A research guide in association with the Make Believe exhibition. This guide aims to keep you informed about misinformation, providing tips and tricks to help you with your own navigation on the high seas of information.
To perform research effectively, it is important to be able to find, evaluate and use information from sources that are credible and legitimate.
Credible sources should provide information that is factual and unbiased, and usually are written by people considered to be experts in their field.
Some indications that a source is credible are currency, relevance, authority, accuracy, and purpose.
The following questions can help us understand whether a source can be considered as credible:
When trying to find information resources, some searches may generate up to thousands of results. Filtering is a good way to ensure that only the most relevant results are shown.
You can filter your searches to ensure that what you see if relevant to your research.
On the right-hand side of the results pane, there is a list of filters that you can refine your results by.
Knowing when an information resource is published or produced is a key aspect of evaluating sources. The date an article or monograph was published tells you how up-to-date it is or whether it is contemporary with the topic you are researching.
Indicators of the currency of the information to look out for can include:
Using the filters on the right-hand side of the results pane, you can filter by date a couple of different ways. One way is to select a range of dates under "Creation date". This will filter the results to only those published within the selected time period.
Another way is to use the "Sort by > Newest date" option to list most recent results first:
The kind of research you are doing will have an impact on the kinds of resources you would want to look at. If you are looking at specific aspects of a topic, general books on the subject may be too broad and surface-level. Conversely, detailed articles about specific aspects of a subject may be too complex if you are only starting to learn about a particular subject.
You can use the filters to specify the kinds of results you are getting. Under "Resource Type", you can select the types of resource you would like to look at:
Another way of determining the relevancy of your results is by using the filters under "Subject" to select only those subjects that align with your research interests.
The source of an information resource can be an important factor in assessing the authority of the content presented within. Who has published or otherwise produced this information? If the subject is climate change and the author has also founded a fossil fuel think-tank, then it's highly possible that the publication could be biased.
Screen grab from library catalogue
Screen grab from Wikipedia
"Peer-review" is a means of reviewing academic articles prior to their publication, by journal editors and academics in the same field of study, to ascertain the relevance and credibility of a given articles' content. This process ensures that articles who have undergone this process are more readily trustworthy sources of accurate and authoritative information.
You can select the "Peer-reviewed journals" filter under "Availability" :
What is the purpose or goal of the information you are looking at? Is it trying to inform, or teach you? Or is it trying to sell, entertain or persuade you of something? Are the authors' intentions clear? Does the information appear to be factual, or it is an opinion piece - or even propaganda? Does the information read as objective and impartial, or is there a clear subjective or personal perspective from the author?
If you are performing a search using the "articles & more" dropdown filter, when selecting an item from the list of results, you can read the summary (for academic articles, these are usually referred to as an abstract). This will tell you more about the purpose and aims of the article, any research methodologies used, and a summary of the conclusions reached. This can help determine whether the article is useful to you.
One trend that has become prevalent recently is the fact that AI platforms like ChatGPT have been creating citations to articles that don't actually exist - leading to some students being caught out when visiting their library to find copies of these invented, or to use machine-learning parlance, "hallucinated", articles.
Multiple studies have shown that OpenAI's GPT-3 and GPT-4 have generated citations with either incorrect or nonexistent digital object identifiers (DOIs), with one study demonstrating that nearly 40% of the citations generated by the AI were either compromised or simply don't exist.
Large Language Models (LLMs) like ChatGPT work by ingesting vast quantities of textual data from the public web, creating connections and inferences using algorithms called "neural networks", so that the model will respond effectively to input from users. What this means, however, is that the algorithms tend to compress the data, in order to provide near-instantaneous response - similar to the way that lossy formats like JPG or MP3 work by excluding redundant information. As a result of this compression, the LLM fills the gaps with information that is probabilistically relevant to the query - in this case, journal articles that seem to be authentic, but are often constructed from existing citations.
For an example of this, I asked ChatGPT for academic articles about the "learning opportunities presented by AI", and, out of 8 results, only 1 title was a match (to a book) but the rest of the citation information was incorrect. Others, such as this citation below, were concocted entirely by ChatGPT:
"The Impact of AI on Education: Adaptive Learning and Personalized Instruction"
Authors: L. Luckin et al.
Published In: Journal of Learning Analytics, 2018
This study explores how AI-driven adaptive learning technologies can enhance personalized instruction and learner autonomy. It includes practical case studies from K-12 and higher education.
It's worth noting that this isn't even an entire or correct citation - the "et al." would be used in-text but the rest of an article's authors would be listed in a correctly formatted bibliography, as well as the issue of the journal and pagination. It can be argued that asking an AI tool for academic references is a recipe for bad learning habits at best, and academic misconduct at worst!
As Hillier notes in the article linked above, the public web is unfiltered so the data ingested by an LLM like ChatGPT "likely contains a decent dose of mis-information and myths in addition to factual data." Like with any source you are looking to cite, it is important to independently verify and fact-check the information you are obtaining. If you require assistance with either finding credible sources or citing them, please do not hesitate to Ask A Librarian!
Bhattacharyya, M., Miller, V. M., Bhattacharyya, D., & Miller, L. E. (2023). High rates of fabricated and inaccurate references in ChatGPT-generated medical content. Cureus, 15(5), e39238. https://doi.org/10.7759/cureus.39238
Hillier, M. (2023, February 20). Why does ChatGPT generate fake references? Teche. https://teche.mq.edu.au/2023/02/why-does-chatgpt-generate-fake-references/
Wolfgang Sievers, Card catalogue, Baillieu Library, University of Melbourne, H2001.40/33
Online databases are a great source for peer-reviewed articles from academic journals. State Library Victoria subscribes to multiple online databases that you can access through the catalogue with a free State Library membership.
Follow the link below for the complete listing of our subscribed databases.
Catalogue screen accessed by the library patrons, State Library of Victoria. H2017.283/9
Remember, when searching a library catalogue, you are searching words in the title, author, subject heading, or occasionally other keywords or notes attached to the item - not the full text content.
If you look at the details of a relevant result, you can use one of the subject headings attached to it to determine better subject terms to conduct a further search on.
For example, fake news is an authorised subject heading, but including broader terms (disinformation or hoaxes), and related terms (journalism) in your searches can increase the amount of relevant results returned.
You can also use Boolean operators such as AND, OR, NOT when searching, e.g. journalism AND reliability (inter-related topics), coronavirus OR COVID-19 (broaden search), cats NOT kittens (narrow search).
