curiosità 27/09/2025 0 Comments

What is the impact of AI in the world of research today?

Today, the real and digital worlds are so interconnected that it is difficult to tell where one ends and the other begins, as if smartphones were an extension of our hands. Similarly, the data market has become extremely important and valuable. Text, images, videos, emails, documents, internet data, social media data, health data, and cybersecurity data are all unstructured data that generates value when extracted. Data that are potentially infinite and used in enormous quantities to train Artificial Intelligence models now widespread and accessible to almost everyone.

Artificial intelligence (AI) refers to the ability of a computer or machine to duplicate human cognitive activity, such as learning, decision-making and problem-solving. Emerging in 2012, AI is now used in a variety of real-world applications and is becoming an increasingly used tool in scientific research, particularly through Machine Learning models (ML). These, are statistical algorithms that can learn autonomously from data, generate results and patterns, and predict model behaviour. ML can be used to automate the analysis and processing of large datasets enabling the identification of complex relationships in a very short amount of time. Thanks to learning techniques such as neural networks, which mimic the structure of the human brain, it can also recognise microscopic images, making it useful for identifying celestial objects or detecting cancer cells, for example. Additionally, Natural Language Processing (NLP) techniques enable software to summarise scientific articles and deliver ongoing updates on new discoveries, accelerating the drafting and review of scientific articles and consequently reducing the time it takes for a discovery to be disseminated. Thus, automating repetitive and time-consuming tasks enables researchers to focus on more complex and creative aspects of their work.

However, AI should be used to support rather than replace human creativity and judgement, in order to continue adding value to one’s work. What is most criticized about ML models is the black box nature of many algorithms whose processes are not transparent, thus raising doubts about the validity and reproducibility of results, which are fundamental aspects in the scientific field*. Consequently, the results of an ML model may be of little use raising also issues regarding the intellectual property of the discoveries themselves. The use of AI requires conscious and critical evaluation by researchers and institutions that promote it, so that it can really contribute to scientific progress. A bit like Michela, who, given the exponential use of ML tools in data analysis, decided to teach a university course on the mathematics and the mechanisms behind so that students are aware of how they work and able to interpret the results. In fact, blindly relying on AI carries significant risks, such as confusing the AI’s knowledge of a subject with one’s own, or restricting research only to those fields that can be explored through AI itself. Another danger is what is known as the ‘illusion of objectivity’, which leads to the belief that AI systems are free from bias and hallucinations, when in fact AI is not neutral at all.

Ethical and governance considerations are fundamental to scientific research. Biases and stereotypes present in the human data used to train AI algorithms can lead to unfair results or discriminatory practices, especially when applied to areas such as health research or criminal justice.  Systematic errors can become frequent and difficult to correct. Given that discriminatory narratives still permeate public and political perception, and given the applications of AI, this too must be decolonised. For these reasons, it is essential to verify the fairness of models and establish clear guidelines for communicating results. It is equally important to create common regulations for the protection of sensitive data included in datasets and to include developing countries in global scientific research.

The future of scientific research is undeniably intertwined with AI. Decisions relating to this technology must be made in collaboration with the people who encounter operational challenges daily, as well as scientists, politicians and ethics experts.

There are many benefits and issues related to AI. For more information, we recommend reading the following articles, which address the topic in greater detail: 

 

*This also happens with generative AI models such as ChatGPT – the best known and most widely used – which provide information without citing the original sources, thus raising questions about transparency, accuracy, and data quality. In addition, AI currently consumes 1% of global energy and water demand, and this is set to increase. Just think that, on average, a query to ChatGPT requires ten times the electricity of a Google search.

Post Views: 311