We are fortunate in the 21st century that AI can help maintain diligence in scientific method.

We can train a machine learning algorithm to detect weaknesses in research proposals and publications.

This can be used by scientists, funders and publishers to check integrity at each stage of the process.

Updates will be required as new weaknesses or manipulations are uncovered.

As a starting point, you can try this prompt in a system such as Claude Opus 4.7, and train the prompt on the content in the free pdf of my book. Of course, you can add other training materials.

Try this prompt:

Scientific method is defined in the pdf in the file.

Design a multi-agent AI system for testing scientific papers for adherence to accurate scientific method, with a clear, structured flowchart. Define each agent’s role, inputs, outputs, and decision logic. Show how tasks are routed, validated, and refined through feedback loops until completion. Include failure handling, optimisation steps, and scalability so the workflow can be reliably implemented in real-world use.