# A Method(ology) to the Madness

## Tags: musings, research

Published on
« Previous post: Machine Learning Needs a Langlands … — Next post: Getting started with a Ph.D. »

Words mean things and as writers1, we should make sure that they mean the right things. Lewis Carroll expressed this marvellously:

“When I use a word,” Humpty Dumpty said, in rather a scornful tone, “it means just what I choose it to mean—neither more nor less.” “The question is,” said Alice, “whether you can make words mean so many different things.”

One of the words that seems to mean too many things is the word methodology. Many research papers in machine learnings devote an entire section to their methodology, so surely, the term must be important. Most authors use the term to refer to the collection of their methods, so you find phrases2 such as ‘Our methodology uses X’ or ‘Our framework has $n$ methodological components’. In fact, this is wrong, at least considering the original meaning of the term:

Methodology is the systematic, theoretical analysis of the methods applied to a field of study. It comprises the theoretical analysis of the body of methods and principles associated with a branch of knowledge. Typically, it encompasses concepts such as paradigm, theoretical model, phases and quantitative or qualitative techniques.

Most research papers, however, do not perform such a theoretical analysis. In fact, most research papers primarily describe a new method, a set of methods, a framework, or something similar3. This is not the same as a methodology! So, in which context would the use of the word be appropriate, then? Suppose you are writing a paper that discusses the proper use of algorithms for machine learning—when to use them, how to interpret their results, and so on. In this case, you would be writing a methodology paper. To put it differently: if you are describing and analysing how to do research in your field, you are probably working on a methodological paper. In all other cases, do not use methodology.