Machine Learning: Redefinition of the Scientific Method - Report for the 3rd semester project

Sep 17, 2026

Uploaded by Finn

Rapid machine learning progress has prompted questions about its influence on science and the Scientific Method’s traditional structure. The report investigates whether machine learning constitutes a new mode of scientific practice or mainly operates as an enhanced instrument within existing methodology. AlphaFold 2 serves as the central case, showing strong predictive performance in biochemistry while not providing causal accounts of protein folding. Comparison to experimental and theory-driven approaches argues that ML expands science’s toolkit without removing hypothesis testing, theory building, and empirical validation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages40

Reading Time1 h 41 min

AccessFree to download

Introduction

Scientific Method

Hypothesis Procedures

Deductive and Inductive logic

Experiments

Interaction between data and theory

Limits to science’s domain

Machine Learning

What is Machine Learning

Learning Models

Deep Learning

What is Deep Learning?

Limitations and Constraints

Application of Deep Learning in Biochemistry (AlphaFold 2)

ML and Biochemistry (Case Study: AlphaFold 2)

Protein Folding

Traditional Methods of Protein Structure Prediction

Machine Learning Methods of Protein Structure Prediction

Discussion

Conclusion

Machine Learning: Redefinition of the Scientific Method? Anne Margaret Rario Lorenzen Emi Maksuti Hugo Daniel da Silva Guiomar Ilie Baciu Maxim Gutu Terezie Maherova Vlad Cojocaru Supervisor: Jens Ulrik Hansen Report for the 3rd semester project International Bachelor in Natural Sciences Autumn 2025 Abstract The rapid development of machine learning has raised multiple questions about its role and influence within the scientific domain. In particular, this discussion centers on the Scientific Method, which traditionally relies on several structural components such as hypothesis formation, logical frameworks, and theoretical explanation. Machine learning frequently operates without explicit theoretical models. This paper examines whether ML represents a new way of doing science or functions primarily as an advanced tool within the established Scientific Method. AlphaFold 2 is presented as the central case study, demonstrating how large-scale data-driven methods can achieve notable predictive success in biochemistry without offering causal explanations of protein folding. By comparing machine learning-driven approaches to traditional experimental and theory-driven methodologies, this analysis contends that machine learning broadens the methodological toolkit of science but does not eliminate the necessity for hypothesis testing, theory development, and empirical validation. The declaration of Generative AI usage Use of GAI in the project process Have you used GAI to search for information? (Yes or no) -Yes Have you used GAI to receive feedback and criticism on text? (Yes or no) -Yes Have you used GAI for proofreading? (Yes or no) -No Have you used GAI to improve the language of the text? (Yes or no) -Yes Have you used GAI to get ideas and new perspectives? (Yes or no) -No Have you used GAI to suggest a research question? (Yes or no) -No Use of GAI in the project report Have you used GAI to figure out which sections of the report should contain and what should be in each section? (Yes or no) -No Have you used GAI to write text in the report (remember to cite the source in the relevant sections of the project report)? (Yes or no) -No If yes, describe where in the project report you have used GAI to write the text. - Have you used GAI to make an abstract for the report (remember to cite the source in the relevant sections of the project report)? (Yes or no) -No Have you used GAI to find out how your empirical data should be analyzed? (Yes or no) -No If yes, describe where in the project report you have used GAI to find out how your empirical data should be analyzed and how you have used GAI here. - Have you used GAI to carry out analyses of your or other people’s empirical data or data?(Yes or no) -No If yes, describe where in the project report you have used GAI to carry out analyses of your or other people’s data. - Have you used GAI to generate figures? (Yes or no) -No Have you used GAI to generate code? (Yes or no) -No If yes, describe how you have generated with GAI and how you have used GAI here. - Have you used GAI in another way than the above? (Yes or no) -No If yes, describe where and how. - Contents 1 Introduction 6 2 Scientific Method 8 2.1 Introductory overview ................................. 8 2.2 General Principles ................................... 9 2.2.1 Hypothesis Procedures ............................. 9 2.2.2 Deductive and Inductive logic ......................... 10 2.2.3 Experiments .................................. 10 2.2.4 Interaction between data and theory ..................... 11 2.2.5 Limits to science’s domain ........................... 12 2.3 What Would It Mean to Redefine the Scientific Method? .............. 13 3 Machine Learning 14 3.1 What is Machine Learning ............................... 14 3.1.1 How does it work? ............................... 15 3.2 Learni

FAQs

Does machine learning replace the Scientific Method or extend it?

The analysis argues that machine learning broadens science’s methodological toolkit but does not eliminate the need for hypothesis testing, theory development, and empirical validation.

Why is AlphaFold 2 used as the central case study?

AlphaFold 2 is used to show how large-scale, data-driven methods can achieve predictive success in biochemistry while lacking causal explanations of protein folding.

What role do hypothesis testing and empirical validation still play with ML?

Even when ML operates without explicit theoretical models, the report contends that science still requires hypothesis testing, theory refinement, and empirical validation to remain scientifically grounded.

Sep 17, 2026

Uploaded by Finn