As a researcher, regardless of your discipline, much of your work revolves around perusing existing literature, understanding and drawing connections, and incorporating this existing information into your own work and publications. You may also be collaborating with other researchers, witnesses, experts, and others; you may wish to visualize existing data or your own insights to present your results or thoughts.
There are AI tools and applications that can support you at every part of research described above.
Very generally, AI may help you:
This section of the LibGuide will walk you through what AI can offer and where caution is appropriate, and will close with a list of AI tools for you to peruse. If you're interested in a more detailed discussion of the ethics of AI-use for researchers and teachers, we recommend UNESCO's Guidance for generative AI in education and research.
When using AI chatbots and generative AI tools, never share:
These services may store your conversations and use them for training. Treat AI interactions like public conversations - if you wouldn't post it publicly online, don't share it with an AI.
Need help determining what's safe to share? Contact the UO Libraries for guidance.
The journal Nature gathered feedback from 1,600 researchers from different fields about the ways they see AI positively and negatively affecting their research. This section will list these pros and cons and offer some more detail on generative AI specifically. You can find the article by Richard Van Noorden and Jeffrey M. Perkel here.
A few key benefits researchers voiced are:
Let's look at them in more detail.
AI translation or editing/proofreading applications can offer non-English natives better access to global science languages. They make information and publications accessible in the languages researchers are most comfortable in. They can help improve language skills while ensuring that proposals and articles meet linguistic expectations of publishers and peers. This is especially important for researchers who could not financially afford human translation or editing services, or when timelines do not leave room for human involvement.
AI technologies process faster than humans, which means that they speed up a variety of processes relevant for researchers. Administration, Literature review, data collation and interpretation, text editing and proofreading, manuscript writing, designing presentations, formatting papers and posters, collaborating are all aspects AI applications can speed up. Faster also often means cheaper. Besides, the applications can offer services that would be more expensive if performed by a human - the flipside of this is that established jobs like web/product designer or proofreader may be destabilized.
AI application's ability to summarize academic literature, and to suggest related papers, topics, or questions can help you draw connections and develop your research approach and focus. Search functions that work without key words may offer you new ways of finding literature. With note-taking aids, you may be able to ease your development process further.
AI applications can take on some laborious and repetitive coding tasks, which saves time and money. This could enable you, for example, to design your website in fast and simple steps.
Despite all the potential benefits explained above, generative AI is not an unproblematic technology for researchers. A few key concerns researchers voiced are:
Let's look at them in more detail.
AI applications may provide information that is incorrect or has indirectly been manipulated through the machine learning of the application. In particular, generative AI applications like ChatGPT have been found to state false events or false citations - this may also be due to incorrect information that was part of the data used in machine learning. AI applications may not (accurately) cite their sources, meaning that you may be plagiarizing if you use their results.
AI applications can process large amounts of data in short amounts of time. Yet overly relying on the pattern recognition of AI technologies can make us blind to the blind spots of the application itself. Especially for researchers who have little experience with pattern recognition methodologies, it's advisable to consult with librarians or other experts to find or interpret results.
Human biases are deeply embedded in AI applications and the possible results they can yield. Relying exclusively on AI applications to find data on which to base your research or using them without reflection in processing your work can reproduce these biases in your research in seemingly covert ways.
Think, for example, about an application that draws on visuals and is able to find images that show humans. If the application's "understanding" of what a human is is based on exclusively light-skinned examples, the application is unlikely to identify humans with different skin tones as humans, excluding large sections of the human population. Basing your research exclusively on these results would therefore mean you are also excluding these sections, and you may not even be aware of it.
Some AI applications are free - or, rather, free of immediate monetary cost. This usually means that their policies grant them access to your data and can repurpose them for various purposes (such as feeding it into machine learning, selling it, or using it for personalized ads).
If they are not "free," they can quickly exceed an individual's budget. Researchers may depend on their institutions paying for access for a specific AI application. Yet not all institutions have the same financial means, so researchers' abilities to use AI applications will differ considerably.
AI requires large data centers, which, in turn, require large amounts of water for cooling. The energy required to power the centers often considerably increases carbon emissions.The memory required to store the data we produce needs material bases, too, and is rapidly depleting the planet's silicon supply.
For reference, ChatGPT can easily use the energy in a week that the average U.S. household uses in a year. It is worth considering this aspect when deciding whether or not and when to use AI applications, as well as which companies and their corporate practices to support.
With AI applications, users often cannot understand or trace how an application reaches certain results, i.e., which data and which algorithm were used to train the application. Furthermore, it can be difficult to cite exactly how you interacted with the application to obtain the results you're including in your research. This means it can be challenging to reproduce your results.
ChatGPT, for example, does not tell you where it found the results it's presenting to you (it may have made them up).
If you're working on a project with the aim to publish, you should be mindful of the guidelines and restrictions for AI use set out by the publisher. This also applies when you're reviewing another scholar's work. The publisher's AI policy will be concerned with, among others, confidentiality, academic integrity, and data security. It may tell you:
Checking these guidelines early in your writing process will ensure that your proposals are in accordance with the publisher's policies and can help you avoid drastic retrospective changes.
This section will highlight a few specific examples of how AI tools may be used in the common process of gathering and writing up data, proofreading, or peer-reviewing a project.
Most research projects require you to review literature to establish the conversation you're joining, as well as to include information, definitions, and arguments in your own writing.
AI applications such as ChatPDF, Elicit, Semantic Scholar, and genei can:
Hence your process could look like this:
You have multiple promising papers downloaded that you think can help you answer your research questions. You feed them into the AI tool and obtain an initial summary for each. Guided by these summaries, you choose one specific chapter/article and ask a particular question about it. Perhaps you're wondering, "What are microaggressions?" You can type that exact question into the tool. Genei, for example, will then show you sections of the text that respond and relate to your question. It also allows you to then add these sections to your notes. That means that, without necessarily reading the whole paper, you were able to grasp key concepts from it. This can save you time during your literature review.
AI applications like Grammarly, ProWritingAid and QuillBo can:
Hence your process could look like this:
You have completed a first draft of a chapter/article. You feed it into the proofreading application and catch any typos, get an assessment of your tone, and suggestions to change phrases or individual synonyms for increased readability or elevated diction. You peruse the suggestions and decide whether they help you meet the tone you're looking for or not and accept them accordingly.
Visuals are often key ways to communicate our research effectively and invitingly.
AI applications such as Adobe Firefly, DALL-E or Midjourney turn your verbal descriptions into visuals. Hence, they allow you to:
Hence, your process could look like this:
You're working on a poster and it's quite text-heavy and includes dense theory. You start with using a text-oriented AI tool. You enter your script and tell the application to shorten your text while keeping the central ideas, perhaps turning parts of it into bullet points. To make the overall poster more visually appealing to the audience, and to break up long blocks of texts, you'd like to depict a key concept in a cartoon or an illustration. The catch? You're not an artist, nor an expert with image-creation software like Photoshop. You use an AI application to turn your verbal thoughts into an image which you can include in your poster.
It can be time-consuming to create intricate designs for each presentation. There are countless AI applications to create presentations. Some examples are autoslide, beautiful.ai, Gamma, and Slidesgo (it can save you further time to combine these tools with text-savvy apps like ChatGPT).
These tools can:
Hence, your process could look like this:
You have the script of what you want to say for your lecture or conference presentation. In a first step, you feed this script into a text-generative AI like ChatGPT and tell it to format your information for a slide format. For individual slides, you may choose to instruct the application to contrast and compare, or to organize the information by concepts or trends it finds.
Next, you use one of these AI tools to create the slides for you, adding your text to the slides where you want it, potentially modifying the design or layout to suit your needs. To facilitate a simple presentation, you add your notes to the presenter view and share a link to the presentation to your email, so that you can easily open it from any device with internet access.
(please note, these often are not research-specific tools)
1501 Kincaid Street
Eugene, OR 97403
P: 541-346-3053
F: 541-346-3485