are frequently wrong.

Students can safely use AI in their research paper to generate a first draft of an abstract or conclusion section. The tools can also create paragraphs from rough notes for the discussion and introduction sections. But, it is not possible to use AI to generate any part of a technical report or publication from scratch. And, it is not possible to use AI writing without heavily editing the product by hand.

AI for Language Translation

AI tools for language translation have become cheap, abundant, and powerful. Platforms such as Google Translate, DeepL Translate, Microsoft Translator, Amazon Translate, Yandex Translate, ChatGPT, Unbabel, and Wordly can make aspects of conducting research that used to be tedious or even impossible, quick and efficient.

For example, it is now reasonable to attend a conference in a foreign language and expect to be able to benefit from the conference using AI-driven language translation tools on a smartphone. This includes reading signs, reading conference proceedings, and networking with attendees. While the translations are not currently accurate for technical details, some slang, and all languages, these do not impede their use for motivated scientific communication.

Reading scientific journals and reports in foreign languages is also quite reasonable, using AI translation tools. However, those doing research in areas that are heavily dependent on subtle or archaic communication, such as free-response surveys, live interviews, interpretations of poetry, diaries or ancient languages, should not depend on AI language tools.

While there is concern that the tools will make mistakes in translation, the risk of bias, hallucination, or misinterpretation when using AI to translate scientific documents or conversations is not much greater than the risk of bias when using a friend to translate the same document.

One area of current weakness for most low-cost and free AI language tools is simultaneous (live) interpretation of a scientific presentation. The information presented often uses terms that do not appear in AIs' language library, and continuous, live language interpretation is spotty in large lecture halls.

Despite the limitations of AI for translation, it is a powerful tool that a high school student can confidently use to travel, work, and learn in a foreign language at distant conferences or in the library with foreign language scientific papers or websites. Only certain researchers in some areas need to exercise caution.

Other Applications of AI in High School Research

AI can also handle many other tasks for scientists and student researchers. One of the most useful is writing software to manage and change data. Simple edits, such as switching columns, can be done using in-app and macro tools. But more complex jobs, like removing text from the middle of entries, translating or rearranging data, or making changes based on certain conditions, can be difficult. An AI tool can automate these tasks in seconds.

AI tools like GitHub Copilot, Cursor, Windsurf, Visual Studio Code and PyCharm (and many others) can write software faster, more accurately, and more concisely than a human. No matter what programming language is being used, including macros in popular document and data editors, there is an AI that can simplify the researcher's life.

A similar tedious task is reformatting references. While most researchers are using reference management tools like Zotero, Mendeley, EndNote, or RefWorks (some of which have their own AI capabilities), it is common to occasionally need to reformat references by hand for older papers or fragments brought together from many members of the team. This task is easily handled by most AI packages.

Most researchers are using tools like NVivo, ATLAS.ti, MaxQDA, or QDA Miner to analyze their free-form survey and interview data. However, most AI tools can also handle this task, and with much less user training.

While there is no substitute for reading the interviews to deeply understand their message and meaning, AI tools can quickly count word and concept occurrences across hundreds of interviews, and AI tools can often do this for free.