4.4 Finding Diverse Voices Using Artificial Intelligence
At the time of writing this resource, the landscape of artificial intelligence (AI) and Machine Learning is changing so rapidly that this chapter is likely already out of date. We recognize, though, that AI is changing the way we seek out and interact with new knowledge.
Many examples of epistemic injustice from Chapter 2 have their origins in historic times. The booming growth of Generative AI provides us an opportunity to critically examine how societal values influence the way knowledge is created and used in real time. We’ll briefly look at three broad types of Generative AI below and then invite you to consider how (and whether) such tools can support us with our knowledge justice work.
How they work
- General Purpose systems like ChatGPT and Google Gemini use large language models (LLMs) to assist with writing, summarizing, and general inquiry.
LLMs have learned to recognize patterns in language to predict and generate human-like text (Shah et al., 2024). These systems rely on search engines and their algorithms for generating answers: for example, ChatGPT uses Bing, and Gemini uses Google. - Domain-Specific Tools are those AI products developed for a specific audience or purpose. They often use a hybrid approach of reasoning and large-language models. Academic examples include:
a) Citation-based literature mapping services
Such as ResearchRabbit
or Connected Papers , which search the open web to find journal articles’ reference lists and recommend additional sources for you to explore.
b) Literature search tools
c) Publisher-owned products
Like Elsevier’s ScienceDirect AI (2025), which analyze full-text and reference lists of journals indexed in Elsevier databases.
- Embedded AI Assistants, like Co-Pilot, Siri, or Google Assistant, blend LLM with search engine and personalization layers. In other words, they customize their search results based on your previous behaviour, relying on search engine algorithms (Google, Bing) to retrieve information.
Can we find diverse voices using Generative AI?
While it can be tempting to start learning about new topics by asking an AI tool, we encourage you to proceed with caution. We explain our reasoning in the video below. To sum up: At this point in time, with AI tools consistently upholding Eurowestern values and understandings of knowledge, we cannot recommend using AI when searching for diverse forms of knowledge.
AI tools change rapidly, which makes videos like this difficult to keep up-to-date. Note that this video was created with ChatGPT 4.5.
So, should we avoid using AI tools entirely?
At the time of publishing this chapter, citation-based AI tools tended to produce the best, most reliable results for recommending academic literature. This is because they rely on human-curated relationships between papers (e.g. the citations and references people have put there) rather than large language models like ChatGPT (Shabanov, 2024 ).
While LLMs are not the right tool for practicing knowledge justice, they can be helpful in other contexts, especially with editing writing. Any time you make use of an AI tool, though, be mindful that it will reproduce biases that are present in its training data and original algorithm.
Stop and Reflect
In the above video, Heather (one of your chapter authors) asked her ChatGPT the following question:
- Should university students use AI tools like you to learn about knowledge justice?
While its response was somewhat impressive, Heather has trained her ChatGPT to talk about knowledge justice in certain ways, over a long period of time. Is this always the case?
Consider asking AI the same question, either using your own account (e.g. Gemini for Google users; AI Assistant built into your Smart Phone) or one of the open and free tools like ChatGPT.
- What results do you get? How does your AI response compare to the one Heather shared in the video ?
- What do the differences tell us about AI algorithms? How does this connect to our overall topic of knowledge justice?
Click the links below to access an LLM without creating an account:
All activities can also be found in a downloadable workbook. Visit the ‘Using this Resource‘ page to access the workbook in MS Word and PDF formats.
A form of generative artificial intelligence (AI) designed to recognize patterns in language and generate human-like text.
