AI Bias in Scientific Research

Introduction:

From the mid-2010s to the present day, AI has exponentially grown in popularity and is used by millions of people worldwide. People often utilize AI tools for school work, ideas, and even as a chat buddy. 51% of adults in the U.S. said they’ve used AI to look up the answer to a question. Over 1.1 billion people are expected to use AI by 2031, making it one of the fastest-adopted technologies in history. 

AI is also often used in scientific environments. For example, AI is speeding up research on complicated neurodegenerative diseases, such as Alzheimer’s disease and Parkinson’s disease, according to Steven Finkbeiner, a senior investigator at the Gladstone Institutes. It is also used in genomics to identify genetic markers that are correlated with diseases and to sequence genomes.

On the other hand, AI presents some potential risks, especially the probability of bias. Combining that with scientific research could spell a recipe for disaster; it can lead to inaccurate results. What does this indicate for the role of AI in science, and is it still good enough to be used as a scientific tool? Today’s blog covers the risk of AI bias in scientific research! But first, what exactly is AI?

What is AI?

AI, formally known as artificial intelligence, is technology that lets computers simulate human learning abilities and adapts the way it behaves based on the information it learns. AI also aims to simulate comprehension, problem-solving, decision-making, creativity, and autonomy. Some of the things it can do are automation of repetitive tasks, generating answers to questions by searching the web, and analyzing patterns to deduce trends.  

The first artificial neural network was created in 1951 by Marvin Minsky and Dean Edmunds. The Stochastic Neural Analog Reinforcement Calculator (SNARC) was an early attempt to recreate learning processes that happen in the brain, particularly through reinforcement learning. It was designed to mimic the behavior of animals that react to rewards and punishments, similar to how a dog would act with treats or not. It achieved this by simulating 40 neuron-like units, and using a network of 3000 vacuum tubes.

In the present day, AI development has come a long way. For example, Google developed MUM, or Multitask Unified Model. It improves the search experience by analyzing text, images, and videos simultaneously, letting it tackle more complex search queries. Generative AI is also on the rise this decade, and some of the most popular AI that have these machine learning models are ChatGPT and Google Gemini. 

As mentioned earlier, AI tools are used by at least one billion people worldwide. Almost all (99%) Americans report using at least one of the most common AI-powered tools every week: weather apps, streaming services, online shopping, social media, virtual assistants, and GPS. These statistics clearly show just how much AI is integrated into our daily lives. Some of us might not even know we’re using it! As much as AI is used in day-to-day life, it’s also used prominently in scientific research.

AI in Scientific Research

AI is commonly used in scientific research to identify patterns, make predictions based on trends, and discover new knowledge. For example, tools like Zotero, Mendeley, and EndNote are used for organizing scientific literature, research materials, and generating citations. This is essential for efficient research, as it speeds up academic writing. AI is also used in research planning, where it outlines effective research plans and constructs a proper study design.

In the MIT School of Science, James DiCarlo, Peter de Florez Professor of Neuroscience, is using machine learning to simulate how humans synthesize information and images in the mind. They’re creating artificial visual networks that can replicate the brain’s structure, so they can produce AI tools that can help AI do visual tasks. AI is also being applied in areas that typically use computational methods.

Noelle Selin, a professor in the Institute for Data, Systems, and Society and the Department of Earth, Atmospheric, and Planetary Sciences, models air pollution trends and environmental issues to better simulate how such environmental threats could affect Earth. AI helps in noticing patterns and trends, which can be handy for scientists like Selin to make more accurate projections about the current climatic conditions.

AI can benefit scientists greatly by automating tasks, detecting patterns, and helping make accurate predictions. But what about the disadvantages? What are the risks of AI used in scientific research, especially with the risk of bias?

AI Bias in Science

AI Bias is when machine learning algorithms portray biased results, due to human biases, which can distort the accuracy of the data that the machine learning model is using to learn. This can lead to inaccurate results and harmful consequences. AI bias can happen at any step of the scientific research process.

Bias is especially prominent in the data collection process. If the data given to the machine learning model is biased in nature and not diverse, it will take that data and produce biased outputs. For example, using data that favors a population, gender, etc, to train a machine learning model will lead to the model illustrating the data. An example of AI bias was associated with racism in US healthcare. In October 2019, researchers discovered that an algorithm used on more than 200 million people in US hospitals to predict which patients would need more medical attention, heavily favored white patients over black patients.

This type of bias is dangerous, especially when handling something as important as healthcare. AI bias continues to ravage scientific research experiments and observations, but scientists are trying to mitigate that risk by training the models on diverse and representative datasets, using bias detection tools such as Trinka AI and IBM AI Fairness 360, and monitoring the output that the models give out, to ensure that the data produced is fair and unbiased.

Final Thoughts

While AI has many benefits in science, such as handling repetitive tasks, detecting trends, and making predictions based on those trends, it can have its fair share of risks, especially the risk of biased and prejudiced results. However, with meticulous monitoring of the data AI produces and security implementations in place, AI can make things highly efficient in scientific research, benefiting both the scientific community and society.

References

Albert. (2025, May 30). AI Statistics In 2025: Key Trends And Usage Data. Digital Silk. https://www.digitalsilk.com/digital-trends/ai-statistics/

Campbell, L. (2024, December 19). Researchers Explore Mutual Benefits of AI and Science. MIT School of Science. https://science.mit.edu/researchers-explore-mutual-benefits-of-ai-and-science/

Datatron. (2021, February 25). Real-life Examples of Discriminating Artificial Intelligence. Datatron. https://datatron.com/real-life-examples-of-discriminating-artificial-intelligence/

Frueh, S. (2023, November 6). How AI Is Shaping Scientific Discovery. Nationalacademies.org. https://www.nationalacademies.org/news/2023/11/how-ai-is-shaping-scientific-discovery

Holdsworth, J. (2023, December 22). What is AI Bias? Ibm. https://www.ibm.com/think/topics/ai-bias

Khalifa, M., & Albadawy, M. (2024). Using Artificial Intelligence in Academic Writing and Research: an Essential Productivity Tool. Computer Methods and Programs in Biomedicine Update, 5(1), 100145. https://doi.org/10.1016/j.cmpbup.2024.100145

Mucci, T. (2024, October 21). History of artificial intelligence. Ibm.com; IBM. https://www.ibm.com/think/topics/history-of-artificial-intelligence

Stryker, C., & Kavlakoglu, E. (2024, August 9). What is artificial intelligence (AI)? IBM. https://www.ibm.com/think/topics/artificial-intelligence

Team, Gptz. (2025, March 7). How Many People Use AI in 2025? AI Detection Resources | GPTZero. https://gptzero.me/news/how-many-people-use-ai/

Wood, T. (2024, December). AI in science and research | Fast Data Science. Fast Data Science; AI in science and research | Fast Data Science. https://fastdatascience.com/ai-in-research/