A tentative taxonomy for predictive models in relation to their falsifiability

Area of Science:

• Biomedical Research
• Interdisciplinary Science
• Scientific Modeling

Background:

• Predictive models are crucial in biomedical research, spanning physics, biology, and medicine.
• Established falsification methods exist within individual disciplines but lack a unified interdisciplinary approach.
• Interdisciplinary research necessitates a broader framework for validating complex scientific models.

Purpose of the Study:

• To propose a general definition of 'scientific model' applicable to predictive models.
• To categorize predictive models and suggest generic falsification strategies.
• To identify areas requiring further research in model falsification.

Main Methods:

• Literature review of falsification methodologies across disciplines.
• Development of a generalized definition for scientific models.
• Categorization of predictive models based on the proposed definition.

Main Results:

• A generalized definition of scientific models is proposed.
• Generic falsification strategies are outlined for most model categories.
• Abductive models (e.g., neural networks, Bayesian models) require further investigation for falsification.

Conclusions:

• A unified framework for falsifying predictive models in interdisciplinary biomedical research is initiated.
• The study highlights the need for continued research into the falsification of abductive models.
• Standardizing falsification methods will enhance the reliability of biomedical predictive models.