Skip to Main Content

Evaluating information & data: Avoiding bias

Bias in research

As much as we would like to think that scientific investigations aren’t biased, this is sadly not the case since there is no way to avoid bias either in research or reporting on that research, there is only how much it is biased. Bias in research and reporting can appear as the result of a selective representation of certain data, or as the willful omission of data which contradicts the results or conclusions of the writer either being bent toward a particular ideological, political, or commercial viewpoint or against it. This is often influenced by who has provided the money for the research. A critical analysis of a source, the author’s background, and the source of funding is needed to identify evidence of bias in scientific research.

Unfortunately, identify biased sources is not the easiest task particularly when one is doing a bachelor’s or master’s degree as digging out biased research is a labor-intensive process often times left to academic researchers or journalists to uncover. With that said, here we will give you some tips on how you might spot bias in both non-academic and academic sources as well as the danger of bias in AI generated texts.

How to identify biased sources

Non-academic sources

If you are using non-academic sources, some good places to check for possible bias are the websites’ “About Us,” “Who we are,” “Colophon,” and the “Contact and/or Copyright” sections. These often times have more information about the source of funding or the organization’s mission though sometimes this can also be concealed by acronyms of the funders or colorful language as you will see in the example below.

A website’s URL might also give you a clue to the trustworthiness of the information found on that site such as: .gov for government; .edu for educational; .org for non-profit; and .com for commercial organizations. Governmental and educational websites might be more trustworthy than corporations or organizations, but this is certainly not a rule and is more of a guideline. However, even if you pay attention to all of the above, it is still not easy to find all of this information without doing some research yourself.

For example, take the “Foundation for a Smoke Free World (FSFW),” a .org, whose stated mission is to, “end smoking in this generation.”



However, if you look the “Who we are” section, they state that they are funded by “charitable gifts from PMI Global Services Inc.” Who is PMI you might ask? It is Philip Morris International, one of the largest sellers of tobacco products in the world, and their mission is to sell tobacco. You can guarantee that the research done by the “Foundation for a Smoke Free World” is highly likely to be biased in favor of the tobacco industry even if they try to convince you to the contrary.


Academic sources

For academic articles, you can look at who has provided funding for the research by checking to see if it comes from a company, research organize, university, or country. This will often times be at the end of an article under the “Acknowledgements” or “Funding” section to see who provided the money for the research as well as what role the funders had in the research being either directly or indirectly involved. See here the funding declaration from the article discussed below.


(Funding: Sacks et al (2020), 2)


For an example of academic bias, in a recent study by Sacks et al (2020), they examined the top 10 peer reviewed nutrition science journals from 2018, and they found that more than 13% of articles had connections to the food industry, mainly from the processed foods and the dietary supplement sectors. Of those articles that had a connection with the food industry, either from a direct connection or direct funding from them, 87% of these peer reviewed articles found results that were favorable to the food industry, compared with only 23% of articles with no affiliation to the food industry. So, even when using peer reviewed articles from top scientific journals one must always be aware that bias could play a strong role in the conclusions reached.


Sacks, G., Riesenberg, D., Mialon, M., Dean, S., & Cameron, A. J. (2020). The characteristics and extent of food industry involvement in peer-reviewed research articles from 10 leading nutrition-related journals in 2018. PLoS One, 15(12), e0243144.


Bias in AI

AI is a new and emerging field where you also need to be aware of biased information, particularly if you are using ChatGPT. As much as we would like to think that an AI is going to be unbiased, this is unfortunately not the case. There are several reasons why an AI will be biased, and the first is the dataset that was used to train it. In the case of ChatGPT, we know the data came from the internet, which is always a great source of unbiased and accurate information, its dataset only includes material up to 2021, and the majority of what it was trained on were English language texts. These three factors mean that vast swaths of information have been left out of ChatGPT’s dataset making it both incomplete and prone to be bias. Another problem is in how the system was trained as it didn’t learn on its own but had human moderators telling it if an answer it generated was good or not. Meaning, not only could the system be biased because of the materials it was given, but also because it had human bias fed into it by the moderators helping it to learn what was the most acceptable answer to a question but not necessarily the most accurate answer to the question.

Picking out the bias in AI generated text is not easy to do since, generally speaking, none of the above suggestions can be used for AI. The funding comes from major corporations and investors, but we don’t know what role they played in the development of the AI, and the dataset and how the AIs were trained are trade secrets. This means that there is no simple way to check for bias other than through researching each statement the AI generates. The best recommendation then is to simply be aware that AI is biased and that the information it produces also needs to be fact checked. This is a problem that OpenAI, the creator of ChatGPT, is well aware of as on their own website they state:


(, Accessed 27.11.2023)


To read about some real-world examples of ChatGPT bias, see the two articles in the suggested box off to the right.